{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.5841311049813402, "eval_steps": 100, "global_step": 1800, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0, "step": 0, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315779162629042e-06, "weight_min": 0.0, "weight_std": 1.503326802776428e-05 }, { "epoch": 0, "step": 0, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2527732906164601e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7906268112710677e-05 }, { "epoch": 0, "step": 0, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.838504249462858e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7415990441804752e-05 }, { "epoch": 0, "step": 0, "weight_max": 4.187939703115262e-05, "weight_mean": 2.586813752714079e-05, "weight_min": 0.0, "weight_std": 1.643578252696898e-05 }, { "epoch": 0.00032451728054518905, "step": 1, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4473338524112478e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0066734577994794e-05 }, { "epoch": 0.00032451728054518905, "step": 1, "weight_max": 3.998322426923551e-05, "weight_mean": 9.576295269653201e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7752712665242143e-05 }, { "epoch": 0.00032451728054518905, "step": 1, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392191143182572e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.924755451909732e-05 }, { "epoch": 0.00032451728054518905, "step": 1, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396732831286499e-06, "weight_min": 0.0, "weight_std": 1.5261864973581396e-05 }, { "epoch": 0.0006490345610903781, "step": 2, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.762896317814011e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.008354931604117e-05 }, { "epoch": 0.0006490345610903781, "step": 2, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2906077245133929e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.822274862206541e-05 }, { "epoch": 0.0006490345610903781, "step": 2, "weight_max": 3.752332122530788e-05, "weight_mean": 7.222552994790021e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4049162018636707e-05 }, { "epoch": 0.0006490345610903781, "step": 2, "weight_max": 3.36925259034615e-05, "weight_mean": 4.2123833736695815e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1911775800399482e-05 }, { "epoch": 0.0009735518416355671, "step": 3, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.357117177278269e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2320688256295398e-05 }, { "epoch": 0.0009735518416355671, "step": 3, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0043636393675115e-05, "weight_min": 0.0, "weight_std": 1.8589451428852044e-05 }, { "epoch": 0.0009735518416355671, "step": 3, "weight_max": 3.572963396436535e-05, "weight_mean": 1.2623790098587051e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.5030902432044968e-05 }, { "epoch": 0.0009735518416355671, "step": 3, "weight_max": 4.252122016623616e-05, "weight_mean": 1.803942723199725e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9941580831073225e-05 }, { "epoch": 0.0012980691221807562, "step": 4, "weight_max": 4.252122016623616e-05, "weight_mean": 1.455882465961622e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0215029508108273e-05 }, { "epoch": 0.0012980691221807562, "step": 4, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.469768540118821e-05, "weight_min": 9.893572627817093e-10, "weight_std": 2.0206509361742064e-05 }, { "epoch": 0.0012980691221807562, "step": 4, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.17925146696507e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.586736470926553e-05 }, { "epoch": 0.0012980691221807562, "step": 4, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.677558071212843e-06, "weight_min": 0.0, "weight_std": 1.6070685887825675e-05 }, { "epoch": 0.0016225864027259451, "step": 5, "weight_max": 2.595289316786875e-09, "weight_mean": 1.0272298567315374e-09, "weight_min": 5.771870759119224e-10, "weight_std": 7.211062302836524e-10 }, { "epoch": 0.0016225864027259451, "step": 5, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393349627207499e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2424131455190945e-05 }, { "epoch": 0.0016225864027259451, "step": 5, "weight_max": 1.0015337448265882e-09, "weight_mean": 4.4179818137202176e-10, "weight_min": 0.0, "weight_std": 3.869508324871873e-10 }, { "epoch": 0.0016225864027259451, "step": 5, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844953764404636e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3697084796149284e-05 }, { "epoch": 0.0019471036832711343, "step": 6, "weight_max": 3.397988257347606e-05, "weight_mean": 7.360142262768932e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3840864994563162e-05 }, { "epoch": 0.0019471036832711343, "step": 6, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3977813068777323e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.937770502991043e-05 }, { "epoch": 0.0019471036832711343, "step": 6, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8390790501143783e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9839044398395345e-05 }, { "epoch": 0.0019471036832711343, "step": 6, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5195297237369232e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0993273210478947e-05 }, { "epoch": 0.0022716209638163234, "step": 7, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1795297723438125e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8516213458497077e-05 }, { "epoch": 0.0022716209638163234, "step": 7, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0154092706216034e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.880256058939267e-05 }, { "epoch": 0.0022716209638163234, "step": 7, "weight_max": 3.662098970380612e-05, "weight_mean": 7.877837560954504e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4838751667411998e-05 }, { "epoch": 0.0022716209638163234, "step": 7, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2117741789552383e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7591069990885444e-05 }, { "epoch": 0.0025961382443615124, "step": 8, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3102386219543405e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8223510778625496e-05 }, { "epoch": 0.0025961382443615124, "step": 8, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2650836652028374e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8602468117023818e-05 }, { "epoch": 0.0025961382443615124, "step": 8, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5880646969890222e-05, "weight_min": 0.0, "weight_std": 2.154638059437275e-05 }, { "epoch": 0.0025961382443615124, "step": 8, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4658920918009244e-05, "weight_min": 7.298507354036587e-10, "weight_std": 2.0292183762649074e-05 }, { "epoch": 0.0029206555249067013, "step": 9, "weight_max": 1.7014956029015593e-05, "weight_mean": 2.2479355266114e-06, "weight_min": 0.0, "weight_std": 5.976290140097262e-06 }, { "epoch": 0.0029206555249067013, "step": 9, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235622539767064e-05, "weight_min": 0.0, "weight_std": 1.897062611533329e-05 }, { "epoch": 0.0029206555249067013, "step": 9, "weight_max": 4.252122016623616e-05, "weight_mean": 9.893376045511104e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.838524258346297e-05 }, { "epoch": 0.0029206555249067013, "step": 9, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7172187654068694e-05, "weight_min": 1.1403916699848082e-09, "weight_std": 1.9262181012891233e-05 }, { "epoch": 0.0032451728054518902, "grad_norm": 4.582557678222656, "learning_rate": 1.0000000000000002e-06, "loss": 0.5938, "step": 10 }, { "epoch": 0.0032451728054518902, "step": 10, "weight_max": 3.752332122530788e-05, "weight_mean": 9.448885066376533e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6103245798149146e-05 }, { "epoch": 0.0032451728054518902, "step": 10, "weight_max": 4.252122016623616e-05, "weight_mean": 1.307422553509241e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8653950974112377e-05 }, { "epoch": 0.0032451728054518902, "step": 10, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.8991473603528e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6491967471665703e-05 }, { "epoch": 0.0032451728054518902, "step": 10, "weight_max": 2.926079287135508e-05, "weight_mean": 5.174776106287027e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0617288353387266e-05 }, { "epoch": 0.003569690085997079, "step": 11, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2256698028068058e-05, "weight_min": 0.0, "weight_std": 1.7133324945461936e-05 }, { "epoch": 0.003569690085997079, "step": 11, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7900863895192742e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.624043215997517e-05 }, { "epoch": 0.003569690085997079, "step": 11, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0236436537525151e-05, "weight_min": 0.0, "weight_std": 1.7700005628284998e-05 }, { "epoch": 0.003569690085997079, "step": 11, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.2901422451250255e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7846561604528688e-05 }, { "epoch": 0.0038942073665422685, "step": 12, "weight_max": 4.252122016623616e-05, "weight_mean": 9.70837936620228e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8082977476296946e-05 }, { "epoch": 0.0038942073665422685, "step": 12, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233334251097403e-05, "weight_min": 0.0, "weight_std": 1.8953978724312037e-05 }, { "epoch": 0.0038942073665422685, "step": 12, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5472058294108137e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.137627416232135e-05 }, { "epoch": 0.0038942073665422685, "step": 12, "weight_max": 4.187939703115262e-05, "weight_mean": 7.570984053018037e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5327024811995216e-05 }, { "epoch": 0.0042187246470874575, "step": 13, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.622455763746984e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.787581823009532e-05 }, { "epoch": 0.0042187246470874575, "step": 13, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0321343981777318e-05, "weight_min": 0.0, "weight_std": 2.1790370738017373e-05 }, { "epoch": 0.0042187246470874575, "step": 13, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0351035345811397e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1781481336802244e-05 }, { "epoch": 0.0042187246470874575, "step": 13, "weight_max": 3.116124207735993e-05, "weight_mean": 7.290910616575275e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.354116466245614e-05 }, { "epoch": 0.004543241927632647, "step": 14, "weight_max": 3.967146767536178e-05, "weight_mean": 1.1040136087103747e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6195332136703655e-05 }, { "epoch": 0.004543241927632647, "step": 14, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6442598280264065e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9682762285810895e-05 }, { "epoch": 0.004543241927632647, "step": 14, "weight_max": 4.252122016623616e-05, "weight_mean": 1.889764826046303e-05, "weight_min": 0.0, "weight_std": 2.0975523511879146e-05 }, { "epoch": 0.004543241927632647, "step": 14, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7982986214046832e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.0724251296778675e-05 }, { "epoch": 0.004867759208177835, "step": 15, "weight_max": 2.2718091852169664e-09, "weight_mean": 8.987356414635883e-10, "weight_min": 0.0, "weight_std": 6.42424113905804e-10 }, { "epoch": 0.004867759208177835, "step": 15, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1227174582018051e-05, "weight_min": 0.0, "weight_std": 1.7925092834047973e-05 }, { "epoch": 0.004867759208177835, "step": 15, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2200077435409185e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7101981939049438e-05 }, { "epoch": 0.004867759208177835, "step": 15, "weight_max": 3.967146767536178e-05, "weight_mean": 1.7306996596744284e-05, "weight_min": 0.0, "weight_std": 1.9023555068997666e-05 }, { "epoch": 0.005192276488723025, "step": 16, "weight_max": 3.572963396436535e-05, "weight_mean": 1.473018892284017e-05, "weight_min": 0.0, "weight_std": 1.6517069525434636e-05 }, { "epoch": 0.005192276488723025, "step": 16, "weight_max": 4.381950566312298e-05, "weight_mean": 1.986209645110648e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.1342593754525296e-05 }, { "epoch": 0.005192276488723025, "step": 16, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2450818758225068e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.856761628005188e-05 }, { "epoch": 0.005192276488723025, "step": 16, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.499903742631432e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5754807463963516e-05 }, { "epoch": 0.005516793769268213, "step": 17, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.7366417523589917e-05, "weight_min": 0.0, "weight_std": 1.877927024906967e-05 }, { "epoch": 0.005516793769268213, "step": 17, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0116977136931382e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8746755813481286e-05 }, { "epoch": 0.005516793769268213, "step": 17, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2535146197478753e-05, "weight_min": 0.0, "weight_std": 1.9243849237682298e-05 }, { "epoch": 0.005516793769268213, "step": 17, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5262428355054e-06, "weight_min": 0.0, "weight_std": 9.970375685952604e-06 }, { "epoch": 0.005841311049813403, "step": 18, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8050424084067345e-06, "weight_min": 0.0, "weight_std": 1.3589214177045505e-05 }, { "epoch": 0.005841311049813403, "step": 18, "weight_max": 4.252122016623616e-05, "weight_mean": 9.746487194206566e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8141266991733573e-05 }, { "epoch": 0.005841311049813403, "step": 18, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0546893210848793e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.1983847545925528e-05 }, { "epoch": 0.005841311049813403, "step": 18, "weight_max": 4.252122016623616e-05, "weight_mean": 1.869108746177517e-05, "weight_min": 7.614652242082798e-10, "weight_std": 2.0301962649682537e-05 }, { "epoch": 0.006165828330358592, "step": 19, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2755333045788575e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7822163499658927e-05 }, { "epoch": 0.006165828330358592, "step": 19, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.203604582173284e-06, "weight_min": 0.0, "weight_std": 1.4753102732356638e-05 }, { "epoch": 0.006165828330358592, "step": 19, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4274220120569225e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.97537628991995e-05 }, { "epoch": 0.006165828330358592, "step": 19, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235576736595249e-06, "weight_min": 0.0, "weight_std": 1.4806339095230214e-05 }, { "epoch": 0.0064903456109037805, "grad_norm": 7.9021220207214355, "learning_rate": 2.0000000000000003e-06, "loss": 0.5631, "step": 20 }, { "epoch": 0.0064903456109037805, "step": 20, "weight_max": 1.504423474685268e-09, "weight_mean": 6.986342060422146e-10, "weight_min": 0.0, "weight_std": 4.0969638792631713e-10 }, { "epoch": 0.0064903456109037805, "step": 20, "weight_max": 4.187939703115262e-05, "weight_mean": 9.963672710000537e-06, "weight_min": 0.0, "weight_std": 1.8479842765373178e-05 }, { "epoch": 0.0064903456109037805, "step": 20, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23541848451714e-06, "weight_min": 0.0, "weight_std": 1.480640366935404e-05 }, { "epoch": 0.0064903456109037805, "step": 20, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.94867662282195e-06, "weight_min": 0.0, "weight_std": 1.682172114669811e-05 }, { "epoch": 0.00681486289144897, "step": 21, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4883856238157023e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.0565747036016546e-05 }, { "epoch": 0.00681486289144897, "step": 21, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0082483640871942e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8703603927860968e-05 }, { "epoch": 0.00681486289144897, "step": 21, "weight_max": 3.998322426923551e-05, "weight_mean": 8.859065928845666e-06, "weight_min": 0.0, "weight_std": 1.658202199905645e-05 }, { "epoch": 0.00681486289144897, "step": 21, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.141179932863452e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5251369404722936e-05 }, { "epoch": 0.007139380171994158, "step": 22, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0163676961383317e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8864320736611262e-05 }, { "epoch": 0.007139380171994158, "step": 22, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5113579138414934e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0866280465270393e-05 }, { "epoch": 0.007139380171994158, "step": 22, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470383131178096e-05, "weight_min": 0.0, "weight_std": 1.9386063286219724e-05 }, { "epoch": 0.007139380171994158, "step": 22, "weight_max": 3.782653220696375e-05, "weight_mean": 1.1640046977845486e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6459553080494516e-05 }, { "epoch": 0.007463897452539348, "step": 23, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.937946944963187e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0890342057100497e-05 }, { "epoch": 0.007463897452539348, "step": 23, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070233899255982e-06, "weight_min": 0.0, "weight_std": 1.1509709111123811e-05 }, { "epoch": 0.007463897452539348, "step": 23, "weight_max": 4.252122016623616e-05, "weight_mean": 2.504611256881617e-05, "weight_min": 1.247039249641091e-09, "weight_std": 2.0812853108509444e-05 }, { "epoch": 0.007463897452539348, "step": 23, "weight_max": 3.4268457966391e-05, "weight_mean": 7.214523066068068e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3665628102899063e-05 }, { "epoch": 0.007788414733084537, "step": 24, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0201218174188398e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.892966429295484e-05 }, { "epoch": 0.007788414733084537, "step": 24, "weight_max": 5.010648251868588e-08, "weight_mean": 7.62875718152145e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.727668852424813e-08 }, { "epoch": 0.007788414733084537, "step": 24, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.4376855688169599e-05, "weight_min": 0.0, "weight_std": 1.984427399293054e-05 }, { "epoch": 0.007788414733084537, "step": 24, "weight_max": 3.967146767536178e-05, "weight_mean": 5.037572918809019e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.39959101943532e-05 }, { "epoch": 0.008112932013629726, "step": 25, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.725403288030066e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.800789323169738e-05 }, { "epoch": 0.008112932013629726, "step": 25, "weight_max": 3.843663216684945e-05, "weight_mean": 4.806603556062328e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.3588584806711879e-05 }, { "epoch": 0.008112932013629726, "step": 25, "weight_max": 3.061215829802677e-05, "weight_mean": 7.619486950716237e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4107411516306456e-05 }, { "epoch": 0.008112932013629726, "step": 25, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5837354769464582e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.676777537795715e-05 }, { "epoch": 0.008437449294174915, "step": 26, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3240853149909526e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.838155549194198e-05 }, { "epoch": 0.008437449294174915, "step": 26, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7534837752464227e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9077766410191543e-05 }, { "epoch": 0.008437449294174915, "step": 26, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.559020165586844e-05, "weight_min": 0.0, "weight_std": 2.153649074898567e-05 }, { "epoch": 0.008437449294174915, "step": 26, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9822460672003217e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1297055354807526e-05 }, { "epoch": 0.008761966574720104, "step": 27, "weight_max": 4.187939703115262e-05, "weight_mean": 1.795854404917918e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.958870780072175e-05 }, { "epoch": 0.008761966574720104, "step": 27, "weight_max": 2.872879122151062e-05, "weight_mean": 3.5952091366198147e-06, "weight_min": 0.0, "weight_std": 1.0155505151487887e-05 }, { "epoch": 0.008761966574720104, "step": 27, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5549749150522985e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1476600522873923e-05 }, { "epoch": 0.008761966574720104, "step": 27, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8468539565219544e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9930803318857215e-05 }, { "epoch": 0.009086483855265294, "step": 28, "weight_max": 3.752332122530788e-05, "weight_mean": 9.23142670217203e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7094613212975673e-05 }, { "epoch": 0.009086483855265294, "step": 28, "weight_max": 3.572963396436535e-05, "weight_mean": 4.4667890506389085e-06, "weight_min": 0.0, "weight_std": 1.2632097423193045e-05 }, { "epoch": 0.009086483855265294, "step": 28, "weight_max": 5.042821715761647e-09, "weight_mean": 1.2300392926434256e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.546539674102121e-09 }, { "epoch": 0.009086483855265294, "step": 28, "weight_max": 3.572963396436535e-05, "weight_mean": 4.466906830202788e-06, "weight_min": 0.0, "weight_std": 1.2632049219973851e-05 }, { "epoch": 0.009411001135810481, "step": 29, "weight_max": 3.632264997577295e-05, "weight_mean": 9.020561265060678e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.666898788243998e-05 }, { "epoch": 0.009411001135810481, "step": 29, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7634160030866042e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.950484511326067e-05 }, { "epoch": 0.009411001135810481, "step": 29, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.394511961436365e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2423661246430129e-05 }, { "epoch": 0.009411001135810481, "step": 29, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0076035323436372e-05, "weight_min": 0.0, "weight_std": 1.8662456568563357e-05 }, { "epoch": 0.00973551841635567, "grad_norm": 7.5632195472717285, "learning_rate": 3e-06, "loss": 0.5526, "step": 30 }, { "epoch": 0.00973551841635567, "step": 30, "weight_max": 5.208601994155515e-09, "weight_mean": 1.7673045249466668e-09, "weight_min": 6.100955851628953e-10, "weight_std": 1.8019461478502308e-09 }, { "epoch": 0.00973551841635567, "step": 30, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0254036169499159e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8367431039223447e-05 }, { "epoch": 0.00973551841635567, "step": 30, "weight_max": 1.783980223990511e-05, "weight_mean": 2.2305148377199657e-06, "weight_min": 0.0, "weight_std": 6.3071047406992875e-06 }, { "epoch": 0.00973551841635567, "step": 30, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8706199625739828e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0082590708625503e-05 }, { "epoch": 0.01006003569690086, "step": 31, "weight_max": 3.116124207735993e-05, "weight_mean": 3.895910140272463e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1016857570211869e-05 }, { "epoch": 0.01006003569690086, "step": 31, "weight_max": 3.998322426923551e-05, "weight_mean": 9.958084774552844e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.843589598138351e-05 }, { "epoch": 0.01006003569690086, "step": 31, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311643563909456e-05, "weight_min": 0.0, "weight_std": 1.9095434254268184e-05 }, { "epoch": 0.01006003569690086, "step": 31, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315558155416511e-06, "weight_min": 0.0, "weight_std": 1.5033358067739755e-05 }, { "epoch": 0.01038455297744605, "step": 32, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315926500770729e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033208910608664e-05 }, { "epoch": 0.01038455297744605, "step": 32, "weight_max": 3.998322426923551e-05, "weight_mean": 9.42775841394905e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.749789407767821e-05 }, { "epoch": 0.01038455297744605, "step": 32, "weight_max": 3.722132169059478e-05, "weight_mean": 1.171159055957105e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.543133839732036e-05 }, { "epoch": 0.01038455297744605, "step": 32, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0632282283040695e-05, "weight_min": 0.0, "weight_std": 1.968679680430796e-05 }, { "epoch": 0.010709070257991239, "step": 33, "weight_max": 3.22740561387036e-05, "weight_mean": 4.4084235923946835e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1307886779832188e-05 }, { "epoch": 0.010709070257991239, "step": 33, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0753408534801565e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.948436511156615e-05 }, { "epoch": 0.010709070257991239, "step": 33, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.142531593563035e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8388627722742967e-05 }, { "epoch": 0.010709070257991239, "step": 33, "weight_max": 3.843663216684945e-05, "weight_mean": 8.805779543763492e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6391017197747715e-05 }, { "epoch": 0.011033587538536426, "step": 34, "weight_max": 3.813097282545641e-05, "weight_mean": 8.130265996442176e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.534857801743783e-05 }, { "epoch": 0.011033587538536426, "step": 34, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105599600734422e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1609555258473847e-05 }, { "epoch": 0.011033587538536426, "step": 34, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5031383554742206e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0902005417156033e-05 }, { "epoch": 0.011033587538536426, "step": 34, "weight_max": 3.843663216684945e-05, "weight_mean": 4.80606604469358e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3588800356956199e-05 }, { "epoch": 0.011358104819081616, "step": 35, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.997421329899225e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8512158931116574e-05 }, { "epoch": 0.011358104819081616, "step": 35, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4487523003481328e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0126792151131667e-05 }, { "epoch": 0.011358104819081616, "step": 35, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212508429191075e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1911725778190885e-05 }, { "epoch": 0.011358104819081616, "step": 35, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.498370673099998e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7601247236598283e-05 }, { "epoch": 0.011682622099626805, "step": 36, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156142378837103e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581074538000394e-05 }, { "epoch": 0.011682622099626805, "step": 36, "weight_max": 4.187939703115262e-05, "weight_mean": 9.447256161365658e-06, "weight_min": 0.0, "weight_std": 1.7627353372517973e-05 }, { "epoch": 0.011682622099626805, "step": 36, "weight_max": 3.9051607018336654e-05, "weight_mean": 2.5980809368775226e-05, "weight_min": 1.1534618815645103e-09, "weight_std": 1.6478852558066137e-05 }, { "epoch": 0.011682622099626805, "step": 36, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.863458005071152e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0918793122982606e-05 }, { "epoch": 0.012007139380171995, "step": 37, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9549901480786502e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.0997118554078043e-05 }, { "epoch": 0.012007139380171995, "step": 37, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.39719985681586e-06, "weight_min": 0.0, "weight_std": 1.5261675798683427e-05 }, { "epoch": 0.012007139380171995, "step": 37, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.80893037194619e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8192016796092503e-05 }, { "epoch": 0.012007139380171995, "step": 37, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8709946743911132e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.009282252402045e-05 }, { "epoch": 0.012331656660717184, "step": 38, "weight_max": 2.7939937353949063e-05, "weight_mean": 6.920982741576154e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.2813517969334498e-05 }, { "epoch": 0.012331656660717184, "step": 38, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.252587005903479e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7239113731193356e-05 }, { "epoch": 0.012331656660717184, "step": 38, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9450966647127643e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1106561689521186e-05 }, { "epoch": 0.012331656660717184, "step": 38, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0239303921698593e-05, "weight_min": 0.0, "weight_std": 2.1672685761586763e-05 }, { "epoch": 0.012656173941262372, "step": 39, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691211870522238e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3266175301396288e-05 }, { "epoch": 0.012656173941262372, "step": 39, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7218830180354416e-05, "weight_min": 0.0, "weight_std": 1.8760851162369363e-05 }, { "epoch": 0.012656173941262372, "step": 39, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653385985875502e-06, "weight_min": 0.0, "weight_std": 1.3159432455722708e-05 }, { "epoch": 0.012656173941262372, "step": 39, "weight_max": 7.937498991417158e-10, "weight_mean": 6.938096208664035e-10, "weight_min": 6.196655966128617e-10, "weight_std": 6.363611720905382e-11 }, { "epoch": 0.012980691221807561, "grad_norm": 4.254098415374756, "learning_rate": 4.000000000000001e-06, "loss": 0.518, "step": 40 }, { "epoch": 0.012980691221807561, "step": 40, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.120427421294153e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7076874428312294e-05 }, { "epoch": 0.012980691221807561, "step": 40, "weight_max": 4.187939703115262e-05, "weight_mean": 9.888250133371912e-06, "weight_min": 0.0, "weight_std": 1.835012881201692e-05 }, { "epoch": 0.012980691221807561, "step": 40, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0116944395122118e-05, "weight_min": 0.0, "weight_std": 1.8746775822364725e-05 }, { "epoch": 0.012980691221807561, "step": 40, "weight_max": 3.4558255720185116e-05, "weight_mean": 9.821063940762542e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5478997738682665e-05 }, { "epoch": 0.01330520850235275, "step": 41, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6475854863529094e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.000777931243647e-05 }, { "epoch": 0.01330520850235275, "step": 41, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550676051934715e-05, "weight_min": 0.0, "weight_std": 1.9535329556674697e-05 }, { "epoch": 0.01330520850235275, "step": 41, "weight_max": 6.949370003894728e-07, "weight_mean": 8.765739778482384e-08, "weight_min": 5.864964625068581e-10, "weight_std": 2.4537851572858926e-07 }, { "epoch": 0.01330520850235275, "step": 41, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5039989193610381e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0786435925401747e-05 }, { "epoch": 0.01362972578289794, "step": 42, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284388069208944e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2115394383727107e-05 }, { "epoch": 0.01362972578289794, "step": 42, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235554453975055e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4806348190177232e-05 }, { "epoch": 0.01362972578289794, "step": 42, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.536924537504092e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.921937109727878e-05 }, { "epoch": 0.01362972578289794, "step": 42, "weight_max": 3.967146767536178e-05, "weight_mean": 9.102800504479092e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.693673402769491e-05 }, { "epoch": 0.013954243063443129, "step": 43, "weight_max": 3.8743506593164057e-05, "weight_mean": 7.152773832785897e-06, "weight_min": 0.0, "weight_std": 1.4307586752693169e-05 }, { "epoch": 0.013954243063443129, "step": 43, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.611739571671933e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7739021132001653e-05 }, { "epoch": 0.013954243063443129, "step": 43, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.967760676459875e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.1211664059374016e-05 }, { "epoch": 0.013954243063443129, "step": 43, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2355776460899506e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4806338185735513e-05 }, { "epoch": 0.014278760343988317, "step": 44, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2695571058429778e-05, "weight_min": 0.0, "weight_std": 1.779795638867654e-05 }, { "epoch": 0.014278760343988317, "step": 44, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998544227419188e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135945093585178e-05 }, { "epoch": 0.014278760343988317, "step": 44, "weight_max": 4.187939703115262e-05, "weight_mean": 1.112069730879739e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6725152818253264e-05 }, { "epoch": 0.014278760343988317, "step": 44, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6529907952644862e-05, "weight_min": 0.0, "weight_std": 1.8437001926940866e-05 }, { "epoch": 0.014603277624533506, "step": 45, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.477059393248055e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7633961761021055e-05 }, { "epoch": 0.014603277624533506, "step": 45, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3164576456765644e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5032994269859046e-05 }, { "epoch": 0.014603277624533506, "step": 45, "weight_max": 3.0886084800840763e-09, "weight_mean": 1.3168172108279919e-09, "weight_min": 0.0, "weight_std": 1.1116088050044937e-09 }, { "epoch": 0.014603277624533506, "step": 45, "weight_max": 3.782653220696375e-05, "weight_mean": 9.195248821924906e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7033717085723765e-05 }, { "epoch": 0.014927794905078695, "step": 46, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2421054634614848e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.745754707371816e-05 }, { "epoch": 0.014927794905078695, "step": 46, "weight_max": 1.4469024790741969e-05, "weight_mean": 1.8093529661200591e-06, "weight_min": 6.29309993005478e-10, "weight_std": 5.115279691381147e-06 }, { "epoch": 0.014927794905078695, "step": 46, "weight_max": 1.504423474685268e-09, "weight_mean": 8.924240235685943e-10, "weight_min": 6.29309993005478e-10, "weight_std": 2.98095714779123e-10 }, { "epoch": 0.014927794905078695, "step": 46, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3821259674150497e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.984442860702984e-05 }, { "epoch": 0.015252312185623885, "step": 47, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2209113265271299e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8931374142994173e-05 }, { "epoch": 0.015252312185623885, "step": 47, "weight_max": 3.813097282545641e-05, "weight_mean": 4.768343842442846e-06, "weight_min": 0.0, "weight_std": 1.3480537745635957e-05 }, { "epoch": 0.015252312185623885, "step": 47, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4538979485223535e-05, "weight_min": 0.0, "weight_std": 2.0107851014472544e-05 }, { "epoch": 0.015252312185623885, "step": 47, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7438958821003325e-05, "weight_min": 0.0, "weight_std": 1.8787954104482196e-05 }, { "epoch": 0.015576829466169074, "step": 48, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235562639427371e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806344552198425e-05 }, { "epoch": 0.015576829466169074, "step": 48, "weight_max": 3.632264997577295e-05, "weight_mean": 8.645871275803074e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6034098734962754e-05 }, { "epoch": 0.015576829466169074, "step": 48, "weight_max": 1.4129907022208954e-08, "weight_mean": 2.963143286294212e-09, "weight_min": 5.771870759119224e-10, "weight_std": 4.849741497281457e-09 }, { "epoch": 0.015576829466169074, "step": 48, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.997411325457506e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8512164388084784e-05 }, { "epoch": 0.015901346746714264, "step": 49, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0637965715432074e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.948252975125797e-05 }, { "epoch": 0.015901346746714264, "step": 49, "weight_max": 3.782653220696375e-05, "weight_mean": 1.4102697605267167e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.705007889540866e-05 }, { "epoch": 0.015901346746714264, "step": 49, "weight_max": 4.252122016623616e-05, "weight_mean": 9.708586730994284e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8082850147038698e-05 }, { "epoch": 0.015901346746714264, "step": 49, "weight_max": 2.1563215341302566e-05, "weight_mean": 2.695897137527936e-06, "weight_min": 0.0, "weight_std": 7.623547844559653e-06 }, { "epoch": 0.01622586402725945, "grad_norm": 20.579547882080078, "learning_rate": 5e-06, "loss": 0.4437, "step": 50 }, { "epoch": 0.01622586402725945, "step": 50, "weight_max": 4.252122016623616e-05, "weight_mean": 9.893636160995811e-06, "weight_min": 0.0, "weight_std": 1.8385082512395456e-05 }, { "epoch": 0.01622586402725945, "step": 50, "weight_max": 3.967146767536178e-05, "weight_mean": 1.016565875033848e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.540331868454814e-05 }, { "epoch": 0.01622586402725945, "step": 50, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1052286936319433e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8749868104350753e-05 }, { "epoch": 0.01622586402725945, "step": 50, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006126103689894e-05, "weight_min": 0.0, "weight_std": 1.8574455680209212e-05 }, { "epoch": 0.016550381307804642, "step": 51, "weight_max": 4.252122016623616e-05, "weight_mean": 1.562742909300141e-05, "weight_min": 0.0, "weight_std": 2.157184098905418e-05 }, { "epoch": 0.016550381307804642, "step": 51, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470463166711852e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9386014173505828e-05 }, { "epoch": 0.016550381307804642, "step": 51, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396478172770003e-06, "weight_min": 0.0, "weight_std": 1.5261968655977398e-05 }, { "epoch": 0.016550381307804642, "step": 51, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.436590875731781e-05, "weight_min": 0.0, "weight_std": 2.032164775300771e-05 }, { "epoch": 0.01687489858834983, "step": 52, "weight_max": 2.846461757144425e-05, "weight_mean": 5.274654540698975e-06, "weight_min": 0.0, "weight_std": 1.0529504834266845e-05 }, { "epoch": 0.01687489858834983, "step": 52, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2865324606536888e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9006516595254652e-05 }, { "epoch": 0.01687489858834983, "step": 52, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.257477591338102e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.4938966160116252e-05 }, { "epoch": 0.01687489858834983, "step": 52, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.6359323177021e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6146237612701952e-05 }, { "epoch": 0.017199415868895018, "step": 53, "weight_max": 1.0698121499785884e-08, "weight_mean": 2.286610900625874e-09, "weight_min": 0.0, "weight_std": 3.448984653431353e-09 }, { "epoch": 0.017199415868895018, "step": 53, "weight_max": 4.187939703115262e-05, "weight_mean": 8.69588529894827e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6540370779694058e-05 }, { "epoch": 0.017199415868895018, "step": 53, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3940887331264094e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.996581158891786e-05 }, { "epoch": 0.017199415868895018, "step": 53, "weight_max": 4.187939703115262e-05, "weight_mean": 1.928188794408925e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0716535800602287e-05 }, { "epoch": 0.01752393314944021, "step": 54, "weight_max": 3.5141503758495674e-05, "weight_mean": 7.857462151150685e-06, "weight_min": 0.0, "weight_std": 1.3983122698846273e-05 }, { "epoch": 0.01752393314944021, "step": 54, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4655127415608149e-05, "weight_min": 0.0, "weight_std": 2.0269324522814713e-05 }, { "epoch": 0.01752393314944021, "step": 54, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.0359398402215447e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.4775711861148011e-05 }, { "epoch": 0.01752393314944021, "step": 54, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.958280315913726e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8442553482600488e-05 }, { "epoch": 0.017848450429985396, "step": 55, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156158749741735e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4581068171537481e-05 }, { "epoch": 0.017848450429985396, "step": 55, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.146651791525073e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7070289686671458e-05 }, { "epoch": 0.017848450429985396, "step": 55, "weight_max": 3.602553260861896e-05, "weight_mean": 8.196594535547774e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5270446965587325e-05 }, { "epoch": 0.017848450429985396, "step": 55, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1269647984590847e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7853684767032973e-05 }, { "epoch": 0.018172967710530587, "step": 56, "weight_max": 4.252122016623616e-05, "weight_mean": 1.14753838715842e-05, "weight_min": 0.0, "weight_std": 1.715959115244914e-05 }, { "epoch": 0.018172967710530587, "step": 56, "weight_max": 3.813097282545641e-05, "weight_mean": 1.3889944966649637e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.733375211188104e-05 }, { "epoch": 0.018172967710530587, "step": 56, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920946594211273e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.391585283272434e-05 }, { "epoch": 0.018172967710530587, "step": 56, "weight_max": 4.187939703115262e-05, "weight_mean": 2.337268233532086e-05, "weight_min": 0.0, "weight_std": 1.9919674741686322e-05 }, { "epoch": 0.018497484991075775, "step": 57, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8463813830749132e-05, "weight_min": 0.0, "weight_std": 2.004503403441049e-05 }, { "epoch": 0.018497484991075775, "step": 57, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.900024451781064e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.842833444243297e-05 }, { "epoch": 0.018497484991075775, "step": 57, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3157964430283755e-06, "weight_min": 0.0, "weight_std": 1.5033261661301367e-05 }, { "epoch": 0.018497484991075775, "step": 57, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4765152627660427e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0395704268594272e-05 }, { "epoch": 0.018822002271620963, "step": 58, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316156602930278e-06, "weight_min": 0.0, "weight_std": 1.5033115232654382e-05 }, { "epoch": 0.018822002271620963, "step": 58, "weight_max": 1.2909623592349817e-06, "weight_mean": 2.1288688856202498e-07, "weight_min": 0.0, "weight_std": 4.5839263407287945e-07 }, { "epoch": 0.018822002271620963, "step": 58, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0037259016826283e-05, "weight_min": 0.0, "weight_std": 1.8593376807984896e-05 }, { "epoch": 0.018822002271620963, "step": 58, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3258922081149649e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.847998282755725e-05 }, { "epoch": 0.019146519552166154, "step": 59, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3954184143804014e-05, "weight_min": 0.0, "weight_std": 2.0025350750074722e-05 }, { "epoch": 0.019146519552166154, "step": 59, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.015370071399957e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.880280251498334e-05 }, { "epoch": 0.019146519552166154, "step": 59, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0043713700724766e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.860971315181814e-05 }, { "epoch": 0.019146519552166154, "step": 59, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0153240509680472e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.879856972664129e-05 }, { "epoch": 0.01947103683271134, "grad_norm": 4.91776180267334, "learning_rate": 5.8e-06, "loss": 0.3229, "step": 60 }, { "epoch": 0.01947103683271134, "step": 60, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5362353224190883e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.124747152265627e-05 }, { "epoch": 0.01947103683271134, "step": 60, "weight_max": 2.639517333591357e-05, "weight_mean": 5.108458026370499e-06, "weight_min": 0.0, "weight_std": 9.98060295387404e-06 }, { "epoch": 0.01947103683271134, "step": 60, "weight_max": 3.843663216684945e-05, "weight_mean": 8.91055697138654e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.625386721570976e-05 }, { "epoch": 0.01947103683271134, "step": 60, "weight_max": 2.319888153579086e-05, "weight_mean": 2.9006737349845935e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.201714990718756e-06 }, { "epoch": 0.019795554113256533, "step": 61, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.430877980776131e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7702108380035497e-05 }, { "epoch": 0.019795554113256533, "step": 61, "weight_max": 4.447596802492626e-05, "weight_mean": 1.5634121155017056e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.161536758649163e-05 }, { "epoch": 0.019795554113256533, "step": 61, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2488058928283863e-05, "weight_min": 0.0, "weight_std": 1.804174462449737e-05 }, { "epoch": 0.019795554113256533, "step": 61, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0551959348958917e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.954357503564097e-05 }, { "epoch": 0.02012007139380172, "step": 62, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.0172016118303873e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.497176617704099e-05 }, { "epoch": 0.02012007139380172, "step": 62, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.302151738665998e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7187043340527453e-05 }, { "epoch": 0.02012007139380172, "step": 62, "weight_max": 4.187939703115262e-05, "weight_mean": 1.969116783584468e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1094316252856515e-05 }, { "epoch": 0.02012007139380172, "step": 62, "weight_max": 3.998322426923551e-05, "weight_mean": 9.841497558227275e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8224303858005442e-05 }, { "epoch": 0.020444588674346908, "step": 63, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315798262017779e-06, "weight_min": 0.0, "weight_std": 1.5033260751806665e-05 }, { "epoch": 0.020444588674346908, "step": 63, "weight_max": 3.4268457966391e-05, "weight_mean": 8.221347343351226e-06, "weight_min": 0.0, "weight_std": 1.4906456272001378e-05 }, { "epoch": 0.020444588674346908, "step": 63, "weight_max": 3.4268457966391e-05, "weight_mean": 4.289148364478024e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2113478987885173e-05 }, { "epoch": 0.020444588674346908, "step": 63, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2699953913397621e-05, "weight_min": 0.0, "weight_std": 1.765797787811607e-05 }, { "epoch": 0.0207691059548921, "step": 64, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.549547939968761e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7753529391484335e-05 }, { "epoch": 0.0207691059548921, "step": 64, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.687356396170799e-06, "weight_min": 0.0, "weight_std": 1.7938364180736244e-05 }, { "epoch": 0.0207691059548921, "step": 64, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7931072256469633e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0726340406108648e-05 }, { "epoch": 0.0207691059548921, "step": 64, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.1735862901550718e-05, "weight_min": 7.614652242082798e-10, "weight_std": 1.9811004676739685e-05 }, { "epoch": 0.021093623235437287, "step": 65, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284448095859261e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.211536982737016e-05 }, { "epoch": 0.021093623235437287, "step": 65, "weight_max": 4.187939703115262e-05, "weight_mean": 8.793487722869031e-06, "weight_min": 0.0, "weight_std": 1.6671254343236797e-05 }, { "epoch": 0.021093623235437287, "step": 65, "weight_max": 4.252122016623616e-05, "weight_mean": 5.317554496286903e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5032552255433984e-05 }, { "epoch": 0.021093623235437287, "step": 65, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3488473996403627e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.884237462945748e-05 }, { "epoch": 0.021418140515982478, "step": 66, "weight_max": 1.6265615521149357e-09, "weight_mean": 9.566722969367447e-10, "weight_min": 6.196655966128617e-10, "weight_std": 4.2134126743142986e-10 }, { "epoch": 0.021418140515982478, "step": 66, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921498202747898e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3915630916017108e-05 }, { "epoch": 0.021418140515982478, "step": 66, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.746985258767381e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.653656545386184e-05 }, { "epoch": 0.021418140515982478, "step": 66, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236514425632777e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4805960745434277e-05 }, { "epoch": 0.021742657796527665, "step": 67, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.7522632333566435e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8832188288797624e-05 }, { "epoch": 0.021742657796527665, "step": 67, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3656193914357573e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9000861357199028e-05 }, { "epoch": 0.021742657796527665, "step": 67, "weight_max": 3.998322426923551e-05, "weight_mean": 2.067807872663252e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7986534658120945e-05 }, { "epoch": 0.021742657796527665, "step": 67, "weight_max": 1.4305072681963793e-09, "weight_mean": 7.203470042682625e-10, "weight_min": 0.0, "weight_std": 4.2042189174473776e-10 }, { "epoch": 0.022067175077072853, "step": 68, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4950961485737935e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0778186808456667e-05 }, { "epoch": 0.022067175077072853, "step": 68, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.445804269285873e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5919275028863922e-05 }, { "epoch": 0.022067175077072853, "step": 68, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3751199730904773e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9054621589020826e-05 }, { "epoch": 0.022067175077072853, "step": 68, "weight_max": 3.967146767536178e-05, "weight_mean": 9.802670319913886e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8150742107536644e-05 }, { "epoch": 0.022391692357618044, "step": 69, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3148435502662323e-05, "weight_min": 8.156453290553145e-10, "weight_std": 1.9473165593808517e-05 }, { "epoch": 0.022391692357618044, "step": 69, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.3968237807566766e-06, "weight_min": 0.0, "weight_std": 1.5261828593793325e-05 }, { "epoch": 0.022391692357618044, "step": 69, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2535016139736399e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8190643459092826e-05 }, { "epoch": 0.022391692357618044, "step": 69, "weight_max": 3.967146767536178e-05, "weight_mean": 8.518139111401979e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6051659258664586e-05 }, { "epoch": 0.02271620963816323, "grad_norm": 1.810129165649414, "learning_rate": 6.800000000000001e-06, "loss": 0.2422, "step": 70 }, { "epoch": 0.02271620963816323, "step": 70, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0197264600719791e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8902790543506853e-05 }, { "epoch": 0.02271620963816323, "step": 70, "weight_max": 3.602553260861896e-05, "weight_mean": 8.823757525533438e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.6341195077984594e-05 }, { "epoch": 0.02271620963816323, "step": 70, "weight_max": 4.252122016623616e-05, "weight_mean": 9.779725587577559e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.788261579349637e-05 }, { "epoch": 0.02271620963816323, "step": 70, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805200205737492e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3589151421911083e-05 }, { "epoch": 0.023040726918708423, "step": 71, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3233204046846367e-05, "weight_min": 0.0, "weight_std": 1.9596634956542403e-05 }, { "epoch": 0.023040726918708423, "step": 71, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077069090475561e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357636246131733e-05 }, { "epoch": 0.023040726918708423, "step": 71, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921281743008876e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3915718227508478e-05 }, { "epoch": 0.023040726918708423, "step": 71, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8867067410610616e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.096550088026561e-05 }, { "epoch": 0.02336524419925361, "step": 72, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4379204003489576e-05, "weight_min": 0.0, "weight_std": 1.985964809136931e-05 }, { "epoch": 0.02336524419925361, "step": 72, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0335768567747436e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.80520710273413e-05 }, { "epoch": 0.02336524419925361, "step": 72, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392283911642153e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9247496311436407e-05 }, { "epoch": 0.02336524419925361, "step": 72, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3509752534446307e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.91004182852339e-05 }, { "epoch": 0.023689761479798798, "step": 73, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3693538676307071e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.904341479530558e-05 }, { "epoch": 0.023689761479798798, "step": 73, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.805513400351629e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8169308532378636e-05 }, { "epoch": 0.023689761479798798, "step": 73, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076800334791187e-06, "weight_min": 0.0, "weight_std": 1.4357743566506542e-05 }, { "epoch": 0.023689761479798798, "step": 73, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155902600672562e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8814991562976502e-05 }, { "epoch": 0.02401427876034399, "step": 74, "weight_max": 4.252122016623616e-05, "weight_mean": 9.455686267756391e-06, "weight_min": 0.0, "weight_std": 1.768698712112382e-05 }, { "epoch": 0.02401427876034399, "step": 74, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.274594049202278e-06, "weight_min": 0.0, "weight_std": 1.7203854440595023e-05 }, { "epoch": 0.02401427876034399, "step": 74, "weight_max": 3.967146767536178e-05, "weight_mean": 4.96010670758551e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4025507880432997e-05 }, { "epoch": 0.02401427876034399, "step": 74, "weight_max": 4.187939703115262e-05, "weight_mean": 6.42706891085254e-06, "weight_min": 0.0, "weight_std": 1.470810911996523e-05 }, { "epoch": 0.024338796040889177, "step": 75, "weight_max": 1.753466261078529e-09, "weight_mean": 7.046235817043112e-10, "weight_min": 0.0, "weight_std": 5.985070772318579e-10 }, { "epoch": 0.024338796040889177, "step": 75, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0037609172286466e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8593162167235278e-05 }, { "epoch": 0.024338796040889177, "step": 75, "weight_max": 3.143761568935588e-05, "weight_mean": 3.973446837335359e-06, "weight_min": 0.0, "weight_std": 1.1097818969574291e-05 }, { "epoch": 0.024338796040889177, "step": 75, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.800390913966112e-06, "weight_min": 0.0, "weight_std": 1.0723416380642448e-05 }, { "epoch": 0.024663313321434368, "step": 76, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728852218249813e-06, "weight_min": 0.0, "weight_std": 1.3373482033784967e-05 }, { "epoch": 0.024663313321434368, "step": 76, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.888937372830696e-06, "weight_min": 0.0, "weight_std": 1.695373430266045e-05 }, { "epoch": 0.024663313321434368, "step": 76, "weight_max": 1.506233004988644e-08, "weight_mean": 2.499640716635554e-09, "weight_min": 5.679522963042416e-10, "weight_std": 5.0812123397747655e-09 }, { "epoch": 0.024663313321434368, "step": 76, "weight_max": 3.843663216684945e-05, "weight_mean": 1.1736452506738715e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7494096027803607e-05 }, { "epoch": 0.024987830601979556, "step": 77, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.506285095994826e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.765497108863201e-05 }, { "epoch": 0.024987830601979556, "step": 77, "weight_max": 3.28377827827353e-05, "weight_mean": 4.1054254324990325e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1609626199060585e-05 }, { "epoch": 0.024987830601979556, "step": 77, "weight_max": 1.6265615521149357e-09, "weight_mean": 7.63984486784608e-10, "weight_min": 0.0, "weight_std": 4.6071177428608223e-10 }, { "epoch": 0.024987830601979556, "step": 77, "weight_max": 4.187939703115262e-05, "weight_mean": 9.394948392582592e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6680867702234536e-05 }, { "epoch": 0.025312347882524743, "step": 78, "weight_max": 8.603301404619401e-10, "weight_mean": 4.996637548160265e-10, "weight_min": 0.0, "weight_std": 3.2457550536157953e-10 }, { "epoch": 0.025312347882524743, "step": 78, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0150765774596948e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.825983781600371e-05 }, { "epoch": 0.025312347882524743, "step": 78, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4147007277642842e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9649058231152594e-05 }, { "epoch": 0.025312347882524743, "step": 78, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0413134077680297e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7897542420541868e-05 }, { "epoch": 0.025636865163069934, "step": 79, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3341015801415779e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.864129808382131e-05 }, { "epoch": 0.025636865163069934, "step": 79, "weight_max": 3.782653220696375e-05, "weight_mean": 9.127286830334924e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.6900190530577675e-05 }, { "epoch": 0.025636865163069934, "step": 79, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0754941942868754e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.224629497504793e-05 }, { "epoch": 0.025636865163069934, "step": 79, "weight_max": 9.413956281179026e-10, "weight_mean": 5.355069720991423e-10, "weight_min": 0.0, "weight_std": 3.48951090245464e-10 }, { "epoch": 0.025961382443615122, "grad_norm": 38.84080123901367, "learning_rate": 7.800000000000002e-06, "loss": 0.5415, "step": 80 }, { "epoch": 0.025961382443615122, "step": 80, "weight_max": 3.967146767536178e-05, "weight_mean": 9.52976097323699e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6064643205027096e-05 }, { "epoch": 0.025961382443615122, "step": 80, "weight_max": 3.061215829802677e-05, "weight_mean": 3.82733742299024e-06, "weight_min": 0.0, "weight_std": 1.082270136976149e-05 }, { "epoch": 0.025961382443615122, "step": 80, "weight_max": 3.199401908204891e-05, "weight_mean": 8.8501574282418e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3969514839118347e-05 }, { "epoch": 0.025961382443615122, "step": 80, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.10818630977883e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4835603906249162e-05 }, { "epoch": 0.026285899724160313, "step": 81, "weight_max": 4.473001524729625e-08, "weight_mean": 6.07957195697395e-09, "weight_min": 0.0, "weight_std": 1.5622212856669648e-08 }, { "epoch": 0.026285899724160313, "step": 81, "weight_max": 3.722132169059478e-05, "weight_mean": 8.54841164255049e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5909776266198605e-05 }, { "epoch": 0.026285899724160313, "step": 81, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0235083866282366e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1666151951649226e-05 }, { "epoch": 0.026285899724160313, "step": 81, "weight_max": 2.3677199351368472e-05, "weight_mean": 2.9661878215847537e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.368529051949736e-06 }, { "epoch": 0.0266104170047055, "step": 82, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578287644108059e-06, "weight_min": 0.0, "weight_std": 1.2947206414537504e-05 }, { "epoch": 0.0266104170047055, "step": 82, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793817541009048e-06, "weight_min": 0.0, "weight_std": 1.0726053915277589e-05 }, { "epoch": 0.0266104170047055, "step": 82, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8952567188534886e-05, "weight_min": 0.0, "weight_std": 2.0425170077942312e-05 }, { "epoch": 0.0266104170047055, "step": 82, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3731738363276236e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.919043825182598e-05 }, { "epoch": 0.02693493428525069, "step": 83, "weight_max": 3.662098970380612e-05, "weight_mean": 9.896131814457476e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.520036494184751e-05 }, { "epoch": 0.02693493428525069, "step": 83, "weight_max": 4.252122016623616e-05, "weight_mean": 2.179699731641449e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8879671188187785e-05 }, { "epoch": 0.02693493428525069, "step": 83, "weight_max": 3.22740561387036e-05, "weight_mean": 7.759546861052513e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.438169419998303e-05 }, { "epoch": 0.02693493428525069, "step": 83, "weight_max": 4.252122016623616e-05, "weight_mean": 1.928375058923848e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.101436439261306e-05 }, { "epoch": 0.02725945156579588, "step": 84, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2899308785563335e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8005634046858177e-05 }, { "epoch": 0.02725945156579588, "step": 84, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.1011827155016363e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.610609979252331e-05 }, { "epoch": 0.02725945156579588, "step": 84, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.682174661254976e-05, "weight_min": 0.0, "weight_std": 1.998736843233928e-05 }, { "epoch": 0.02725945156579588, "step": 84, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.972005161922425e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6802436221041717e-05 }, { "epoch": 0.027583968846341067, "step": 85, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155840881343465e-06, "weight_min": 0.0, "weight_std": 1.4581196410290431e-05 }, { "epoch": 0.027583968846341067, "step": 85, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7822341760620475e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9486329620121978e-05 }, { "epoch": 0.027583968846341067, "step": 85, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.99547740118578e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6863768905750476e-05 }, { "epoch": 0.027583968846341067, "step": 85, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4610958714911249e-05, "weight_min": 0.0, "weight_std": 2.0135015802225098e-05 }, { "epoch": 0.027908486126886258, "step": 86, "weight_max": 3.967146767536178e-05, "weight_mean": 8.65077799971914e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6244019207078964e-05 }, { "epoch": 0.027908486126886258, "step": 86, "weight_max": 3.813097282545641e-05, "weight_mean": 6.552280865435023e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3703386684937868e-05 }, { "epoch": 0.027908486126886258, "step": 86, "weight_max": 4.252122016623616e-05, "weight_mean": 9.860877071332652e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.832053931138944e-05 }, { "epoch": 0.027908486126886258, "step": 86, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3738309462496545e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.92758616321953e-05 }, { "epoch": 0.028233003407431446, "step": 87, "weight_max": 3.967146767536178e-05, "weight_mean": 8.456689101876691e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.595682442712132e-05 }, { "epoch": 0.028233003407431446, "step": 87, "weight_max": 1.3727149417164242e-09, "weight_mean": 8.667220829927658e-10, "weight_min": 5.958802340444436e-10, "weight_std": 3.252345059934214e-10 }, { "epoch": 0.028233003407431446, "step": 87, "weight_max": 3.998322426923551e-05, "weight_mean": 5.13888380737626e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4084479516895954e-05 }, { "epoch": 0.028233003407431446, "step": 87, "weight_max": 2.7679425329552032e-05, "weight_mean": 3.460686457401607e-06, "weight_min": 6.100955851628953e-10, "weight_std": 9.785848305909894e-06 }, { "epoch": 0.028557520687976633, "step": 88, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077041350887157e-06, "weight_min": 0.0, "weight_std": 1.4357646250573453e-05 }, { "epoch": 0.028557520687976633, "step": 88, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4457382349064574e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.997168510570191e-05 }, { "epoch": 0.028557520687976633, "step": 88, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.68667518463917e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7935395590029657e-05 }, { "epoch": 0.028557520687976633, "step": 88, "weight_max": 4.356234967417549e-06, "weight_mean": 5.454484721667541e-07, "weight_min": 0.0, "weight_std": 1.539790673632524e-06 }, { "epoch": 0.028882037968521825, "step": 89, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4580286006093957e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.017748192884028e-05 }, { "epoch": 0.028882037968521825, "step": 89, "weight_max": 3.998322426923551e-05, "weight_mean": 9.918685464072041e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.83649444807088e-05 }, { "epoch": 0.028882037968521825, "step": 89, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.367355232825503e-06, "weight_min": 0.0, "weight_std": 1.7460764865973033e-05 }, { "epoch": 0.028882037968521825, "step": 89, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.410850124259014e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2417155630828347e-05 }, { "epoch": 0.029206555249067012, "grad_norm": 0.7956805229187012, "learning_rate": 8.8e-06, "loss": 0.1772, "step": 90 }, { "epoch": 0.029206555249067012, "step": 90, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.713093848200515e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.631134728086181e-05 }, { "epoch": 0.029206555249067012, "step": 90, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.452215449011419e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0615541870938614e-05 }, { "epoch": 0.029206555249067012, "step": 90, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.439481800654903e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.757610880304128e-05 }, { "epoch": 0.029206555249067012, "step": 90, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070014711032854e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1509798241604585e-05 }, { "epoch": 0.029531072529612203, "step": 91, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0206098522758111e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8897359041147865e-05 }, { "epoch": 0.029531072529612203, "step": 91, "weight_max": 2.952862632810138e-05, "weight_mean": 7.1515341915073805e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3249985386210028e-05 }, { "epoch": 0.029531072529612203, "step": 91, "weight_max": 3.632264997577295e-05, "weight_mean": 7.398178240691777e-06, "weight_min": 0.0, "weight_std": 1.4131699572317302e-05 }, { "epoch": 0.029531072529612203, "step": 91, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.149100722221192e-06, "weight_min": 0.0, "weight_std": 1.513600454927655e-05 }, { "epoch": 0.02985558981015739, "step": 92, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5140993127715774e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.10209509532433e-05 }, { "epoch": 0.02985558981015739, "step": 92, "weight_max": 3.9360926166409627e-05, "weight_mean": 7.337237548199482e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4600692338717636e-05 }, { "epoch": 0.02985558981015739, "step": 92, "weight_max": 4.381950566312298e-05, "weight_mean": 1.8036294932244346e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0215542463120073e-05 }, { "epoch": 0.02985558981015739, "step": 92, "weight_max": 1.1403916699848082e-09, "weight_mean": 6.081406489499841e-10, "weight_min": 0.0, "weight_std": 3.097556378062194e-10 }, { "epoch": 0.03018010709070258, "step": 93, "weight_max": 1.7859370649020434e-09, "weight_mean": 6.662970730708651e-10, "weight_min": 0.0, "weight_std": 6.743401947950645e-10 }, { "epoch": 0.03018010709070258, "step": 93, "weight_max": 4.252122016623616e-05, "weight_mean": 5.31594741914887e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5033199815661646e-05 }, { "epoch": 0.03018010709070258, "step": 93, "weight_max": 3.843663216684945e-05, "weight_mean": 1.4186052794684656e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9580220396164805e-05 }, { "epoch": 0.03018010709070258, "step": 93, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313645361748058e-05, "weight_min": 0.0, "weight_std": 1.9107746993540786e-05 }, { "epoch": 0.03050462437124777, "step": 94, "weight_max": 3.843663216684945e-05, "weight_mean": 9.34559284360148e-06, "weight_min": 0.0, "weight_std": 1.731221163936425e-05 }, { "epoch": 0.03050462437124777, "step": 94, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1367888873792253e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7328846297459677e-05 }, { "epoch": 0.03050462437124777, "step": 94, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9773280655499548e-05, "weight_min": 0.0, "weight_std": 2.119866621796973e-05 }, { "epoch": 0.03050462437124777, "step": 94, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.71639747149311e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8136290236725472e-05 }, { "epoch": 0.030829141651792957, "step": 95, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235791377344867e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806252693233546e-05 }, { "epoch": 0.030829141651792957, "step": 95, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.325088856508955e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.5455938410013914e-05 }, { "epoch": 0.030829141651792957, "step": 95, "weight_max": 4.187939703115262e-05, "weight_mean": 1.056370365404291e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.851225533755496e-05 }, { "epoch": 0.030829141651792957, "step": 95, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9984591896645725e-06, "weight_min": 0.0, "weight_std": 1.4135979654383846e-05 }, { "epoch": 0.03115365893233815, "step": 96, "weight_max": 4.252122016623616e-05, "weight_mean": 9.420633432455361e-06, "weight_min": 0.0, "weight_std": 1.7632710296311416e-05 }, { "epoch": 0.03115365893233815, "step": 96, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4420769730350003e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9930608686991036e-05 }, { "epoch": 0.03115365893233815, "step": 96, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396710548666306e-06, "weight_min": 0.0, "weight_std": 1.5261874068528414e-05 }, { "epoch": 0.03115365893233815, "step": 96, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0537790330999997e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8306080164620653e-05 }, { "epoch": 0.03147817621288334, "step": 97, "weight_max": 4.187939703115262e-05, "weight_mean": 9.8507953225635e-06, "weight_min": 0.0, "weight_std": 1.828623680921737e-05 }, { "epoch": 0.03147817621288334, "step": 97, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3441801456792746e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8575537978904322e-05 }, { "epoch": 0.03147817621288334, "step": 97, "weight_max": 4.252122016623616e-05, "weight_mean": 2.2646152501692995e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0925039279973134e-05 }, { "epoch": 0.03147817621288334, "step": 97, "weight_max": 3.998322426923551e-05, "weight_mean": 9.391280400450341e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7435653717257082e-05 }, { "epoch": 0.03180269349342853, "step": 98, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0712032235460356e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9833387341350317e-05 }, { "epoch": 0.03180269349342853, "step": 98, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3784693692286965e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9130246073473245e-05 }, { "epoch": 0.03180269349342853, "step": 98, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.390468398516532e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0501371182035655e-05 }, { "epoch": 0.03180269349342853, "step": 98, "weight_max": 9.160966101262602e-07, "weight_mean": 1.1496208429662147e-07, "weight_min": 0.0, "weight_std": 3.2370738267673005e-07 }, { "epoch": 0.032127210773973715, "step": 99, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4300486479187384e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2527876606327482e-05 }, { "epoch": 0.032127210773973715, "step": 99, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.792198397917673e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8249456843477674e-05 }, { "epoch": 0.032127210773973715, "step": 99, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4296005247160792e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9879673345712945e-05 }, { "epoch": 0.032127210773973715, "step": 99, "weight_max": 2.1793222913402133e-05, "weight_mean": 2.7247976959188236e-06, "weight_min": 6.100955851628953e-10, "weight_std": 7.704807103436906e-06 }, { "epoch": 0.0324517280545189, "grad_norm": 0.8745288252830505, "learning_rate": 9.800000000000001e-06, "loss": 0.2376, "step": 100 }, { "epoch": 0.0324517280545189, "eval_accuracy": 0.494, "eval_auc": 0.6186075794914467, "eval_loss": 2.247490406036377, "eval_runtime": 155.5932, "eval_samples_per_second": 12.854, "eval_steps_per_second": 1.607, "step": 100 }, { "epoch": 0.0324517280545189, "step": 100, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9118986529065296e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.066408342216164e-05 }, { "epoch": 0.0324517280545189, "step": 100, "weight_max": 3.6920544516760856e-05, "weight_mean": 7.282433216460049e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4111868949839845e-05 }, { "epoch": 0.0324517280545189, "step": 100, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.6544623576919548e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8426895621814765e-05 }, { "epoch": 0.0324517280545189, "step": 100, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0165803359996062e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7179912902065553e-05 }, { "epoch": 0.03277624533506409, "step": 101, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0126266463194042e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1587236915365793e-05 }, { "epoch": 0.03277624533506409, "step": 101, "weight_max": 4.381950566312298e-05, "weight_mean": 9.103118827624712e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.7311751435045153e-05 }, { "epoch": 0.03277624533506409, "step": 101, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0114788892678916e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.873230212368071e-05 }, { "epoch": 0.03277624533506409, "step": 101, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4305233889899682e-05, "weight_min": 0.0, "weight_std": 1.976594103325624e-05 }, { "epoch": 0.033100762615609285, "step": 102, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3319513527676463e-05, "weight_min": 0.0, "weight_std": 1.8946775526273996e-05 }, { "epoch": 0.033100762615609285, "step": 102, "weight_max": 3.4849264920921996e-05, "weight_mean": 5.101460374135058e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2196671377751045e-05 }, { "epoch": 0.033100762615609285, "step": 102, "weight_max": 3.843663216684945e-05, "weight_mean": 8.495970178046264e-06, "weight_min": 0.0, "weight_std": 1.5909847206785344e-05 }, { "epoch": 0.033100762615609285, "step": 102, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2552158295875415e-05, "weight_min": 0.0, "weight_std": 1.7578631741343997e-05 }, { "epoch": 0.03342527989615447, "step": 103, "weight_max": 3.662098970380612e-05, "weight_mean": 6.347906492010225e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3133097127138171e-05 }, { "epoch": 0.03342527989615447, "step": 103, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0552842468314338e-05, "weight_min": 0.0, "weight_std": 1.9533994418452494e-05 }, { "epoch": 0.03342527989615447, "step": 103, "weight_max": 1.0509849657225345e-09, "weight_mean": 6.886976544606682e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.5825638022270994e-10 }, { "epoch": 0.03342527989615447, "step": 103, "weight_max": 4.252122016623616e-05, "weight_mean": 1.031364627124276e-05, "weight_min": 0.0, "weight_std": 1.9107746993540786e-05 }, { "epoch": 0.03374979717669966, "step": 104, "weight_max": 3.843663216684945e-05, "weight_mean": 4.806116521649528e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.358878034807276e-05 }, { "epoch": 0.03374979717669966, "step": 104, "weight_max": 4.0610404539620504e-05, "weight_mean": 6.956535798963159e-06, "weight_min": 0.0, "weight_std": 1.4580842616851442e-05 }, { "epoch": 0.03374979717669966, "step": 104, "weight_max": 3.998322426923551e-05, "weight_mean": 1.9203356714569964e-05, "weight_min": 0.0, "weight_std": 2.0129058611928485e-05 }, { "epoch": 0.03374979717669966, "step": 104, "weight_max": 3.255531191825867e-05, "weight_mean": 8.069367140706163e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4940587789169513e-05 }, { "epoch": 0.03407431445724485, "step": 105, "weight_max": 3.843663216684945e-05, "weight_mean": 1.8766124412650242e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.007649527513422e-05 }, { "epoch": 0.03407431445724485, "step": 105, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9396589777898043e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0984796719858423e-05 }, { "epoch": 0.03407431445724485, "step": 105, "weight_max": 3.143761568935588e-05, "weight_mean": 3.9304368328885175e-06, "weight_min": 0.0, "weight_std": 1.1114579137938563e-05 }, { "epoch": 0.03407431445724485, "step": 105, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9258513930253685e-05, "weight_min": 0.0, "weight_std": 2.0758583559654653e-05 }, { "epoch": 0.034398831737790035, "step": 106, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4522425772156566e-05, "weight_min": 0.0, "weight_std": 2.0083351046196185e-05 }, { "epoch": 0.034398831737790035, "step": 106, "weight_max": 4.187939703115262e-05, "weight_mean": 1.904878081404604e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0452442186069675e-05 }, { "epoch": 0.034398831737790035, "step": 106, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4381603250512853e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.987247924262192e-05 }, { "epoch": 0.034398831737790035, "step": 106, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.797566619643476e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9772433006437495e-05 }, { "epoch": 0.03472334901833523, "step": 107, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4379147362196818e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0389754354255274e-05 }, { "epoch": 0.03472334901833523, "step": 107, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2055854313075542e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.596870060893707e-05 }, { "epoch": 0.03472334901833523, "step": 107, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397004770202329e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.526175401522778e-05 }, { "epoch": 0.03472334901833523, "step": 107, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3400672944262624e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.9802884708042257e-05 }, { "epoch": 0.03504786629888042, "step": 108, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729677129944321e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3373149158724118e-05 }, { "epoch": 0.03504786629888042, "step": 108, "weight_max": 3.998322426923551e-05, "weight_mean": 6.2872304624761455e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4085112525208388e-05 }, { "epoch": 0.03504786629888042, "step": 108, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311292498954572e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9091130525339395e-05 }, { "epoch": 0.03504786629888042, "step": 108, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.15471685200464e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5173015526670497e-05 }, { "epoch": 0.035372383579425605, "step": 109, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5198627806967124e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.101233803841751e-05 }, { "epoch": 0.035372383579425605, "step": 109, "weight_max": 3.813097282545641e-05, "weight_mean": 4.805127900908701e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.3466096788761206e-05 }, { "epoch": 0.035372383579425605, "step": 109, "weight_max": 3.3961886636291183e-09, "weight_mean": 1.059719423324168e-09, "weight_min": 5.864964625068581e-10, "weight_std": 9.522296284814047e-10 }, { "epoch": 0.035372383579425605, "step": 109, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2456191143428441e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8711616576183587e-05 }, { "epoch": 0.03569690085997079, "grad_norm": 4.3439154624938965, "learning_rate": 9.983673469387756e-06, "loss": 0.1912, "step": 110 }, { "epoch": 0.03569690085997079, "step": 110, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.4204524632077664e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.8580105461296625e-05 }, { "epoch": 0.03569690085997079, "step": 110, "weight_max": 3.4268457966391e-05, "weight_mean": 8.110624548862688e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.5048367458803114e-05 }, { "epoch": 0.03569690085997079, "step": 110, "weight_max": 3.36925259034615e-05, "weight_mean": 4.3555451156862546e-06, "weight_min": 0.0, "weight_std": 1.186063491331879e-05 }, { "epoch": 0.03569690085997079, "step": 110, "weight_max": 3.752332122530788e-05, "weight_mean": 1.4972843928262591e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.806517320801504e-05 }, { "epoch": 0.03602141814051598, "step": 111, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.656436506891623e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.858559699030593e-05 }, { "epoch": 0.03602141814051598, "step": 111, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.84145959187299e-06, "weight_min": 0.0, "weight_std": 1.5088926375028677e-05 }, { "epoch": 0.03602141814051598, "step": 111, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7630689399084076e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8274697140441276e-05 }, { "epoch": 0.03602141814051598, "step": 111, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076934485259699e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357689906319138e-05 }, { "epoch": 0.036345935421061175, "step": 112, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4688901501358487e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0285791833885014e-05 }, { "epoch": 0.036345935421061175, "step": 112, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843594524572836e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3697633221454453e-05 }, { "epoch": 0.036345935421061175, "step": 112, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5158575479290448e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0952227714587934e-05 }, { "epoch": 0.036345935421061175, "step": 112, "weight_max": 3.22740561387036e-05, "weight_mean": 4.0351042116526514e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.141026041295845e-05 }, { "epoch": 0.03667045270160636, "step": 113, "weight_max": 3.397988257347606e-05, "weight_mean": 8.318087566294707e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5403877114295028e-05 }, { "epoch": 0.03667045270160636, "step": 113, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4273895430960692e-05, "weight_min": 0.0, "weight_std": 1.975548184418585e-05 }, { "epoch": 0.03667045270160636, "step": 113, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.562800389365293e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1579549866146408e-05 }, { "epoch": 0.03667045270160636, "step": 113, "weight_max": 1.2650940334424376e-05, "weight_mean": 1.6400803133365116e-06, "weight_min": 0.0, "weight_std": 4.452032499102643e-06 }, { "epoch": 0.03699496998215155, "step": 114, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2808708308730274e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.932507802848704e-05 }, { "epoch": 0.03699496998215155, "step": 114, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.850941412674729e-06, "weight_min": 0.0, "weight_std": 1.6522250007255934e-05 }, { "epoch": 0.03699496998215155, "step": 114, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5039418030937668e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0782048522960395e-05 }, { "epoch": 0.03699496998215155, "step": 114, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396706455940148e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5261875887517817e-05 }, { "epoch": 0.03731948726269674, "step": 115, "weight_max": 3.998322426923551e-05, "weight_mean": 5.0016710702038836e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4134684533928521e-05 }, { "epoch": 0.03731948726269674, "step": 115, "weight_max": 3.1715208024252206e-05, "weight_mean": 6.2739445638726465e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.214285657624714e-05 }, { "epoch": 0.03731948726269674, "step": 115, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2174008588772267e-05, "weight_min": 0.0, "weight_std": 1.7615970136830583e-05 }, { "epoch": 0.03731948726269674, "step": 115, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233385182800703e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8953947801492177e-05 }, { "epoch": 0.037644004543241925, "step": 116, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.45305484693381e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.005728310905397e-05 }, { "epoch": 0.037644004543241925, "step": 116, "weight_max": 4.252122016623616e-05, "weight_mean": 6.726235824316973e-06, "weight_min": 0.0, "weight_std": 1.4992628166510258e-05 }, { "epoch": 0.037644004543241925, "step": 116, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0486137398402207e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9394450646359473e-05 }, { "epoch": 0.037644004543241925, "step": 116, "weight_max": 3.752332122530788e-05, "weight_mean": 9.978823982237373e-06, "weight_min": 0.0, "weight_std": 1.577208240632899e-05 }, { "epoch": 0.03796852182378712, "step": 117, "weight_max": 1.2882609423670033e-09, "weight_mean": 6.565909482780796e-10, "weight_min": 0.0, "weight_std": 3.457199804213218e-10 }, { "epoch": 0.03796852182378712, "step": 117, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.07715594721958e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357600775838364e-05 }, { "epoch": 0.03796852182378712, "step": 117, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2257490197953302e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.759700899128802e-05 }, { "epoch": 0.03796852182378712, "step": 117, "weight_max": 4.187939703115262e-05, "weight_mean": 8.444708328170236e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6216181393247098e-05 }, { "epoch": 0.03829303910433231, "step": 118, "weight_max": 3.843663216684945e-05, "weight_mean": 1.4100954103923868e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.850828448368702e-05 }, { "epoch": 0.03829303910433231, "step": 118, "weight_max": 2.4786623864514468e-09, "weight_mean": 9.411593726582623e-10, "weight_min": 5.864964625068581e-10, "weight_std": 6.277279807065383e-10 }, { "epoch": 0.03829303910433231, "step": 118, "weight_max": 4.187939703115262e-05, "weight_mean": 7.286477739398833e-06, "weight_min": 0.0, "weight_std": 1.5111023458302952e-05 }, { "epoch": 0.03829303910433231, "step": 118, "weight_max": 4.187939703115262e-05, "weight_mean": 1.039095877786167e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9239276298321784e-05 }, { "epoch": 0.038617556384877495, "step": 119, "weight_max": 1.987018549698405e-09, "weight_mean": 9.253196542324815e-10, "weight_min": 6.053385570581327e-10, "weight_std": 4.658850527583525e-10 }, { "epoch": 0.038617556384877495, "step": 119, "weight_max": 4.252122016623616e-05, "weight_mean": 8.808103302726522e-06, "weight_min": 0.0, "weight_std": 1.6767455235822126e-05 }, { "epoch": 0.038617556384877495, "step": 119, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998778877052246e-06, "weight_min": 0.0, "weight_std": 1.4135851415630896e-05 }, { "epoch": 0.038617556384877495, "step": 119, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1943043975625187e-05, "weight_min": 8.267048712262692e-10, "weight_std": 1.8856186215998605e-05 }, { "epoch": 0.03894207366542268, "grad_norm": 1.534057378768921, "learning_rate": 9.96326530612245e-06, "loss": 0.3396, "step": 120 }, { "epoch": 0.03894207366542268, "step": 120, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.893799526442308e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0779330952791497e-05 }, { "epoch": 0.03894207366542268, "step": 120, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.92218338069506e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.838988100644201e-05 }, { "epoch": 0.03894207366542268, "step": 120, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392291187599767e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9247490854468197e-05 }, { "epoch": 0.03894207366542268, "step": 120, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315667294780724e-06, "weight_min": 0.0, "weight_std": 1.503331441199407e-05 }, { "epoch": 0.03926659094596787, "step": 121, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.523771399864927e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0955572836101055e-05 }, { "epoch": 0.03926659094596787, "step": 121, "weight_max": 3.8743506593164057e-05, "weight_mean": 7.685104719712399e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4856545931252185e-05 }, { "epoch": 0.03926659094596787, "step": 121, "weight_max": 3.602553260861896e-05, "weight_mean": 4.503826858126558e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2736692042381037e-05 }, { "epoch": 0.03926659094596787, "step": 121, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.668714144732803e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.635925036680419e-05 }, { "epoch": 0.039591108226513065, "step": 122, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2815217814932112e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8766799257718958e-05 }, { "epoch": 0.039591108226513065, "step": 122, "weight_max": 2.7939937353949063e-05, "weight_mean": 3.49309857483604e-06, "weight_min": 0.0, "weight_std": 9.87801467999816e-06 }, { "epoch": 0.039591108226513065, "step": 122, "weight_max": 3.255531191825867e-05, "weight_mean": 8.215643902076408e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4939331776986364e-05 }, { "epoch": 0.039591108226513065, "step": 122, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5351664842455648e-05, "weight_min": 0.0, "weight_std": 2.1204468794167042e-05 }, { "epoch": 0.03991562550705825, "step": 123, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2400386367517058e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.717995291983243e-05 }, { "epoch": 0.03991562550705825, "step": 123, "weight_max": 1.982200501515763e-06, "weight_mean": 2.4828273126331624e-07, "weight_min": 0.0, "weight_std": 7.006087230365665e-07 }, { "epoch": 0.03991562550705825, "step": 123, "weight_max": 4.252122016623616e-05, "weight_mean": 2.36444848269457e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9906232409994118e-05 }, { "epoch": 0.03991562550705825, "step": 123, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23674680152908e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4805866157985292e-05 }, { "epoch": 0.04024014278760344, "step": 124, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235488060861826e-06, "weight_min": 0.0, "weight_std": 1.4806375475018285e-05 }, { "epoch": 0.04024014278760344, "step": 124, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1565211833803914e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4580920833395794e-05 }, { "epoch": 0.04024014278760344, "step": 124, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0771072892530356e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357619875227101e-05 }, { "epoch": 0.04024014278760344, "step": 124, "weight_max": 3.602553260861896e-05, "weight_mean": 8.71530210133642e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.6148300346685573e-05 }, { "epoch": 0.04056466006814863, "step": 125, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4276734873419628e-05, "weight_min": 0.0, "weight_std": 1.9770508515648544e-05 }, { "epoch": 0.04056466006814863, "step": 125, "weight_max": 3.843663216684945e-05, "weight_mean": 9.382788448419888e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7378877601004206e-05 }, { "epoch": 0.04056466006814863, "step": 125, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9212485464522615e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3915730960434303e-05 }, { "epoch": 0.04056466006814863, "step": 125, "weight_max": 3.003409743129737e-09, "weight_mean": 1.3678846944031875e-09, "weight_min": 6.053385570581327e-10, "weight_std": 9.086358887522294e-10 }, { "epoch": 0.040889177348693816, "step": 126, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0194848982791882e-05, "weight_min": 0.0, "weight_std": 1.888396764115896e-05 }, { "epoch": 0.040889177348693816, "step": 126, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.475812476011924e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.763472937454935e-05 }, { "epoch": 0.040889177348693816, "step": 126, "weight_max": 2.639517333591357e-05, "weight_mean": 3.3021080980688566e-06, "weight_min": 0.0, "weight_std": 9.33100818656385e-06 }, { "epoch": 0.040889177348693816, "step": 126, "weight_max": 4.187939703115262e-05, "weight_mean": 5.242225142865209e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4803663361817598e-05 }, { "epoch": 0.04121369462923901, "step": 127, "weight_max": 4.252122016623616e-05, "weight_mean": 1.450805393687915e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0236955606378615e-05 }, { "epoch": 0.04121369462923901, "step": 127, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.131862311216537e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.1263649614411406e-05 }, { "epoch": 0.04121369462923901, "step": 127, "weight_max": 3.998322426923551e-05, "weight_mean": 9.802792192203924e-06, "weight_min": 0.0, "weight_std": 1.8155184079660103e-05 }, { "epoch": 0.04121369462923901, "step": 127, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.7647149434196763e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.897832225949969e-05 }, { "epoch": 0.0415382119097842, "step": 128, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3655331713380292e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9156548660248518e-05 }, { "epoch": 0.0415382119097842, "step": 128, "weight_max": 3.782653220696375e-05, "weight_mean": 9.232512638845947e-06, "weight_min": 0.0, "weight_std": 1.709958814899437e-05 }, { "epoch": 0.0415382119097842, "step": 128, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140805685892701e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1709960745065473e-05 }, { "epoch": 0.0415382119097842, "step": 128, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3913302609580569e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9329892893438227e-05 }, { "epoch": 0.041862729190329385, "step": 129, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4100993212196045e-05, "weight_min": 0.0, "weight_std": 1.974219412659295e-05 }, { "epoch": 0.041862729190329385, "step": 129, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.015363432088634e-05, "weight_min": 0.0, "weight_std": 1.880284253275022e-05 }, { "epoch": 0.041862729190329385, "step": 129, "weight_max": 1.4159473593622351e-09, "weight_mean": 9.196129968636058e-10, "weight_min": 5.771870759119224e-10, "weight_std": 2.817644451091894e-10 }, { "epoch": 0.041862729190329385, "step": 129, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631889381329529e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.968703872989863e-05 }, { "epoch": 0.04218724647087457, "grad_norm": 4.102730751037598, "learning_rate": 9.942857142857145e-06, "loss": 0.1748, "step": 130 }, { "epoch": 0.04218724647087457, "step": 130, "weight_max": 4.187939703115262e-05, "weight_mean": 9.664951903687324e-06, "weight_min": 0.0, "weight_std": 1.797708864614833e-05 }, { "epoch": 0.04218724647087457, "step": 130, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.2997532394365408e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.794359013729263e-05 }, { "epoch": 0.04218724647087457, "step": 130, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.406491254689172e-06, "weight_min": 0.0, "weight_std": 1.525794777990086e-05 }, { "epoch": 0.04218724647087457, "step": 130, "weight_max": 2.2490561605081894e-05, "weight_mean": 6.341216248983983e-06, "weight_min": 6.390288853630466e-10, "weight_std": 9.953440894605592e-06 }, { "epoch": 0.04251176375141976, "step": 131, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396638698584866e-06, "weight_min": 0.0, "weight_std": 1.526190317235887e-05 }, { "epoch": 0.04251176375141976, "step": 131, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.3702087825513445e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4506965271721128e-05 }, { "epoch": 0.04251176375141976, "step": 131, "weight_max": 4.187939703115262e-05, "weight_mean": 1.141619577538222e-05, "weight_min": 0.0, "weight_std": 1.7177455447381362e-05 }, { "epoch": 0.04251176375141976, "step": 131, "weight_max": 3.967146767536178e-05, "weight_mean": 9.892146408674307e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.828457425290253e-05 }, { "epoch": 0.042836281031964955, "step": 132, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.081872586742975e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.435570084140636e-05 }, { "epoch": 0.042836281031964955, "step": 132, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.985411113826558e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6869857063284144e-05 }, { "epoch": 0.042836281031964955, "step": 132, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9118713680654764e-05, "weight_min": 0.0, "weight_std": 2.0663503164541908e-05 }, { "epoch": 0.042836281031964955, "step": 132, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.447971590096131e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0097904780413955e-05 }, { "epoch": 0.04316079831251014, "step": 133, "weight_max": 1.901931279135738e-09, "weight_mean": 8.413623131531267e-10, "weight_min": 6.100955851628953e-10, "weight_std": 4.371243644829548e-10 }, { "epoch": 0.04316079831251014, "step": 133, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390946954430547e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9239281755289994e-05 }, { "epoch": 0.04316079831251014, "step": 133, "weight_max": 3.0203543133211497e-07, "weight_mean": 3.8810483715678856e-08, "weight_min": 6.053385570581327e-10, "weight_std": 1.0636311742473481e-07 }, { "epoch": 0.04316079831251014, "step": 133, "weight_max": 3.312147964606993e-05, "weight_mean": 4.141701083426597e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.170959967566887e-05 }, { "epoch": 0.04348531559305533, "step": 134, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.2821881429990754e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9241499103372917e-05 }, { "epoch": 0.04348531559305533, "step": 134, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.660150681156665e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.668229560891632e-05 }, { "epoch": 0.04348531559305533, "step": 134, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.737175989546813e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8802897102432325e-05 }, { "epoch": 0.04348531559305533, "step": 134, "weight_max": 4.187939703115262e-05, "weight_mean": 9.709654477774166e-06, "weight_min": 0.0, "weight_std": 1.7950049368664622e-05 }, { "epoch": 0.04380983287360052, "step": 135, "weight_max": 4.252122016623616e-05, "weight_mean": 1.967453681572806e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.116569157806225e-05 }, { "epoch": 0.04380983287360052, "step": 135, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076887646282557e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357709005707875e-05 }, { "epoch": 0.04380983287360052, "step": 135, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.12992618395947e-06, "weight_min": 0.0, "weight_std": 1.695715582172852e-05 }, { "epoch": 0.04380983287360052, "step": 135, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4827131053607445e-05, "weight_min": 0.0, "weight_std": 2.0579862393788062e-05 }, { "epoch": 0.044134350154145706, "step": 136, "weight_max": 3.572963396436535e-05, "weight_mean": 9.720145499159116e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5273320968844928e-05 }, { "epoch": 0.044134350154145706, "step": 136, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.331613000540528e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7404005120624788e-05 }, { "epoch": 0.044134350154145706, "step": 136, "weight_max": 3.199401908204891e-05, "weight_mean": 4.020839242002694e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.130302462115651e-05 }, { "epoch": 0.044134350154145706, "step": 136, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430137778399512e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.252784022653941e-05 }, { "epoch": 0.0444588674346909, "step": 137, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5397103197756223e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.129222957591992e-05 }, { "epoch": 0.0444588674346909, "step": 137, "weight_max": 4.252122016623616e-05, "weight_mean": 2.342904372198973e-05, "weight_min": 7.721521200210191e-10, "weight_std": 1.963122485904023e-05 }, { "epoch": 0.0444588674346909, "step": 137, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0001526788983028e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8523172911955044e-05 }, { "epoch": 0.0444588674346909, "step": 137, "weight_max": 4.187939703115262e-05, "weight_mean": 7.115454536688048e-06, "weight_min": 0.0, "weight_std": 1.4999977793195285e-05 }, { "epoch": 0.04478338471523609, "step": 138, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.48471144711948e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.052795207418967e-05 }, { "epoch": 0.04478338471523609, "step": 138, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396618689701427e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5261910448316485e-05 }, { "epoch": 0.04478338471523609, "step": 138, "weight_max": 3.1316551574178675e-09, "weight_mean": 1.0610914369379998e-09, "weight_min": 5.958802340444436e-10, "weight_std": 8.540915197308152e-10 }, { "epoch": 0.04478338471523609, "step": 138, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311591722711455e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9090946807409637e-05 }, { "epoch": 0.045107901995781276, "step": 139, "weight_max": 4.187939703115262e-05, "weight_mean": 9.926719940267503e-06, "weight_min": 0.0, "weight_std": 1.8414146325085312e-05 }, { "epoch": 0.045107901995781276, "step": 139, "weight_max": 3.813097282545641e-05, "weight_mean": 1.1557982361409813e-05, "weight_min": 7.721521200210191e-10, "weight_std": 1.630525730433874e-05 }, { "epoch": 0.045107901995781276, "step": 139, "weight_max": 3.860909814079605e-09, "weight_mean": 1.0170488895511198e-09, "weight_min": 0.0, "weight_std": 1.1785096232230785e-09 }, { "epoch": 0.045107901995781276, "step": 139, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.709145741799148e-06, "weight_min": 0.0, "weight_std": 1.221837828779826e-05 }, { "epoch": 0.04543241927632646, "grad_norm": 6.351010322570801, "learning_rate": 9.922448979591838e-06, "loss": 0.2391, "step": 140 }, { "epoch": 0.04543241927632646, "step": 140, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070107934239786e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1509760952321813e-05 }, { "epoch": 0.04543241927632646, "step": 140, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7773516447050497e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8675875253393315e-05 }, { "epoch": 0.04543241927632646, "step": 140, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8851464119507e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0362571376608685e-05 }, { "epoch": 0.04543241927632646, "step": 140, "weight_max": 3.813097282545641e-05, "weight_mean": 1.675713610893581e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8264088794239797e-05 }, { "epoch": 0.04575693655687165, "step": 141, "weight_max": 3.722132169059478e-05, "weight_mean": 9.305827006755862e-06, "weight_min": 0.0, "weight_std": 1.7229816876351833e-05 }, { "epoch": 0.04575693655687165, "step": 141, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.46941920320387e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.032443808275275e-05 }, { "epoch": 0.04575693655687165, "step": 141, "weight_max": 3.143761568935588e-05, "weight_mean": 6.293023034231737e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2122312000428792e-05 }, { "epoch": 0.04575693655687165, "step": 141, "weight_max": 4.187939703115262e-05, "weight_mean": 8.860318303050008e-06, "weight_min": 0.0, "weight_std": 1.6760943253757432e-05 }, { "epoch": 0.046081453837416846, "step": 142, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.473126551194582e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0273886548238806e-05 }, { "epoch": 0.046081453837416846, "step": 142, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.519780749338679e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6231088011409156e-05 }, { "epoch": 0.046081453837416846, "step": 142, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9328153939568438e-05, "weight_min": 8.267048712262692e-10, "weight_std": 2.0794212105101906e-05 }, { "epoch": 0.046081453837416846, "step": 142, "weight_max": 3.662098970380612e-05, "weight_mean": 4.5810247684130445e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2946103197464254e-05 }, { "epoch": 0.04640597111796203, "step": 143, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.23338120628614e-06, "weight_min": 0.0, "weight_std": 1.7232654499821365e-05 }, { "epoch": 0.04640597111796203, "step": 143, "weight_max": 3.199401908204891e-05, "weight_mean": 1.596154470462352e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6700167179806158e-05 }, { "epoch": 0.04640597111796203, "step": 143, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.111119652516209e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7015623598126695e-05 }, { "epoch": 0.04640597111796203, "step": 143, "weight_max": 3.843663216684945e-05, "weight_mean": 9.382529242429882e-06, "weight_min": 0.0, "weight_std": 1.7379037672071718e-05 }, { "epoch": 0.04673048839850722, "step": 144, "weight_max": 3.578502383305704e-09, "weight_mean": 1.7226198245623436e-09, "weight_min": 6.390288853630466e-10, "weight_std": 1.184886189165013e-09 }, { "epoch": 0.04673048839850722, "step": 144, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8277922208653763e-05, "weight_min": 0.0, "weight_std": 1.9724913727259263e-05 }, { "epoch": 0.04673048839850722, "step": 144, "weight_max": 4.252122016623616e-05, "weight_mean": 6.101367034716532e-06, "weight_min": 0.0, "weight_std": 1.4879259651934262e-05 }, { "epoch": 0.04673048839850722, "step": 144, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430534318089485e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2527680155471899e-05 }, { "epoch": 0.04705500567905241, "step": 145, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805320259038126e-06, "weight_min": 0.0, "weight_std": 1.358910321869189e-05 }, { "epoch": 0.04705500567905241, "step": 145, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7932541090412997e-06, "weight_min": 0.0, "weight_std": 1.072628037945833e-05 }, { "epoch": 0.04705500567905241, "step": 145, "weight_max": 3.782653220696375e-05, "weight_mean": 9.306473657488823e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7233936887350865e-05 }, { "epoch": 0.04705500567905241, "step": 145, "weight_max": 3.813097282545641e-05, "weight_mean": 4.97534756505047e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3409552593657281e-05 }, { "epoch": 0.047379522959597596, "step": 146, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1544712581089698e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.95503616851056e-05 }, { "epoch": 0.047379522959597596, "step": 146, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1028167136828415e-05, "weight_min": 0.0, "weight_std": 1.8188969988841563e-05 }, { "epoch": 0.047379522959597596, "step": 146, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5317098586820066e-05, "weight_min": 0.0, "weight_std": 2.117613621521741e-05 }, { "epoch": 0.047379522959597596, "step": 146, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4225033737602644e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9802366296062246e-05 }, { "epoch": 0.04770404024014279, "step": 147, "weight_max": 3.998322426923551e-05, "weight_mean": 9.957694601325784e-06, "weight_min": 0.0, "weight_std": 1.8436139725963585e-05 }, { "epoch": 0.04770404024014279, "step": 147, "weight_max": 4.187939703115262e-05, "weight_mean": 1.929460995597765e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0925972421537153e-05 }, { "epoch": 0.04770404024014279, "step": 147, "weight_max": 3.0886084800840763e-09, "weight_mean": 1.0085902113488032e-09, "weight_min": 0.0, "weight_std": 9.287662861012791e-10 }, { "epoch": 0.04770404024014279, "step": 147, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9821560272248462e-05, "weight_min": 0.0, "weight_std": 2.1378331439336762e-05 }, { "epoch": 0.04802855752068798, "step": 148, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2649060408875812e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8684657334233634e-05 }, { "epoch": 0.04802855752068798, "step": 148, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.671570134931244e-05, "weight_min": 0.0, "weight_std": 1.8068885765387677e-05 }, { "epoch": 0.04802855752068798, "step": 148, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3787855095870327e-05, "weight_min": 7.721521200210191e-10, "weight_std": 1.9545541363186203e-05 }, { "epoch": 0.04802855752068798, "step": 148, "weight_max": 3.967146767536178e-05, "weight_mean": 1.855835762398783e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9934295778512023e-05 }, { "epoch": 0.048353074801233166, "step": 149, "weight_max": 3.143761568935588e-05, "weight_mean": 7.587844265799504e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4060657122172415e-05 }, { "epoch": 0.048353074801233166, "step": 149, "weight_max": 4.252122016623616e-05, "weight_mean": 1.178316324512707e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.821172008931171e-05 }, { "epoch": 0.048353074801233166, "step": 149, "weight_max": 2.674530374946471e-09, "weight_mean": 9.866715222628386e-10, "weight_min": 5.771870759119224e-10, "weight_std": 7.505745469593705e-10 }, { "epoch": 0.048353074801233166, "step": 149, "weight_max": 4.252122016623616e-05, "weight_mean": 2.2578027710551396e-05, "weight_min": 0.0, "weight_std": 1.973296639334876e-05 }, { "epoch": 0.048677592081778354, "grad_norm": 3.9809370040893555, "learning_rate": 9.902040816326531e-06, "loss": 0.2281, "step": 150 }, { "epoch": 0.048677592081778354, "step": 150, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0969290087814443e-05, "weight_min": 0.0, "weight_std": 1.716150472930167e-05 }, { "epoch": 0.048677592081778354, "step": 150, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767203336086823e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3480999768944457e-05 }, { "epoch": 0.048677592081778354, "step": 150, "weight_max": 3.998322426923551e-05, "weight_mean": 9.613540896680206e-06, "weight_min": 0.0, "weight_std": 1.7818223568610847e-05 }, { "epoch": 0.048677592081778354, "step": 150, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1160494977957569e-05, "weight_min": 0.0, "weight_std": 1.6113346646307036e-05 }, { "epoch": 0.04900210936232354, "step": 151, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8859418560168706e-05, "weight_min": 0.0, "weight_std": 2.041171865130309e-05 }, { "epoch": 0.04900210936232354, "step": 151, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3861490515409969e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9191471437807195e-05 }, { "epoch": 0.04900210936232354, "step": 151, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1721001101250295e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8789083696901798e-05 }, { "epoch": 0.04900210936232354, "step": 151, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0221699085377622e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8757271391223185e-05 }, { "epoch": 0.049326626642868736, "step": 152, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.648393845243845e-05, "weight_min": 0.0, "weight_std": 1.8062066374113783e-05 }, { "epoch": 0.049326626642868736, "step": 152, "weight_max": 3.752332122530788e-05, "weight_mean": 4.69119231638615e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3266183486848604e-05 }, { "epoch": 0.049326626642868736, "step": 152, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.459470675210468e-06, "weight_min": 0.0, "weight_std": 1.7526599549455568e-05 }, { "epoch": 0.049326626642868736, "step": 152, "weight_max": 4.252122016623616e-05, "weight_mean": 1.497419270890532e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.074684925901238e-05 }, { "epoch": 0.04965114392341392, "step": 153, "weight_max": 4.252122016623616e-05, "weight_mean": 1.000661177386064e-05, "weight_min": 0.0, "weight_std": 1.8574159184936434e-05 }, { "epoch": 0.04965114392341392, "step": 153, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430246463016374e-06, "weight_min": 0.0, "weight_std": 1.2527796570793726e-05 }, { "epoch": 0.04965114392341392, "step": 153, "weight_max": 4.187939703115262e-05, "weight_mean": 1.619403155928012e-05, "weight_min": 0.0, "weight_std": 1.9038965547224507e-05 }, { "epoch": 0.04965114392341392, "step": 153, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.60573800309794e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5942547179292887e-05 }, { "epoch": 0.04997566120395911, "step": 154, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076973138784524e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357674444909208e-05 }, { "epoch": 0.04997566120395911, "step": 154, "weight_max": 3.397988257347606e-05, "weight_mean": 8.143276318151038e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5095655726327095e-05 }, { "epoch": 0.04997566120395911, "step": 154, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0706720533780754e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.2159905711305328e-05 }, { "epoch": 0.04997566120395911, "step": 154, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.286595402751118e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.535664705443196e-05 }, { "epoch": 0.0503001784845043, "step": 155, "weight_max": 4.187939703115262e-05, "weight_mean": 2.361779479542747e-05, "weight_min": 0.0, "weight_std": 1.977382999029942e-05 }, { "epoch": 0.0503001784845043, "step": 155, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.377261989749968e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9758470443775877e-05 }, { "epoch": 0.0503001784845043, "step": 155, "weight_max": 3.998322426923551e-05, "weight_mean": 2.288157156726811e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.908058038679883e-05 }, { "epoch": 0.0503001784845043, "step": 155, "weight_max": 9.413956281179026e-10, "weight_mean": 6.924854023537819e-10, "weight_min": 5.771870759119224e-10, "weight_std": 1.13573830895497e-10 }, { "epoch": 0.050624695765049486, "step": 156, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805372554983478e-06, "weight_min": 0.0, "weight_std": 1.3589081390819047e-05 }, { "epoch": 0.050624695765049486, "step": 156, "weight_max": 3.998322426923551e-05, "weight_mean": 9.42792757996358e-06, "weight_min": 0.0, "weight_std": 1.7497788576292805e-05 }, { "epoch": 0.050624695765049486, "step": 156, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1274032658548094e-05, "weight_min": 0.0, "weight_std": 1.63230379257584e-05 }, { "epoch": 0.050624695765049486, "step": 156, "weight_max": 3.998322426923551e-05, "weight_mean": 5.090866125101456e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.410099048371194e-05 }, { "epoch": 0.05094921304559468, "step": 157, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3891823073208798e-05, "weight_min": 0.0, "weight_std": 1.921100374602247e-05 }, { "epoch": 0.05094921304559468, "step": 157, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077145033283159e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357604413817171e-05 }, { "epoch": 0.05094921304559468, "step": 157, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1691977306327317e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8448594346409664e-05 }, { "epoch": 0.05094921304559468, "step": 157, "weight_max": 1.2064879095774472e-09, "weight_mean": 7.159960957459077e-10, "weight_min": 0.0, "weight_std": 3.783838242732429e-10 }, { "epoch": 0.05127373032613987, "step": 158, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397090717451647e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5261719454429112e-05 }, { "epoch": 0.05127373032613987, "step": 158, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.88076192495646e-06, "weight_min": 0.0, "weight_std": 1.6724228771636263e-05 }, { "epoch": 0.05127373032613987, "step": 158, "weight_max": 4.252122016623616e-05, "weight_mean": 1.383642666041851e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.937527304107789e-05 }, { "epoch": 0.05127373032613987, "step": 158, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.4189638022799045e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9601295207394287e-05 }, { "epoch": 0.051598247606685056, "step": 159, "weight_max": 5.628989470096712e-07, "weight_mean": 7.578984195788507e-08, "weight_min": 0.0, "weight_std": 1.972299088492946e-07 }, { "epoch": 0.051598247606685056, "step": 159, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4668269614048768e-05, "weight_min": 0.0, "weight_std": 2.034100361925084e-05 }, { "epoch": 0.051598247606685056, "step": 159, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9117534975521266e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0656203560065478e-05 }, { "epoch": 0.051598247606685056, "step": 159, "weight_max": 1.1403916699848082e-09, "weight_mean": 7.824738634809592e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.1358392832127038e-10 }, { "epoch": 0.051922764887230244, "grad_norm": 3.0776946544647217, "learning_rate": 9.881632653061225e-06, "loss": 0.2904, "step": 160 }, { "epoch": 0.051922764887230244, "step": 160, "weight_max": 2.639517333591357e-05, "weight_mean": 3.3004455417540157e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.331679393653758e-06 }, { "epoch": 0.051922764887230244, "step": 160, "weight_max": 3.061215829802677e-05, "weight_mean": 3.82721918867901e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0822749572980683e-05 }, { "epoch": 0.051922764887230244, "step": 160, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2155977856309619e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8118655134458095e-05 }, { "epoch": 0.051922764887230244, "step": 160, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.683474485820625e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0037990907439962e-05 }, { "epoch": 0.05224728216777543, "step": 161, "weight_max": 3.4849264920921996e-05, "weight_mean": 7.4377098826516885e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.403802161803469e-05 }, { "epoch": 0.05224728216777543, "step": 161, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.881010217010044e-06, "weight_min": 0.0, "weight_std": 1.6551506632822566e-05 }, { "epoch": 0.05224728216777543, "step": 161, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4983219443820417e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.951116428244859e-05 }, { "epoch": 0.05224728216777543, "step": 161, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.548750313115306e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7754020518623292e-05 }, { "epoch": 0.052571799448320626, "step": 162, "weight_max": 3.116124207735993e-05, "weight_mean": 5.717055501008872e-06, "weight_min": 0.0, "weight_std": 1.0821815521921962e-05 }, { "epoch": 0.052571799448320626, "step": 162, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236430297372863e-06, "weight_min": 0.0, "weight_std": 1.4805994396738242e-05 }, { "epoch": 0.052571799448320626, "step": 162, "weight_max": 4.252122016623616e-05, "weight_mean": 6.395518994395388e-06, "weight_min": 0.0, "weight_std": 1.4906733667885419e-05 }, { "epoch": 0.052571799448320626, "step": 162, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550813385634683e-05, "weight_min": 0.0, "weight_std": 1.9535245883162133e-05 }, { "epoch": 0.052896316728865814, "step": 163, "weight_max": 4.381950566312298e-05, "weight_mean": 1.76681496668607e-05, "weight_min": 0.0, "weight_std": 1.943602728715632e-05 }, { "epoch": 0.052896316728865814, "step": 163, "weight_max": 1.1118992006231565e-05, "weight_mean": 1.3909827885072445e-06, "weight_min": 5.864964625068581e-10, "weight_std": 3.9307092265516985e-06 }, { "epoch": 0.052896316728865814, "step": 163, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0189060958509799e-05, "weight_min": 0.0, "weight_std": 1.75230252352776e-05 }, { "epoch": 0.052896316728865814, "step": 163, "weight_max": 3.998322426923551e-05, "weight_mean": 8.234828783315606e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.570464883116074e-05 }, { "epoch": 0.053220834009411, "step": 164, "weight_max": 3.998322426923551e-05, "weight_mean": 7.307407031476032e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4700144674861804e-05 }, { "epoch": 0.053220834009411, "step": 164, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316052011039574e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5033157069410663e-05 }, { "epoch": 0.053220834009411, "step": 164, "weight_max": 3.998322426923551e-05, "weight_mean": 9.464611139264889e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7560645574121736e-05 }, { "epoch": 0.053220834009411, "step": 164, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5167610147036612e-05, "weight_min": 0.0, "weight_std": 2.109176966769155e-05 }, { "epoch": 0.05354535128995619, "step": 165, "weight_max": 3.998322426923551e-05, "weight_mean": 9.39107485464774e-06, "weight_min": 0.0, "weight_std": 1.743578104651533e-05 }, { "epoch": 0.05354535128995619, "step": 165, "weight_max": 7.894064424363023e-08, "weight_mean": 1.1230447682919475e-08, "weight_min": 0.0, "weight_std": 2.7523238799176397e-08 }, { "epoch": 0.05354535128995619, "step": 165, "weight_max": 2.517239972021912e-09, "weight_mean": 9.671836664892908e-10, "weight_min": 5.771870759119224e-10, "weight_std": 6.408282238190566e-10 }, { "epoch": 0.05354535128995619, "step": 165, "weight_max": 4.252122016623616e-05, "weight_mean": 5.5569644246133976e-06, "weight_min": 0.0, "weight_std": 1.4951070625102147e-05 }, { "epoch": 0.05386986857050138, "step": 166, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236245669948403e-06, "weight_min": 0.0, "weight_std": 1.4806068975303788e-05 }, { "epoch": 0.05386986857050138, "step": 166, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.326610779680777e-06, "weight_min": 0.0, "weight_std": 1.7549553376738913e-05 }, { "epoch": 0.05386986857050138, "step": 166, "weight_max": 1.7424939869670197e-05, "weight_mean": 2.1799276055389782e-06, "weight_min": 6.586902134841921e-10, "weight_std": 6.159915756143164e-06 }, { "epoch": 0.05386986857050138, "step": 166, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959649686497869e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4025693417352159e-05 }, { "epoch": 0.05419438585104657, "step": 167, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3701146599487402e-05, "weight_min": 0.0, "weight_std": 1.892785803647712e-05 }, { "epoch": 0.05419438585104657, "step": 167, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4539765288645867e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.010720163525548e-05 }, { "epoch": 0.05419438585104657, "step": 167, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.190333003061824e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.718328712740913e-05 }, { "epoch": 0.05419438585104657, "step": 167, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235852641926613e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8970484234159812e-05 }, { "epoch": 0.05451890313159176, "step": 168, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.72579097529524e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0534634384384844e-05 }, { "epoch": 0.05451890313159176, "step": 168, "weight_max": 4.252122016623616e-05, "weight_mean": 1.337687444902258e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8839933545677923e-05 }, { "epoch": 0.05451890313159176, "step": 168, "weight_max": 3.306819484905077e-09, "weight_mean": 1.2924081804754906e-09, "weight_min": 6.586902134841921e-10, "weight_std": 9.36803523643448e-10 }, { "epoch": 0.05451890313159176, "step": 168, "weight_max": 1.3161142176087992e-09, "weight_mean": 6.506062355526865e-10, "weight_min": 0.0, "weight_std": 3.57517043747535e-10 }, { "epoch": 0.054843420412136946, "step": 169, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.805161425902043e-06, "weight_min": 0.0, "weight_std": 1.8169526811107062e-05 }, { "epoch": 0.054843420412136946, "step": 169, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8055890147225e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3588993169832975e-05 }, { "epoch": 0.054843420412136946, "step": 169, "weight_max": 2.563926318543963e-05, "weight_mean": 3.7828212953172624e-06, "weight_min": 0.0, "weight_std": 8.977894140116405e-06 }, { "epoch": 0.054843420412136946, "step": 169, "weight_max": 3.662098970380612e-05, "weight_mean": 1.762318061082624e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8903041564044543e-05 }, { "epoch": 0.055167937692682134, "grad_norm": 0.7597931623458862, "learning_rate": 9.861224489795918e-06, "loss": 0.1796, "step": 170 }, { "epoch": 0.055167937692682134, "step": 170, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.664614481152967e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6882600903045386e-05 }, { "epoch": 0.055167937692682134, "step": 170, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8283859390066937e-05, "weight_min": 0.0, "weight_std": 1.9894421711796895e-05 }, { "epoch": 0.055167937692682134, "step": 170, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.477584348525852e-05, "weight_min": 0.0, "weight_std": 2.059235703200102e-05 }, { "epoch": 0.055167937692682134, "step": 170, "weight_max": 3.632264997577295e-05, "weight_mean": 4.540872851066524e-06, "weight_min": 0.0, "weight_std": 1.2841776879213285e-05 }, { "epoch": 0.05549245497322732, "step": 171, "weight_max": 2.5890016331686638e-05, "weight_mean": 3.2376365197706036e-06, "weight_min": 6.390288853630466e-10, "weight_std": 9.152943675871938e-06 }, { "epoch": 0.05549245497322732, "step": 171, "weight_max": 4.381950566312298e-05, "weight_mean": 2.7724821848096326e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.812572736525908e-05 }, { "epoch": 0.05549245497322732, "step": 171, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653343239624519e-06, "weight_min": 0.0, "weight_std": 1.3159450645616744e-05 }, { "epoch": 0.05549245497322732, "step": 171, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8651518985279836e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0116787709412165e-05 }, { "epoch": 0.055816972253772516, "step": 172, "weight_max": 3.998322426923551e-05, "weight_mean": 1.378629167447798e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9094242816208862e-05 }, { "epoch": 0.055816972253772516, "step": 172, "weight_max": 3.662098970380612e-05, "weight_mean": 9.007488188217394e-06, "weight_min": 0.0, "weight_std": 1.668041841185186e-05 }, { "epoch": 0.055816972253772516, "step": 172, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8866221580537967e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.043216045422014e-05 }, { "epoch": 0.055816972253772516, "step": 172, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.767405572347343e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8102828107657842e-05 }, { "epoch": 0.056141489534317704, "step": 173, "weight_max": 1.6359139465293993e-07, "weight_mean": 2.191604231427391e-08, "weight_min": 0.0, "weight_std": 5.7304720968431866e-08 }, { "epoch": 0.056141489534317704, "step": 173, "weight_max": 3.572963396436535e-05, "weight_mean": 8.137570148392115e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3880150618206244e-05 }, { "epoch": 0.056141489534317704, "step": 173, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.1158614142914303e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.5437672118423507e-05 }, { "epoch": 0.056141489534317704, "step": 173, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1368119885446504e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7760876289685257e-05 }, { "epoch": 0.05646600681486289, "step": 174, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9704128135344945e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.116377800120972e-05 }, { "epoch": 0.05646600681486289, "step": 174, "weight_max": 4.252122016623616e-05, "weight_mean": 6.5079557316494174e-06, "weight_min": 0.0, "weight_std": 1.4928877135389484e-05 }, { "epoch": 0.05646600681486289, "step": 174, "weight_max": 3.967146767536178e-05, "weight_mean": 9.65004255704116e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.787591645552311e-05 }, { "epoch": 0.05646600681486289, "step": 174, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4929555618437007e-05, "weight_min": 7.091466858177853e-10, "weight_std": 2.0662449969677255e-05 }, { "epoch": 0.05679052409540808, "step": 175, "weight_max": 4.252122016623616e-05, "weight_mean": 9.856052201939747e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8323502445127815e-05 }, { "epoch": 0.05679052409540808, "step": 175, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1833795472048223e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9316456018714234e-05 }, { "epoch": 0.05679052409540808, "step": 175, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1563170018198434e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581003597413655e-05 }, { "epoch": 0.05679052409540808, "step": 175, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.333753728744341e-06, "weight_min": 0.0, "weight_std": 1.4855025256110821e-05 }, { "epoch": 0.05711504137595327, "step": 176, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.883060228166869e-06, "weight_min": 0.0, "weight_std": 1.3806177776132245e-05 }, { "epoch": 0.05711504137595327, "step": 176, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5100935343070887e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.029126335401088e-05 }, { "epoch": 0.05711504137595327, "step": 176, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2842306205129717e-05, "weight_min": 0.0, "weight_std": 1.8138822269975208e-05 }, { "epoch": 0.05711504137595327, "step": 176, "weight_max": 4.252122016623616e-05, "weight_mean": 1.372367114527151e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.949632314790506e-05 }, { "epoch": 0.05743955865649846, "step": 177, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7120084521593526e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8554754205979407e-05 }, { "epoch": 0.05743955865649846, "step": 177, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9315945034322795e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3911583664594218e-05 }, { "epoch": 0.05743955865649846, "step": 177, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.686898920335807e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.793525916582439e-05 }, { "epoch": 0.05743955865649846, "step": 177, "weight_max": 2.407077559496429e-08, "weight_mean": 3.8490908238486554e-09, "weight_min": 5.679522963042416e-10, "weight_std": 8.202120760358866e-09 }, { "epoch": 0.05776407593704365, "step": 178, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313689926988445e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9107719708699733e-05 }, { "epoch": 0.05776407593704365, "step": 178, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3626116924569942e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8754790289676748e-05 }, { "epoch": 0.05776407593704365, "step": 178, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653278210753342e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3159477020963095e-05 }, { "epoch": 0.05776407593704365, "step": 178, "weight_max": 3.572963396436535e-05, "weight_mean": 1.5743336916784756e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.707700539554935e-05 }, { "epoch": 0.05808859321758884, "step": 179, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3088232410373166e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0354686057544313e-05 }, { "epoch": 0.05808859321758884, "step": 179, "weight_max": 1.7212929970256141e-09, "weight_mean": 9.133093170632378e-10, "weight_min": 0.0, "weight_std": 5.640402589435212e-10 }, { "epoch": 0.05808859321758884, "step": 179, "weight_max": 3.998322426923551e-05, "weight_mean": 1.0017729437095113e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8495198673917912e-05 }, { "epoch": 0.05808859321758884, "step": 179, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.3969761211192235e-06, "weight_min": 0.0, "weight_std": 1.5261766748153605e-05 }, { "epoch": 0.058413110498134024, "grad_norm": 1.7110168933868408, "learning_rate": 9.840816326530613e-06, "loss": 0.1511, "step": 180 }, { "epoch": 0.058413110498134024, "step": 180, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.767295523488428e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8102897229255177e-05 }, { "epoch": 0.058413110498134024, "step": 180, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.171907105250284e-06, "weight_min": 0.0, "weight_std": 1.4574771739717107e-05 }, { "epoch": 0.058413110498134024, "step": 180, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.3964595281286165e-06, "weight_min": 0.0, "weight_std": 1.5261975931935012e-05 }, { "epoch": 0.058413110498134024, "step": 180, "weight_max": 4.252122016623616e-05, "weight_mean": 5.318036073731491e-06, "weight_min": 0.0, "weight_std": 1.5032357623567805e-05 }, { "epoch": 0.05873762777867921, "step": 181, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156430233910214e-06, "weight_min": 0.0, "weight_std": 1.4580958122678567e-05 }, { "epoch": 0.05873762777867921, "step": 181, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156283805263229e-06, "weight_min": 0.0, "weight_std": 1.4581017239834182e-05 }, { "epoch": 0.05873762777867921, "step": 181, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.08959055878222e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4352619473356754e-05 }, { "epoch": 0.05873762777867921, "step": 181, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.680047696747351e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8079863366438076e-05 }, { "epoch": 0.05906214505922441, "step": 182, "weight_max": 3.782653220696375e-05, "weight_mean": 9.321468496636953e-06, "weight_min": 0.0, "weight_std": 1.7224732800968923e-05 }, { "epoch": 0.05906214505922441, "step": 182, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235677235759795e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806299077463336e-05 }, { "epoch": 0.05906214505922441, "step": 182, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805312983080512e-06, "weight_min": 0.0, "weight_std": 1.3589105037681293e-05 }, { "epoch": 0.05906214505922441, "step": 182, "weight_max": 3.813097282545641e-05, "weight_mean": 6.3487682382401545e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3583407053374685e-05 }, { "epoch": 0.059386662339769594, "step": 183, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4871387975290418e-05, "weight_min": 0.0, "weight_std": 2.066542219836265e-05 }, { "epoch": 0.059386662339769594, "step": 183, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9102175428997725e-05, "weight_min": 0.0, "weight_std": 2.098961158480961e-05 }, { "epoch": 0.059386662339769594, "step": 183, "weight_max": 4.187939703115262e-05, "weight_mean": 9.270721420762129e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.73531734617427e-05 }, { "epoch": 0.059386662339769594, "step": 183, "weight_max": 3.722132169059478e-05, "weight_mean": 7.29199064153363e-06, "weight_min": 0.0, "weight_std": 1.4170096619636752e-05 }, { "epoch": 0.05971117962031478, "step": 184, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0037494575954042e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8581942640594207e-05 }, { "epoch": 0.05971117962031478, "step": 184, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2889118806924671e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8993543562828563e-05 }, { "epoch": 0.05971117962031478, "step": 184, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.2592900020536035e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.750487535900902e-05 }, { "epoch": 0.05971117962031478, "step": 184, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3922847756475676e-05, "weight_min": 0.0, "weight_std": 1.95350239664549e-05 }, { "epoch": 0.06003569690085997, "step": 185, "weight_max": 4.187939703115262e-05, "weight_mean": 9.759209206094965e-06, "weight_min": 0.0, "weight_std": 1.8086353520629928e-05 }, { "epoch": 0.06003569690085997, "step": 185, "weight_max": 3.998322426923551e-05, "weight_mean": 9.880947800411377e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8293718312634155e-05 }, { "epoch": 0.06003569690085997, "step": 185, "weight_max": 1.0888551171817085e-09, "weight_mean": 7.089768772061689e-10, "weight_min": 0.0, "weight_std": 3.4585484476323813e-10 }, { "epoch": 0.06003569690085997, "step": 185, "weight_max": 4.252122016623616e-05, "weight_mean": 9.8189393611392e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.826237712521106e-05 }, { "epoch": 0.06036021418140516, "step": 186, "weight_max": 3.632264997577295e-05, "weight_mean": 5.066091034677811e-06, "weight_min": 0.0, "weight_std": 1.271488144993782e-05 }, { "epoch": 0.06036021418140516, "step": 186, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471740097273141e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9389746739761904e-05 }, { "epoch": 0.06036021418140516, "step": 186, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1555117453099228e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.8806087609846145e-05 }, { "epoch": 0.06036021418140516, "step": 186, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0849121180363e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4354478480527177e-05 }, { "epoch": 0.06068473146195035, "step": 187, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.012054028455168e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9884901121258736e-05 }, { "epoch": 0.06068473146195035, "step": 187, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0775479394360445e-06, "weight_min": 0.0, "weight_std": 1.4357442523760255e-05 }, { "epoch": 0.06068473146195035, "step": 187, "weight_max": 5.740389497077558e-06, "weight_mean": 7.187628057181428e-07, "weight_min": 6.053385570581327e-10, "weight_std": 2.029043798756902e-06 }, { "epoch": 0.06068473146195035, "step": 187, "weight_max": 2.5283219073912733e-08, "weight_mean": 4.550127830071915e-09, "weight_min": 0.0, "weight_std": 8.503780790647397e-09 }, { "epoch": 0.06100924874249554, "step": 188, "weight_max": 3.4268457966391e-05, "weight_mean": 1.2054766557412222e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.6721280189813115e-05 }, { "epoch": 0.06100924874249554, "step": 188, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.791398613539059e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6304269593092613e-05 }, { "epoch": 0.06100924874249554, "step": 188, "weight_max": 1.0469849257788155e-05, "weight_mean": 1.310304014623398e-06, "weight_min": 0.0, "weight_std": 3.7010158848715946e-06 }, { "epoch": 0.06100924874249554, "step": 188, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.01804378372617e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.685027928033378e-05 }, { "epoch": 0.06133376602304073, "step": 189, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.5039078789413907e-05, "weight_min": 0.0, "weight_std": 1.9539143977453932e-05 }, { "epoch": 0.06133376602304073, "step": 189, "weight_max": 3.662098970380612e-05, "weight_mean": 7.909798114269506e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.488420912210131e-05 }, { "epoch": 0.06133376602304073, "step": 189, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.84659982350422e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8257271221955307e-05 }, { "epoch": 0.06133376602304073, "step": 189, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.38376433623489e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7386177205480635e-05 }, { "epoch": 0.061658283303585915, "grad_norm": 0.3234117031097412, "learning_rate": 9.820408163265308e-06, "loss": 0.1531, "step": 190 }, { "epoch": 0.061658283303585915, "step": 190, "weight_max": 3.4413205618477605e-09, "weight_mean": 1.4182632845916032e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.245053171672339e-09 }, { "epoch": 0.061658283303585915, "step": 190, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.667801921546925e-06, "weight_min": 0.0, "weight_std": 1.6359801520593464e-05 }, { "epoch": 0.061658283303585915, "step": 190, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9313769371365197e-05, "weight_min": 0.0, "weight_std": 2.0724510250147432e-05 }, { "epoch": 0.061658283303585915, "step": 190, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2387312128557824e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8789840396493673e-05 }, { "epoch": 0.0619828005841311, "step": 191, "weight_max": 3.998322426923551e-05, "weight_mean": 9.613720976631157e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.781811261025723e-05 }, { "epoch": 0.0619828005841311, "step": 191, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076826255390188e-06, "weight_min": 0.0, "weight_std": 1.4357734471559525e-05 }, { "epoch": 0.0619828005841311, "step": 191, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4299257600214332e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.005003989324905e-05 }, { "epoch": 0.0619828005841311, "step": 191, "weight_max": 4.187939703115262e-05, "weight_mean": 9.269744623452425e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7353770090267062e-05 }, { "epoch": 0.0623073178646763, "step": 192, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.635021004010923e-06, "weight_min": 0.0, "weight_std": 1.6313582818838768e-05 }, { "epoch": 0.0623073178646763, "step": 192, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728940893983236e-06, "weight_min": 0.0, "weight_std": 1.3373446563491598e-05 }, { "epoch": 0.0623073178646763, "step": 192, "weight_max": 3.998322426923551e-05, "weight_mean": 9.174355000141077e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.707669798634015e-05 }, { "epoch": 0.0623073178646763, "step": 192, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235425305727404e-06, "weight_min": 0.0, "weight_std": 1.4806400940869935e-05 }, { "epoch": 0.06263183514522148, "step": 193, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.930842518049758e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.069626862066798e-05 }, { "epoch": 0.06263183514522148, "step": 193, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.167074505152414e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4576689864043146e-05 }, { "epoch": 0.06263183514522148, "step": 193, "weight_max": 3.572963396436535e-05, "weight_mean": 5.176882496016333e-06, "weight_min": 0.0, "weight_std": 1.2504191545303911e-05 }, { "epoch": 0.06263183514522148, "step": 193, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212553903926164e-06, "weight_min": 0.0, "weight_std": 1.1911706678802148e-05 }, { "epoch": 0.06295635242576668, "step": 194, "weight_max": 4.187939703115262e-05, "weight_mean": 1.793539195205085e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.008470801229123e-05 }, { "epoch": 0.06295635242576668, "step": 194, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3300978025654331e-05, "weight_min": 0.0, "weight_std": 1.8563900084700435e-05 }, { "epoch": 0.06295635242576668, "step": 194, "weight_max": 3.143761568935588e-05, "weight_mean": 3.9318106246355455e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1114024346170481e-05 }, { "epoch": 0.06295635242576668, "step": 194, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.1864611224154942e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7646849300945178e-05 }, { "epoch": 0.06328086970631186, "step": 195, "weight_max": 3.782653220696375e-05, "weight_mean": 8.755277121963445e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.563313344377093e-05 }, { "epoch": 0.06328086970631186, "step": 195, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.927097961626714e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3913373550167307e-05 }, { "epoch": 0.06328086970631186, "step": 195, "weight_max": 2.5141427613561973e-05, "weight_mean": 5.614606379822362e-06, "weight_min": 0.0, "weight_std": 1.0492441106180195e-05 }, { "epoch": 0.06328086970631186, "step": 195, "weight_max": 3.602553260861896e-05, "weight_mean": 1.2804664038412739e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.774616976035759e-05 }, { "epoch": 0.06360538698685705, "step": 196, "weight_max": 4.187939703115262e-05, "weight_mean": 9.018260243465193e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.654737934586592e-05 }, { "epoch": 0.06360538698685705, "step": 196, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0311461412347853e-05, "weight_min": 0.0, "weight_std": 2.173958819184918e-05 }, { "epoch": 0.06360538698685705, "step": 196, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6207000953727402e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.031568146776408e-05 }, { "epoch": 0.06360538698685705, "step": 196, "weight_max": 4.252122016623616e-05, "weight_mean": 1.55472425831249e-05, "weight_min": 0.0, "weight_std": 2.1462477889144793e-05 }, { "epoch": 0.06392990426740223, "step": 197, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8141392502002418e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9048015019507147e-05 }, { "epoch": 0.06392990426740223, "step": 197, "weight_max": 3.572963396436535e-05, "weight_mean": 7.766266207909212e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4593505511584226e-05 }, { "epoch": 0.06392990426740223, "step": 197, "weight_max": 2.464847420924343e-05, "weight_mean": 3.082014472965966e-06, "weight_min": 6.29309993005478e-10, "weight_std": 8.714165232959203e-06 }, { "epoch": 0.06392990426740223, "step": 197, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9209925236937124e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.3915834642830305e-05 }, { "epoch": 0.06425442154794743, "step": 198, "weight_max": 1.0211713430408054e-07, "weight_mean": 1.3504399376529364e-08, "weight_min": 5.679522963042416e-10, "weight_std": 3.580585072882059e-08 }, { "epoch": 0.06425442154794743, "step": 198, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0001769624068402e-05, "weight_min": 0.0, "weight_std": 1.854558831837494e-05 }, { "epoch": 0.06425442154794743, "step": 198, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2906075426144525e-05, "weight_min": 0.0, "weight_std": 1.8006088794209063e-05 }, { "epoch": 0.06425442154794743, "step": 198, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4004008335177787e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.936008447955828e-05 }, { "epoch": 0.06457893882849262, "step": 199, "weight_max": 4.252122016623616e-05, "weight_mean": 9.655974281486124e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.795411117200274e-05 }, { "epoch": 0.06457893882849262, "step": 199, "weight_max": 3.312147964606993e-05, "weight_mean": 1.2350361430435441e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.704473834251985e-05 }, { "epoch": 0.06457893882849262, "step": 199, "weight_max": 2.4894337911973707e-05, "weight_mean": 4.921007075608941e-06, "weight_min": 6.29309993005478e-10, "weight_std": 9.527030670142267e-06 }, { "epoch": 0.06457893882849262, "step": 199, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920890205539763e-06, "weight_min": 0.0, "weight_std": 1.3915874660597183e-05 }, { "epoch": 0.0649034561090378, "grad_norm": 2.555039405822754, "learning_rate": 9.800000000000001e-06, "loss": 0.2582, "step": 200 }, { "epoch": 0.0649034561090378, "eval_accuracy": 0.494, "eval_auc": 0.3980613208301996, "eval_loss": 1.8595376014709473, "eval_runtime": 155.2456, "eval_samples_per_second": 12.883, "eval_steps_per_second": 1.61, "step": 200 }, { "epoch": 0.0649034561090378, "step": 200, "weight_max": 3.813097282545641e-05, "weight_mean": 4.7668754632468335e-06, "weight_min": 0.0, "weight_std": 1.3481130736181512e-05 }, { "epoch": 0.0649034561090378, "step": 200, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0429137293831445e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1854073565918952e-05 }, { "epoch": 0.0649034561090378, "step": 200, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4380148058990017e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0159130144747905e-05 }, { "epoch": 0.0649034561090378, "step": 200, "weight_max": 4.187939703115262e-05, "weight_mean": 1.004087425826583e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.861146120063495e-05 }, { "epoch": 0.065227973389583, "step": 201, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4186700354912318e-05, "weight_min": 0.0, "weight_std": 1.9901579435099848e-05 }, { "epoch": 0.065227973389583, "step": 201, "weight_max": 4.3167921830900013e-05, "weight_mean": 7.653025022591464e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.567840990901459e-05 }, { "epoch": 0.065227973389583, "step": 201, "weight_max": 4.252122016623616e-05, "weight_mean": 9.781989319890272e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8201893908553757e-05 }, { "epoch": 0.065227973389583, "step": 201, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.899628947896417e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3675484296982177e-05 }, { "epoch": 0.06555249067012818, "step": 202, "weight_max": 2.639517333591357e-05, "weight_mean": 4.344095941632986e-06, "weight_min": 5.864964625068581e-10, "weight_std": 9.376909474667627e-06 }, { "epoch": 0.06555249067012818, "step": 202, "weight_max": 2.6141986381844617e-05, "weight_mean": 3.28093869939039e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.237324775313027e-06 }, { "epoch": 0.06555249067012818, "step": 202, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930867933377158e-06, "weight_min": 0.0, "weight_std": 1.1114404514955822e-05 }, { "epoch": 0.06555249067012818, "step": 202, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767181053466629e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3481007044902071e-05 }, { "epoch": 0.06587700795067337, "step": 203, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.309627785114571e-06, "weight_min": 0.0, "weight_std": 1.7255693819606677e-05 }, { "epoch": 0.06587700795067337, "step": 203, "weight_max": 4.252122016623616e-05, "weight_mean": 9.968551239580847e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.8510632799007e-05 }, { "epoch": 0.06587700795067337, "step": 203, "weight_max": 3.813097282545641e-05, "weight_mean": 9.2726559159928e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7171209037769586e-05 }, { "epoch": 0.06587700795067337, "step": 203, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.429970431374386e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2527908438642044e-05 }, { "epoch": 0.06620152523121857, "step": 204, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2632484615314752e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9379484001547098e-05 }, { "epoch": 0.06620152523121857, "step": 204, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.631716809759382e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.4639478649769444e-05 }, { "epoch": 0.06620152523121857, "step": 204, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470445886312518e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.938602508744225e-05 }, { "epoch": 0.06620152523121857, "step": 204, "weight_max": 7.466283591384126e-08, "weight_mean": 1.0113139659040371e-08, "weight_min": 5.771870759119224e-10, "weight_std": 2.6083943893695505e-08 }, { "epoch": 0.06652604251176375, "step": 205, "weight_max": 3.662098970380612e-05, "weight_mean": 9.007753760670312e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6680256521794945e-05 }, { "epoch": 0.06652604251176375, "step": 205, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0552264939178713e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9543385860743e-05 }, { "epoch": 0.06652604251176375, "step": 205, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1586114598903805e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.8689290300244465e-05 }, { "epoch": 0.06652604251176375, "step": 205, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397238055593334e-06, "weight_min": 0.0, "weight_std": 1.52616612467682e-05 }, { "epoch": 0.06685055979230894, "step": 206, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4120139894657768e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.9677494492498226e-05 }, { "epoch": 0.06685055979230894, "step": 206, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3328257409739308e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8564413039712235e-05 }, { "epoch": 0.06685055979230894, "step": 206, "weight_max": 3.662098970380612e-05, "weight_mean": 8.404924301430583e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5643616279703565e-05 }, { "epoch": 0.06685055979230894, "step": 206, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4301086745690554e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2527852959465235e-05 }, { "epoch": 0.06717507707285413, "step": 207, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.129753379966132e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6957261323113926e-05 }, { "epoch": 0.06717507707285413, "step": 207, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471178029547445e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9390092347748578e-05 }, { "epoch": 0.06717507707285413, "step": 207, "weight_max": 4.252122016623616e-05, "weight_mean": 1.012029497360345e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8769525922834873e-05 }, { "epoch": 0.06717507707285413, "step": 207, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5675483155064285e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.079316800518427e-05 }, { "epoch": 0.06749959435339932, "step": 208, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.211795259034261e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7170274077216163e-05 }, { "epoch": 0.06749959435339932, "step": 208, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3476579372072592e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9843475456582382e-05 }, { "epoch": 0.06749959435339932, "step": 208, "weight_max": 2.6141986381844617e-05, "weight_mean": 3.2688083138054935e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.242159649147652e-06 }, { "epoch": 0.06749959435339932, "step": 208, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156190127308946e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4581054529116955e-05 }, { "epoch": 0.06782411163394451, "step": 209, "weight_max": 4.252122016623616e-05, "weight_mean": 9.385556950292084e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7579179257154465e-05 }, { "epoch": 0.06782411163394451, "step": 209, "weight_max": 3.752332122530788e-05, "weight_mean": 1.6491532733198255e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7788219338399358e-05 }, { "epoch": 0.06782411163394451, "step": 209, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3970790860184934e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9303235603729263e-05 }, { "epoch": 0.06782411163394451, "step": 209, "weight_max": 2.5561155414521863e-09, "weight_mean": 9.277412171826427e-10, "weight_min": 5.679522963042416e-10, "weight_std": 6.850717770845449e-10 }, { "epoch": 0.0681486289144897, "grad_norm": 4.62890100479126, "learning_rate": 9.779591836734694e-06, "loss": 0.3963, "step": 210 }, { "epoch": 0.0681486289144897, "step": 210, "weight_max": 3.340639523230493e-05, "weight_mean": 1.145470014307648e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.5987785445759073e-05 }, { "epoch": 0.0681486289144897, "step": 210, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959506895829691e-06, "weight_min": 0.0, "weight_std": 1.4025749806023668e-05 }, { "epoch": 0.0681486289144897, "step": 210, "weight_max": 3.143761568935588e-05, "weight_mean": 8.863594302965794e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.3632597074320074e-05 }, { "epoch": 0.0681486289144897, "step": 210, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0363335352158174e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9369013898540288e-05 }, { "epoch": 0.06847314619503489, "step": 211, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691120011557359e-06, "weight_min": 0.0, "weight_std": 1.326621259067906e-05 }, { "epoch": 0.06847314619503489, "step": 211, "weight_max": 4.252122016623616e-05, "weight_mean": 9.24904270505067e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7299091268796474e-05 }, { "epoch": 0.06847314619503489, "step": 211, "weight_max": 7.540548629947352e-09, "weight_mean": 1.7216981174072998e-09, "weight_min": 0.0, "weight_std": 2.4851196656072716e-09 }, { "epoch": 0.06847314619503489, "step": 211, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5787853044457734e-05, "weight_min": 0.0, "weight_std": 2.179999319196213e-05 }, { "epoch": 0.06879766347558007, "step": 212, "weight_max": 3.662098970380612e-05, "weight_mean": 8.647707545605954e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.604761928319931e-05 }, { "epoch": 0.06879766347558007, "step": 212, "weight_max": 3.340639523230493e-05, "weight_mean": 6.511971150757745e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.2682044143730309e-05 }, { "epoch": 0.06879766347558007, "step": 212, "weight_max": 3.255531191825867e-05, "weight_mean": 4.094264113518875e-06, "weight_min": 0.0, "weight_std": 1.1500102118588984e-05 }, { "epoch": 0.06879766347558007, "step": 212, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8629685655469075e-05, "weight_min": 0.0, "weight_std": 1.9996707123937085e-05 }, { "epoch": 0.06912218075612526, "step": 213, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2949956726515666e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8154412828153e-05 }, { "epoch": 0.06912218075612526, "step": 213, "weight_max": 3.967146767536178e-05, "weight_mean": 5.333165518095484e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3914130249759182e-05 }, { "epoch": 0.06912218075612526, "step": 213, "weight_max": 3.602553260861896e-05, "weight_mean": 1.1308724424452521e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6022491763578728e-05 }, { "epoch": 0.06912218075612526, "step": 213, "weight_max": 3.782653220696375e-05, "weight_mean": 4.736500613944372e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3370404303714167e-05 }, { "epoch": 0.06944669803667046, "step": 214, "weight_max": 3.782653220696375e-05, "weight_mean": 9.381965355714783e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.737034835969098e-05 }, { "epoch": 0.06944669803667046, "step": 214, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2772667105309665e-05, "weight_min": 0.0, "weight_std": 1.7885613488033414e-05 }, { "epoch": 0.06944669803667046, "step": 214, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.648682155064307e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8481990991858765e-05 }, { "epoch": 0.06944669803667046, "step": 214, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0551161722105462e-05, "weight_min": 0.0, "weight_std": 1.9535031242412515e-05 }, { "epoch": 0.06977121531721564, "step": 215, "weight_max": 1.0385104998178463e-09, "weight_mean": 7.135570467831087e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.5746902393143358e-10 }, { "epoch": 0.06977121531721564, "step": 215, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4665232811239548e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.032739394053351e-05 }, { "epoch": 0.06977121531721564, "step": 215, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355661288485862e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9195897039026022e-05 }, { "epoch": 0.06977121531721564, "step": 215, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.098962446936639e-06, "weight_min": 0.0, "weight_std": 1.4348916010931134e-05 }, { "epoch": 0.07009573259776083, "step": 216, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960513125406578e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.845697988756001e-05 }, { "epoch": 0.07009573259776083, "step": 216, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2146879928186536e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.939214052981697e-05 }, { "epoch": 0.07009573259776083, "step": 216, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235631306277355e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806317267357372e-05 }, { "epoch": 0.07009573259776083, "step": 216, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077705282019451e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357378859131131e-05 }, { "epoch": 0.07042024987830602, "step": 217, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4357107829710003e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.987183713936247e-05 }, { "epoch": 0.07042024987830602, "step": 217, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.2265425539226271e-05, "weight_min": 0.0, "weight_std": 1.6697713363100775e-05 }, { "epoch": 0.07042024987830602, "step": 217, "weight_max": 4.252122016623616e-05, "weight_mean": 1.578827141202055e-05, "weight_min": 0.0, "weight_std": 1.852890636655502e-05 }, { "epoch": 0.07042024987830602, "step": 217, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.58508189796703e-06, "weight_min": 0.0, "weight_std": 1.7814867533161305e-05 }, { "epoch": 0.07074476715885121, "step": 218, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8659357010619715e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.015413156186696e-05 }, { "epoch": 0.07074476715885121, "step": 218, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070159775437787e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.150973912444897e-05 }, { "epoch": 0.07074476715885121, "step": 218, "weight_max": 1.987018549698405e-09, "weight_mean": 9.705338754883996e-10, "weight_min": 6.196655966128617e-10, "weight_std": 4.539765785516181e-10 }, { "epoch": 0.07074476715885121, "step": 218, "weight_max": 4.187939703115262e-05, "weight_mean": 9.813309588935226e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8223110600956716e-05 }, { "epoch": 0.0710692844393964, "step": 219, "weight_max": 3.572963396436535e-05, "weight_mean": 8.293991413665935e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.541515302960761e-05 }, { "epoch": 0.0710692844393964, "step": 219, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5114427696971688e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0871801098110154e-05 }, { "epoch": 0.0710692844393964, "step": 219, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.111414328799583e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.701544169918634e-05 }, { "epoch": 0.0710692844393964, "step": 219, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4499115422950126e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0046800273121335e-05 }, { "epoch": 0.07139380171994159, "grad_norm": 2.05782413482666, "learning_rate": 9.75918367346939e-06, "loss": 0.2362, "step": 220 }, { "epoch": 0.07139380171994159, "step": 220, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9396216885070316e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.083506842609495e-05 }, { "epoch": 0.07139380171994159, "step": 220, "weight_max": 4.381950566312298e-05, "weight_mean": 9.653937013354152e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8089254808728583e-05 }, { "epoch": 0.07139380171994159, "step": 220, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.046385452558752e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7755577573552728e-05 }, { "epoch": 0.07139380171994159, "step": 220, "weight_max": 4.187939703115262e-05, "weight_mean": 8.766869541432243e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6623376723146066e-05 }, { "epoch": 0.07171831900048678, "step": 221, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155910457688151e-06, "weight_min": 0.0, "weight_std": 1.4581168215954676e-05 }, { "epoch": 0.07171831900048678, "step": 221, "weight_max": 2.846461757144425e-05, "weight_mean": 6.458844836743083e-06, "weight_min": 0.0, "weight_std": 1.2039750799885951e-05 }, { "epoch": 0.07171831900048678, "step": 221, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.3322196486551547e-06, "weight_min": 5.958802340444436e-10, "weight_std": 9.42163842410082e-06 }, { "epoch": 0.07171831900048678, "step": 221, "weight_max": 3.8743506593164057e-05, "weight_mean": 7.858150638639927e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4762095815967768e-05 }, { "epoch": 0.07204283628103196, "step": 222, "weight_max": 1.0224462165808745e-08, "weight_mean": 2.0275940926239855e-09, "weight_min": 5.864964625068581e-10, "weight_std": 3.3334202065304908e-09 }, { "epoch": 0.07204283628103196, "step": 222, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7733676941134036e-05, "weight_min": 0.0, "weight_std": 1.9182929463568144e-05 }, { "epoch": 0.07204283628103196, "step": 222, "weight_max": 3.312147964606993e-05, "weight_mean": 5.449493073683698e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1766019269998651e-05 }, { "epoch": 0.07204283628103196, "step": 222, "weight_max": 7.937498991417158e-10, "weight_mean": 6.985131917325305e-10, "weight_min": 5.864964625068581e-10, "weight_std": 8.204587009785769e-11 }, { "epoch": 0.07236735356157716, "step": 223, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5352710761362687e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.119784039678052e-05 }, { "epoch": 0.07236735356157716, "step": 223, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4957538951421157e-05, "weight_min": 0.0, "weight_std": 2.0653611500165425e-05 }, { "epoch": 0.07236735356157716, "step": 223, "weight_max": 3.255531191825867e-05, "weight_mean": 8.10418714536354e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5004952729213983e-05 }, { "epoch": 0.07236735356157716, "step": 223, "weight_max": 3.998322426923551e-05, "weight_mean": 8.393879397772253e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.591395994182676e-05 }, { "epoch": 0.07269187084212235, "step": 224, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5468240235350095e-05, "weight_min": 0.0, "weight_std": 2.1355268472689204e-05 }, { "epoch": 0.07269187084212235, "step": 224, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.57335487328237e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.773421354300808e-05 }, { "epoch": 0.07269187084212235, "step": 224, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5417605178954545e-06, "weight_min": 0.0, "weight_std": 1.2841418538300786e-05 }, { "epoch": 0.07269187084212235, "step": 224, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.3188142222352326e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9504868760122918e-05 }, { "epoch": 0.07301638812266753, "step": 225, "weight_max": 3.998322426923551e-05, "weight_mean": 9.72251746134134e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6738707927288488e-05 }, { "epoch": 0.07301638812266753, "step": 225, "weight_max": 4.252122016623616e-05, "weight_mean": 1.625605727895163e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.143797610187903e-05 }, { "epoch": 0.07301638812266753, "step": 225, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550751539994963e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9535284081939608e-05 }, { "epoch": 0.07301638812266753, "step": 225, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0274783310014755e-05, "weight_min": 0.0, "weight_std": 2.1710053260903805e-05 }, { "epoch": 0.07334090540321273, "step": 226, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3651148947246838e-05, "weight_min": 0.0, "weight_std": 1.89017373486422e-05 }, { "epoch": 0.07334090540321273, "step": 226, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8824903236236423e-05, "weight_min": 0.0, "weight_std": 2.0380140995257534e-05 }, { "epoch": 0.07334090540321273, "step": 226, "weight_max": 3.255531191825867e-05, "weight_mean": 7.369365903286962e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3743014278588817e-05 }, { "epoch": 0.07334090540321273, "step": 226, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691067260864656e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3266234418551903e-05 }, { "epoch": 0.0736654226837579, "step": 227, "weight_max": 3.116124207735993e-05, "weight_mean": 3.8957350625423715e-06, "weight_min": 0.0, "weight_std": 1.1016929420293309e-05 }, { "epoch": 0.0736654226837579, "step": 227, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3919072191347368e-05, "weight_min": 7.721521200210191e-10, "weight_std": 2.0013445464428514e-05 }, { "epoch": 0.0736654226837579, "step": 227, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7937286379019497e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0726089385570958e-05 }, { "epoch": 0.0736654226837579, "step": 227, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3745789146923926e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9185039491276257e-05 }, { "epoch": 0.0739899399643031, "step": 228, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5194917978078593e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0992025383748114e-05 }, { "epoch": 0.0739899399643031, "step": 228, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5082765457918867e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0857818526565097e-05 }, { "epoch": 0.0739899399643031, "step": 228, "weight_max": 4.187939703115262e-05, "weight_mean": 9.777348168427125e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8159793398808688e-05 }, { "epoch": 0.0739899399643031, "step": 228, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5283696484402753e-05, "weight_min": 0.0, "weight_std": 2.115522147505544e-05 }, { "epoch": 0.0743144572448483, "step": 229, "weight_max": 3.967146767536178e-05, "weight_mean": 9.658673661760986e-06, "weight_min": 0.0, "weight_std": 1.7870608644443564e-05 }, { "epoch": 0.0743144572448483, "step": 229, "weight_max": 3.28377827827353e-05, "weight_mean": 9.777323612070177e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.444290410290705e-05 }, { "epoch": 0.0743144572448483, "step": 229, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.5477264545043e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5043735402286984e-05 }, { "epoch": 0.0743144572448483, "step": 229, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077213245385792e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357576219481416e-05 }, { "epoch": 0.07463897452539348, "grad_norm": 0.8887665271759033, "learning_rate": 9.738775510204083e-06, "loss": 0.264, "step": 230 }, { "epoch": 0.07463897452539348, "step": 230, "weight_max": 3.76558251247161e-09, "weight_mean": 9.810394718812177e-10, "weight_min": 0.0, "weight_std": 1.1602783178688014e-09 }, { "epoch": 0.07463897452539348, "step": 230, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8036680557997897e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9841259927488863e-05 }, { "epoch": 0.07463897452539348, "step": 230, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8616323359310627e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0358858819236048e-05 }, { "epoch": 0.07463897452539348, "step": 230, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3956410839455202e-05, "weight_min": 0.0, "weight_std": 1.941225491464138e-05 }, { "epoch": 0.07496349180593867, "step": 231, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0194407877861522e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8884240489569493e-05 }, { "epoch": 0.07496349180593867, "step": 231, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8182316125603393e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0192695956211537e-05 }, { "epoch": 0.07496349180593867, "step": 231, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.9772925952565856e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8101551177096553e-05 }, { "epoch": 0.07496349180593867, "step": 231, "weight_max": 4.698116811141517e-07, "weight_mean": 6.113569384069706e-08, "weight_min": 7.937498991417158e-10, "weight_std": 1.6515095069280505e-07 }, { "epoch": 0.07528800908648385, "step": 232, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8567869119578972e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9983173842774704e-05 }, { "epoch": 0.07528800908648385, "step": 232, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4002907846588641e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9355131371412426e-05 }, { "epoch": 0.07528800908648385, "step": 232, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541793259704718e-06, "weight_min": 0.0, "weight_std": 1.284140489588026e-05 }, { "epoch": 0.07528800908648385, "step": 232, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0197258234256878e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8902796000475064e-05 }, { "epoch": 0.07561252636702905, "step": 233, "weight_max": 4.252122016623616e-05, "weight_mean": 1.489080568717327e-05, "weight_min": 0.0, "weight_std": 2.0563649741234258e-05 }, { "epoch": 0.07561252636702905, "step": 233, "weight_max": 6.0400980146368966e-06, "weight_mean": 7.559173695881327e-07, "weight_min": 6.100955851628953e-10, "weight_std": 2.13513158087153e-06 }, { "epoch": 0.07561252636702905, "step": 233, "weight_max": 4.187939703115262e-05, "weight_mean": 1.421182241756469e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9725093807210214e-05 }, { "epoch": 0.07561252636702905, "step": 233, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.134132980951108e-05, "weight_min": 0.0, "weight_std": 1.7457919966545887e-05 }, { "epoch": 0.07593704364757424, "step": 234, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4104132787906565e-05, "weight_min": 0.0, "weight_std": 1.9756222172873095e-05 }, { "epoch": 0.07593704364757424, "step": 234, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.0761125849967357e-05, "weight_min": 0.0, "weight_std": 1.6684180081938393e-05 }, { "epoch": 0.07593704364757424, "step": 234, "weight_max": 3.4558255720185116e-05, "weight_mean": 5.3172625484876335e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2111151590943336e-05 }, { "epoch": 0.07593704364757424, "step": 234, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4157754776533693e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.9583538232836872e-05 }, { "epoch": 0.07626156092811942, "step": 235, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2422760846675374e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8962758986162953e-05 }, { "epoch": 0.07626156092811942, "step": 235, "weight_max": 3.813097282545641e-05, "weight_mean": 4.766933670907747e-06, "weight_min": 0.0, "weight_std": 1.3481107998813968e-05 }, { "epoch": 0.07626156092811942, "step": 235, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4285372344602365e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9822247850243002e-05 }, { "epoch": 0.07626156092811942, "step": 235, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9927354514948092e-05, "weight_min": 0.0, "weight_std": 2.1360485334298573e-05 }, { "epoch": 0.07658607820866462, "step": 236, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0818180170608684e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4355722669279203e-05 }, { "epoch": 0.07658607820866462, "step": 236, "weight_max": 3.722132169059478e-05, "weight_mean": 1.9403765691095032e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.705459908407647e-05 }, { "epoch": 0.07658607820866462, "step": 236, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315926046023378e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5033208910608664e-05 }, { "epoch": 0.07658607820866462, "step": 236, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235568096395582e-06, "weight_min": 0.0, "weight_std": 1.4806342733209021e-05 }, { "epoch": 0.0769105954892098, "step": 237, "weight_max": 3.998322426923551e-05, "weight_mean": 8.893743142834865e-06, "weight_min": 0.0, "weight_std": 1.6634267012705095e-05 }, { "epoch": 0.0769105954892098, "step": 237, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960542229237035e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8456961697665974e-05 }, { "epoch": 0.0769105954892098, "step": 237, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5429846826009452e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1307874703779817e-05 }, { "epoch": 0.0769105954892098, "step": 237, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0772637223417405e-06, "weight_min": 0.0, "weight_std": 1.4357557120092679e-05 }, { "epoch": 0.07723511276975499, "step": 238, "weight_max": 4.252122016623616e-05, "weight_mean": 1.54690642375499e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1358562662499025e-05 }, { "epoch": 0.07723511276975499, "step": 238, "weight_max": 4.252122016623616e-05, "weight_mean": 2.390880763414316e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.039396895270329e-05 }, { "epoch": 0.07723511276975499, "step": 238, "weight_max": 3.843663216684945e-05, "weight_mean": 1.367352160741575e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8934217223431915e-05 }, { "epoch": 0.07723511276975499, "step": 238, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236194283497753e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806089893681929e-05 }, { "epoch": 0.07755963005030019, "step": 239, "weight_max": 2.2256894226302393e-05, "weight_mean": 2.7830174076370895e-06, "weight_min": 6.100955851628953e-10, "weight_std": 7.868634384067263e-06 }, { "epoch": 0.07755963005030019, "step": 239, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5117488146643154e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0887648133793846e-05 }, { "epoch": 0.07755963005030019, "step": 239, "weight_max": 3.36925259034615e-05, "weight_mean": 7.904351150500588e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4674058547825553e-05 }, { "epoch": 0.07755963005030019, "step": 239, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.766554305504542e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9282702851342037e-05 }, { "epoch": 0.07788414733084537, "grad_norm": 4.052671432495117, "learning_rate": 9.718367346938776e-06, "loss": 0.2398, "step": 240 }, { "epoch": 0.07788414733084537, "step": 240, "weight_max": 3.967146767536178e-05, "weight_mean": 9.17102261155378e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7055164789780974e-05 }, { "epoch": 0.07788414733084537, "step": 240, "weight_max": 3.843663216684945e-05, "weight_mean": 9.08864149096189e-06, "weight_min": 0.0, "weight_std": 1.6864480130607262e-05 }, { "epoch": 0.07788414733084537, "step": 240, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5707628335803747e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1688447304768488e-05 }, { "epoch": 0.07788414733084537, "step": 240, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077014520793455e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357658074004576e-05 }, { "epoch": 0.07820866461139056, "step": 241, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4238927178666927e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9877328668371774e-05 }, { "epoch": 0.07820866461139056, "step": 241, "weight_max": 3.061215829802677e-05, "weight_mean": 5.249222340353299e-06, "weight_min": 0.0, "weight_std": 1.0890569683397189e-05 }, { "epoch": 0.07820866461139056, "step": 241, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4507278137898538e-05, "weight_min": 0.0, "weight_std": 2.009264972002711e-05 }, { "epoch": 0.07820866461139056, "step": 241, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9770697690546513e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1185032892390154e-05 }, { "epoch": 0.07853318189193574, "step": 242, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2664102289127186e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9238861568737775e-05 }, { "epoch": 0.07853318189193574, "step": 242, "weight_max": 2.639517333591357e-05, "weight_mean": 3.332365395181114e-06, "weight_min": 5.958802340444436e-10, "weight_std": 9.319222044723574e-06 }, { "epoch": 0.07853318189193574, "step": 242, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233336070086807e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8953976905322634e-05 }, { "epoch": 0.07853318189193574, "step": 242, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313954589946661e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9107557818642817e-05 }, { "epoch": 0.07885769917248094, "step": 243, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.967930287122726e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5578661987092346e-05 }, { "epoch": 0.07885769917248094, "step": 243, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1809763236669824e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.7468106307205744e-05 }, { "epoch": 0.07885769917248094, "step": 243, "weight_max": 4.187939703115262e-05, "weight_mean": 1.689003693172708e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8954309780383483e-05 }, { "epoch": 0.07885769917248094, "step": 243, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.4952420315239578e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7225258488906547e-05 }, { "epoch": 0.07918221645302613, "step": 244, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9093844457529485e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9447026716079563e-05 }, { "epoch": 0.07918221645302613, "step": 244, "weight_max": 4.187939703115262e-05, "weight_mean": 9.59195131144952e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7857813872979023e-05 }, { "epoch": 0.07918221645302613, "step": 244, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4877831745252479e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.053410207736306e-05 }, { "epoch": 0.07918221645302613, "step": 244, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0925082177855074e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.4351458958117291e-05 }, { "epoch": 0.07950673373357131, "step": 245, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0233280590909999e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.8944978364743292e-05 }, { "epoch": 0.07950673373357131, "step": 245, "weight_max": 4.187939703115262e-05, "weight_mean": 1.7895119526656345e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9407179934205487e-05 }, { "epoch": 0.07950673373357131, "step": 245, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5468793208128773e-05, "weight_min": 0.0, "weight_std": 2.135878639819566e-05 }, { "epoch": 0.07950673373357131, "step": 245, "weight_max": 3.967146767536178e-05, "weight_mean": 9.784709618543275e-06, "weight_min": 0.0, "weight_std": 1.5770221580169164e-05 }, { "epoch": 0.0798312510141165, "step": 246, "weight_max": 3.624825550829769e-09, "weight_mean": 1.0268599304197323e-09, "weight_min": 0.0, "weight_std": 1.2543831529043814e-09 }, { "epoch": 0.0798312510141165, "step": 246, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578326297632884e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2947190953127574e-05 }, { "epoch": 0.0798312510141165, "step": 246, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.433763024280779e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2526378668553662e-05 }, { "epoch": 0.0798312510141165, "step": 246, "weight_max": 4.252122016623616e-05, "weight_mean": 2.2618574803345837e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.9018276361748576e-05 }, { "epoch": 0.08015576829466169, "step": 247, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5787934898980893e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.17999258893542e-05 }, { "epoch": 0.08015576829466169, "step": 247, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965024461649591e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1212766366952565e-05 }, { "epoch": 0.08015576829466169, "step": 247, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3473625585902482e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8763848856906407e-05 }, { "epoch": 0.08015576829466169, "step": 247, "weight_max": 4.252122016623616e-05, "weight_mean": 3.0935047107050195e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.910820537887048e-05 }, { "epoch": 0.08048028557520688, "step": 248, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9554297978174873e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1017942344769835e-05 }, { "epoch": 0.08048028557520688, "step": 248, "weight_max": 3.967146767536178e-05, "weight_mean": 9.803165994526353e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.8150436517316848e-05 }, { "epoch": 0.08048028557520688, "step": 248, "weight_max": 4.252122016623616e-05, "weight_mean": 1.276525199500611e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8048523998004384e-05 }, { "epoch": 0.08048028557520688, "step": 248, "weight_max": 3.006794759130571e-05, "weight_mean": 7.483852641598787e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3855988072464243e-05 }, { "epoch": 0.08080480285575208, "step": 249, "weight_max": 2.716206836339552e-05, "weight_mean": 3.3959688607865246e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.60295437835157e-06 }, { "epoch": 0.08080480285575208, "step": 249, "weight_max": 3.006794759130571e-05, "weight_mean": 3.759126684599323e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0630369615682866e-05 }, { "epoch": 0.08080480285575208, "step": 249, "weight_max": 3.998322426923551e-05, "weight_mean": 9.920333468471654e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8363927665632218e-05 }, { "epoch": 0.08080480285575208, "step": 249, "weight_max": 5.548205450622845e-09, "weight_mean": 2.0909383113831836e-09, "weight_min": 6.488223291967188e-10, "weight_std": 1.8249356470434464e-09 }, { "epoch": 0.08112932013629726, "grad_norm": 3.7305004596710205, "learning_rate": 9.697959183673469e-06, "loss": 0.1856, "step": 250 }, { "epoch": 0.08112932013629726, "step": 250, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3357919417321682e-05, "weight_min": 9.413956281179026e-10, "weight_std": 1.9593071556300856e-05 }, { "epoch": 0.08112932013629726, "step": 250, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.5800726032466628e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7938496966962703e-05 }, { "epoch": 0.08112932013629726, "step": 250, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0470873348822352e-05, "weight_min": 0.0, "weight_std": 1.9390281522646546e-05 }, { "epoch": 0.08112932013629726, "step": 250, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7933689327473985e-06, "weight_min": 0.0, "weight_std": 1.072623399522854e-05 }, { "epoch": 0.08145383741684245, "step": 251, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8789769455906935e-05, "weight_min": 0.0, "weight_std": 2.0342411517049186e-05 }, { "epoch": 0.08145383741684245, "step": 251, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.533396223559976e-06, "weight_min": 0.0, "weight_std": 1.7474314518040046e-05 }, { "epoch": 0.08145383741684245, "step": 251, "weight_max": 3.602553260861896e-05, "weight_mean": 8.538297151972074e-06, "weight_min": 0.0, "weight_std": 1.5839523257454857e-05 }, { "epoch": 0.08145383741684245, "step": 251, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1886881111422554e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7361675418214872e-05 }, { "epoch": 0.08177835469738763, "step": 252, "weight_max": 3.752332122530788e-05, "weight_mean": 1.3201408364693634e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8314442058908753e-05 }, { "epoch": 0.08177835469738763, "step": 252, "weight_max": 3.752332122530788e-05, "weight_mean": 9.157789463642985e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.696081926638726e-05 }, { "epoch": 0.08177835469738763, "step": 252, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0037691936304327e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8593109416542575e-05 }, { "epoch": 0.08177835469738763, "step": 252, "weight_max": 3.998322426923551e-05, "weight_mean": 8.962821993918624e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6741118088248186e-05 }, { "epoch": 0.08210287197793283, "step": 253, "weight_max": 2.5561155414521863e-09, "weight_mean": 9.344438556269097e-10, "weight_min": 5.679522963042416e-10, "weight_std": 6.635450522374242e-10 }, { "epoch": 0.08210287197793283, "step": 253, "weight_max": 3.843663216684945e-05, "weight_mean": 9.609537301003002e-06, "weight_min": 0.0, "weight_std": 1.779246849764604e-05 }, { "epoch": 0.08210287197793283, "step": 253, "weight_max": 9.160967096022432e-09, "weight_mean": 1.8673331769747392e-09, "weight_min": 6.196655966128617e-10, "weight_std": 2.955842681728882e-09 }, { "epoch": 0.08210287197793283, "step": 253, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274989108438604e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.746176349115558e-05 }, { "epoch": 0.08242738925847802, "step": 254, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504026037466247e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2736612006847281e-05 }, { "epoch": 0.08242738925847802, "step": 254, "weight_max": 6.177455702527368e-08, "weight_mean": 1.4695102024120388e-08, "weight_min": 0.0, "weight_std": 2.6633511396312315e-08 }, { "epoch": 0.08242738925847802, "step": 254, "weight_max": 4.252122016623616e-05, "weight_mean": 9.747303920448758e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8140766769647598e-05 }, { "epoch": 0.08242738925847802, "step": 254, "weight_max": 3.4268457966391e-05, "weight_mean": 7.583703336422332e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4197242308000568e-05 }, { "epoch": 0.0827519065390232, "step": 255, "weight_max": 3.998322426923551e-05, "weight_mean": 7.868578904890455e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.526234882476274e-05 }, { "epoch": 0.0827519065390232, "step": 255, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.339882252679672e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2210179193061776e-05 }, { "epoch": 0.0827519065390232, "step": 255, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236103334027575e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806127182964701e-05 }, { "epoch": 0.0827519065390232, "step": 255, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0283658411935903e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9058068573940545e-05 }, { "epoch": 0.0830764238195684, "step": 256, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6142827007570304e-05, "weight_min": 0.0, "weight_std": 1.806292129913345e-05 }, { "epoch": 0.0830764238195684, "step": 256, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8322400137549266e-05, "weight_min": 0.0, "weight_std": 1.9937082470278256e-05 }, { "epoch": 0.0830764238195684, "step": 256, "weight_max": 4.252122016623616e-05, "weight_mean": 8.64715639181668e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.656186577747576e-05 }, { "epoch": 0.0830764238195684, "step": 256, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5171859558904544e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.102789221680723e-05 }, { "epoch": 0.08340094110011358, "step": 257, "weight_max": 4.187939703115262e-05, "weight_mean": 1.480809623899404e-05, "weight_min": 0.0, "weight_std": 2.0474219127208926e-05 }, { "epoch": 0.08340094110011358, "step": 257, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5042771337903105e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.886092832137365e-05 }, { "epoch": 0.08340094110011358, "step": 257, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8833507056115195e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.027181653829757e-05 }, { "epoch": 0.08340094110011358, "step": 257, "weight_max": 3.632264997577295e-05, "weight_mean": 6.014455721015111e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2922263522341382e-05 }, { "epoch": 0.08372545838065877, "step": 258, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.456123573007062e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0387111362651922e-05 }, { "epoch": 0.08372545838065877, "step": 258, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.417315797880292e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9666678781504743e-05 }, { "epoch": 0.08372545838065877, "step": 258, "weight_max": 4.252122016623616e-05, "weight_mean": 1.697277548373677e-05, "weight_min": 0.0, "weight_std": 1.8550583263277076e-05 }, { "epoch": 0.08372545838065877, "step": 258, "weight_max": 3.602553260861896e-05, "weight_mean": 1.184007396659581e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.653852996241767e-05 }, { "epoch": 0.08404997566120397, "step": 259, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.357470970717259e-06, "weight_min": 0.0, "weight_std": 1.2320546375121921e-05 }, { "epoch": 0.08404997566120397, "step": 259, "weight_max": 3.22740561387036e-05, "weight_mean": 4.03509329771623e-06, "weight_min": 0.0, "weight_std": 1.1410264050937258e-05 }, { "epoch": 0.08404997566120397, "step": 259, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7138792827608995e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8488512068870477e-05 }, { "epoch": 0.08404997566120397, "step": 259, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470258530403953e-05, "weight_min": 0.0, "weight_std": 1.9386141502764076e-05 }, { "epoch": 0.08437449294174915, "grad_norm": 2.034040689468384, "learning_rate": 9.677551020408164e-06, "loss": 0.2782, "step": 260 }, { "epoch": 0.08437449294174915, "step": 260, "weight_max": 1.5955802235012584e-09, "weight_mean": 9.316662996639025e-10, "weight_min": 5.958802340444436e-10, "weight_std": 4.5854839370029765e-10 }, { "epoch": 0.08437449294174915, "step": 260, "weight_max": 4.187939703115262e-05, "weight_mean": 1.511485424998682e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0873427274636924e-05 }, { "epoch": 0.08437449294174915, "step": 260, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8822200217982754e-05, "weight_min": 0.0, "weight_std": 2.036848854913842e-05 }, { "epoch": 0.08437449294174915, "step": 260, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3355251212487929e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8795000869431533e-05 }, { "epoch": 0.08469901022229434, "step": 261, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2110971510992385e-05, "weight_min": 0.0, "weight_std": 1.733465978759341e-05 }, { "epoch": 0.08469901022229434, "step": 261, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396535016188864e-06, "weight_min": 0.0, "weight_std": 1.5261945009115152e-05 }, { "epoch": 0.08469901022229434, "step": 261, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.773530680628028e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6360161680495366e-05 }, { "epoch": 0.08469901022229434, "step": 261, "weight_max": 3.255531191825867e-05, "weight_mean": 4.07017159886891e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.1509735486470163e-05 }, { "epoch": 0.08502352750283952, "step": 262, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076708475826308e-06, "weight_min": 0.0, "weight_std": 1.4357781765284017e-05 }, { "epoch": 0.08502352750283952, "step": 262, "weight_max": 3.116124207735993e-05, "weight_mean": 5.0220546654600184e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.102185979107162e-05 }, { "epoch": 0.08502352750283952, "step": 262, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9984591896645725e-06, "weight_min": 0.0, "weight_std": 1.4135979654383846e-05 }, { "epoch": 0.08502352750283952, "step": 262, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.031177362165181e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9090834030066617e-05 }, { "epoch": 0.08534804478338472, "step": 263, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2364512157510035e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4805986211285926e-05 }, { "epoch": 0.08534804478338472, "step": 263, "weight_max": 4.252122016623616e-05, "weight_mean": 9.176381354336627e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.727242488414049e-05 }, { "epoch": 0.08534804478338472, "step": 263, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235715434537269e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4806283616053406e-05 }, { "epoch": 0.08534804478338472, "step": 263, "weight_max": 3.752332122530788e-05, "weight_mean": 7.30121428205166e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4230418855731841e-05 }, { "epoch": 0.08567256206392991, "step": 264, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397426775743952e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.526158484921325e-05 }, { "epoch": 0.08567256206392991, "step": 264, "weight_max": 4.538538789233826e-08, "weight_mean": 6.13946227190354e-09, "weight_min": 0.0, "weight_std": 1.585932807302015e-08 }, { "epoch": 0.08567256206392991, "step": 264, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235979642748134e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806177205173299e-05 }, { "epoch": 0.08567256206392991, "step": 264, "weight_max": 3.782653220696375e-05, "weight_mean": 1.0092857337440364e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.6665237126289867e-05 }, { "epoch": 0.08599707934447509, "step": 265, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.091039828490466e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.687989424681291e-05 }, { "epoch": 0.08599707934447509, "step": 265, "weight_max": 3.967146767536178e-05, "weight_mean": 9.841138307820074e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8221138816443272e-05 }, { "epoch": 0.08599707934447509, "step": 265, "weight_max": 4.187939703115262e-05, "weight_mean": 1.039073868014384e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.923941272252705e-05 }, { "epoch": 0.08599707934447509, "step": 265, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960464012692682e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8457008991390467e-05 }, { "epoch": 0.08632159662502029, "step": 266, "weight_max": 3.967146767536178e-05, "weight_mean": 1.80501738213934e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9449516912573017e-05 }, { "epoch": 0.08632159662502029, "step": 266, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0712161383708008e-05, "weight_min": 0.0, "weight_std": 1.9833305486827157e-05 }, { "epoch": 0.08632159662502029, "step": 266, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155499694519676e-05, "weight_min": 0.0, "weight_std": 1.8815240764524788e-05 }, { "epoch": 0.08632159662502029, "step": 266, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0838976777449716e-05, "weight_min": 0.0, "weight_std": 1.900401730381418e-05 }, { "epoch": 0.08664611390556547, "step": 267, "weight_max": 3.752332122530788e-05, "weight_mean": 8.796949259703979e-06, "weight_min": 0.0, "weight_std": 1.6332392988260835e-05 }, { "epoch": 0.08664611390556547, "step": 267, "weight_max": 1.9113685993943363e-05, "weight_mean": 2.616173333080951e-06, "weight_min": 0.0, "weight_std": 6.696074706269428e-06 }, { "epoch": 0.08664611390556547, "step": 267, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.442437329998938e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2288543985050637e-05 }, { "epoch": 0.08664611390556547, "step": 267, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316847818903625e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5032836017780937e-05 }, { "epoch": 0.08697063118611066, "step": 268, "weight_max": 4.252122016623616e-05, "weight_mean": 9.856188626145013e-06, "weight_min": 0.0, "weight_std": 1.8323418771615252e-05 }, { "epoch": 0.08697063118611066, "step": 268, "weight_max": 9.652274535199012e-10, "weight_mean": 7.275652302851654e-10, "weight_min": 6.196655966128617e-10, "weight_std": 1.2192502563124208e-10 }, { "epoch": 0.08697063118611066, "step": 268, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4356093743117526e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.0196290279272944e-05 }, { "epoch": 0.08697063118611066, "step": 268, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232235581497662e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8945622286992148e-05 }, { "epoch": 0.08729514846665586, "step": 269, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3182639021542855e-06, "weight_min": 0.0, "weight_std": 1.5032265764602926e-05 }, { "epoch": 0.08729514846665586, "step": 269, "weight_max": 2.952862632810138e-05, "weight_mean": 7.62367471907055e-06, "weight_min": 0.0, "weight_std": 1.3234953257779125e-05 }, { "epoch": 0.08729514846665586, "step": 269, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0162864782614633e-05, "weight_min": 0.0, "weight_std": 1.819679710024502e-05 }, { "epoch": 0.08729514846665586, "step": 269, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2070046068402007e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6986510672722943e-05 }, { "epoch": 0.08761966574720104, "grad_norm": 0.4418581426143646, "learning_rate": 9.657142857142859e-06, "loss": 0.3063, "step": 270 }, { "epoch": 0.08761966574720104, "step": 270, "weight_max": 3.998322426923551e-05, "weight_mean": 9.53885137278121e-06, "weight_min": 0.0, "weight_std": 1.7688073057797737e-05 }, { "epoch": 0.08761966574720104, "step": 270, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4457429642789066e-05, "weight_min": 0.0, "weight_std": 1.9973713278886862e-05 }, { "epoch": 0.08761966574720104, "step": 270, "weight_max": 4.187939703115262e-05, "weight_mean": 1.546838211652357e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1355022909119725e-05 }, { "epoch": 0.08761966574720104, "step": 270, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4740902770427056e-05, "weight_min": 0.0, "weight_std": 2.0425375623744912e-05 }, { "epoch": 0.08794418302774623, "step": 271, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.111183317145333e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7015581761370413e-05 }, { "epoch": 0.08794418302774623, "step": 271, "weight_max": 3.0339444492710754e-05, "weight_mean": 7.124575404304778e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3226627743279096e-05 }, { "epoch": 0.08794418302774623, "step": 271, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5688397727208212e-05, "weight_min": 0.0, "weight_std": 1.876831811387092e-05 }, { "epoch": 0.08794418302774623, "step": 271, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.96041126199998e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.845704173319973e-05 }, { "epoch": 0.08826870030829141, "step": 272, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.276614966627676e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7862726963358e-05 }, { "epoch": 0.08826870030829141, "step": 272, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3332766684470698e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8752296455204487e-05 }, { "epoch": 0.08826870030829141, "step": 272, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2463519851735327e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8729930161498487e-05 }, { "epoch": 0.08826870030829141, "step": 272, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578416792355711e-06, "weight_min": 0.0, "weight_std": 1.2947155482834205e-05 }, { "epoch": 0.0885932175888366, "step": 273, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.6160635065461975e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.30529824673431e-05 }, { "epoch": 0.0885932175888366, "step": 273, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.213312761043198e-06, "weight_min": 0.0, "weight_std": 1.4558848306478467e-05 }, { "epoch": 0.0885932175888366, "step": 273, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2728678484563716e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.76590692717582e-05 }, { "epoch": 0.0885932175888366, "step": 273, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.562320721859578e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8546421415521763e-05 }, { "epoch": 0.0889177348693818, "step": 274, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396760570874903e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5261854059644975e-05 }, { "epoch": 0.0889177348693818, "step": 274, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.341899956576526e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5741121387691237e-05 }, { "epoch": 0.0889177348693818, "step": 274, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.76726005319506e-06, "weight_min": 0.0, "weight_std": 1.810291905712802e-05 }, { "epoch": 0.0889177348693818, "step": 274, "weight_max": 1.1429997215373078e-07, "weight_mean": 1.6739809538535155e-08, "weight_min": 5.771870759119224e-10, "weight_std": 3.9709490806671965e-08 }, { "epoch": 0.08924225214992698, "step": 275, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2085725504439324e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.881390744529199e-05 }, { "epoch": 0.08924225214992698, "step": 275, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4522748642775696e-05, "weight_min": 0.0, "weight_std": 2.0177021724521182e-05 }, { "epoch": 0.08924225214992698, "step": 275, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0044001101050526e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8638842448126525e-05 }, { "epoch": 0.08924225214992698, "step": 275, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4046888281882275e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.944095674843993e-05 }, { "epoch": 0.08956676943047218, "step": 276, "weight_max": 4.381950566312298e-05, "weight_mean": 2.1265008399495855e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8219776393380016e-05 }, { "epoch": 0.08956676943047218, "step": 276, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4247352737584151e-05, "weight_min": 0.0, "weight_std": 1.976403837034013e-05 }, { "epoch": 0.08956676943047218, "step": 276, "weight_max": 3.116124207735993e-05, "weight_mean": 4.8610127123538405e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0964939065161161e-05 }, { "epoch": 0.08956676943047218, "step": 276, "weight_max": 3.967146767536178e-05, "weight_mean": 2.259438952023629e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.888653787318617e-05 }, { "epoch": 0.08989128671101736, "step": 277, "weight_max": 2.517239972021912e-09, "weight_mean": 7.51145701194389e-10, "weight_min": 0.0, "weight_std": 7.852281602716005e-10 }, { "epoch": 0.08989128671101736, "step": 277, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.257095664186636e-05, "weight_min": 0.0, "weight_std": 1.7381904399371706e-05 }, { "epoch": 0.08989128671101736, "step": 277, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.48873332364019e-05, "weight_min": 0.0, "weight_std": 2.0751760530401953e-05 }, { "epoch": 0.08989128671101736, "step": 277, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827040927717462e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0822822332556825e-05 }, { "epoch": 0.09021580399156255, "step": 278, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0394302080385387e-05, "weight_min": 0.0, "weight_std": 1.9264311049482785e-05 }, { "epoch": 0.09021580399156255, "step": 278, "weight_max": 3.813097282545641e-05, "weight_mean": 9.153309292742051e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.689562304818537e-05 }, { "epoch": 0.09021580399156255, "step": 278, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6711477655917406e-05, "weight_min": 0.0, "weight_std": 1.9346611225046217e-05 }, { "epoch": 0.09021580399156255, "step": 278, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9381088350201026e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0757179299835116e-05 }, { "epoch": 0.09054032127210775, "step": 279, "weight_max": 1.7859370649020434e-09, "weight_mean": 9.105910470097456e-10, "weight_min": 0.0, "weight_std": 5.178442119557758e-10 }, { "epoch": 0.09054032127210775, "step": 279, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4324392547423486e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.25269116324489e-05 }, { "epoch": 0.09054032127210775, "step": 279, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.325160716689425e-06, "weight_min": 0.0, "weight_std": 1.2216021787025966e-05 }, { "epoch": 0.09054032127210775, "step": 279, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5317671568482183e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.114404742314946e-05 }, { "epoch": 0.09086483855265293, "grad_norm": 2.9732701778411865, "learning_rate": 9.636734693877552e-06, "loss": 0.1916, "step": 280 }, { "epoch": 0.09086483855265293, "step": 280, "weight_max": 2.7420132028055377e-05, "weight_mean": 3.4282090837223222e-06, "weight_min": 6.488223291967188e-10, "weight_std": 9.694201253296342e-06 }, { "epoch": 0.09086483855265293, "step": 280, "weight_max": 3.998322426923551e-05, "weight_mean": 9.764935384737328e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8086291674990207e-05 }, { "epoch": 0.09086483855265293, "step": 280, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2305183190619573e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7425747500965372e-05 }, { "epoch": 0.09086483855265293, "step": 280, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.874905137810856e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0431034499779344e-05 }, { "epoch": 0.09118935583319812, "step": 281, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7763552023097873e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.0853349269600585e-05 }, { "epoch": 0.09118935583319812, "step": 281, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.4073302635515574e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5257605809892993e-05 }, { "epoch": 0.09118935583319812, "step": 281, "weight_max": 1.0310614584341238e-07, "weight_mean": 1.3494880768405437e-08, "weight_min": 5.771870759119224e-10, "weight_std": 3.620877819798807e-08 }, { "epoch": 0.09118935583319812, "step": 281, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882492248725612e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3806406968797091e-05 }, { "epoch": 0.0915138731137433, "step": 282, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9435014110058546e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0876597773167305e-05 }, { "epoch": 0.0915138731137433, "step": 282, "weight_max": 3.36925259034615e-05, "weight_mean": 6.364628006849671e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2579029316839296e-05 }, { "epoch": 0.0915138731137433, "step": 282, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0049437807756476e-05, "weight_min": 0.0, "weight_std": 2.1511175873456523e-05 }, { "epoch": 0.0915138731137433, "step": 282, "weight_max": 2.872879122151062e-05, "weight_mean": 6.412914899556199e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0383709195593838e-05 }, { "epoch": 0.0918383903942885, "step": 283, "weight_max": 3.998322426923551e-05, "weight_mean": 9.57606880547246e-06, "weight_min": 0.0, "weight_std": 1.7752850908436812e-05 }, { "epoch": 0.0918383903942885, "step": 283, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.653438610257581e-06, "weight_min": 0.0, "weight_std": 1.786704706319142e-05 }, { "epoch": 0.0918383903942885, "step": 283, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4096347513259389e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9558834537747316e-05 }, { "epoch": 0.0918383903942885, "step": 283, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998758868168807e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.413585869158851e-05 }, { "epoch": 0.09216290767483369, "step": 284, "weight_max": 1.1273959543700585e-09, "weight_mean": 7.023253645321859e-10, "weight_min": 0.0, "weight_std": 3.3928965192941973e-10 }, { "epoch": 0.09216290767483369, "step": 284, "weight_max": 3.199401908204891e-05, "weight_mean": 6.934988959983457e-06, "weight_min": 0.0, "weight_std": 1.3028429748374037e-05 }, { "epoch": 0.09216290767483369, "step": 284, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6908330508158542e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0426778064575046e-05 }, { "epoch": 0.09216290767483369, "step": 284, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3402468539425172e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8669494238565676e-05 }, { "epoch": 0.09248742495537887, "step": 285, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5116369468159974e-05, "weight_min": 0.0, "weight_std": 2.0882072931271978e-05 }, { "epoch": 0.09248742495537887, "step": 285, "weight_max": 6.8851719881024565e-09, "weight_mean": 1.3586474167937013e-09, "weight_min": 0.0, "weight_std": 2.2466273286170235e-09 }, { "epoch": 0.09248742495537887, "step": 285, "weight_max": 3.813097282545641e-05, "weight_mean": 8.871690624800976e-06, "weight_min": 0.0, "weight_std": 1.648651232244447e-05 }, { "epoch": 0.09248742495537887, "step": 285, "weight_max": 3.813097282545641e-05, "weight_mean": 4.771708518092055e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3479180779540911e-05 }, { "epoch": 0.09281194223592407, "step": 286, "weight_max": 3.0886087188264355e-05, "weight_mean": 6.070280051062582e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0939933417830616e-05 }, { "epoch": 0.09281194223592407, "step": 286, "weight_max": 3.752332122530788e-05, "weight_mean": 4.692174570664065e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3265786947158631e-05 }, { "epoch": 0.09281194223592407, "step": 286, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998510121367872e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135959645500407e-05 }, { "epoch": 0.09281194223592407, "step": 286, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.129622412729077e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6957343177637085e-05 }, { "epoch": 0.09313645951646925, "step": 287, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.472414078307338e-05, "weight_min": 0.0, "weight_std": 2.03302606678335e-05 }, { "epoch": 0.09313645951646925, "step": 287, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.534895070828497e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7660047888057306e-05 }, { "epoch": 0.09313645951646925, "step": 287, "weight_max": 3.967146767536178e-05, "weight_mean": 9.576513548381627e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.774129668774549e-05 }, { "epoch": 0.09313645951646925, "step": 287, "weight_max": 3.752332122530788e-05, "weight_mean": 9.161559319181833e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6958496416918933e-05 }, { "epoch": 0.09346097679701444, "step": 288, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.869833436619956e-06, "weight_min": 0.0, "weight_std": 1.6648875316604972e-05 }, { "epoch": 0.09346097679701444, "step": 288, "weight_max": 4.187939703115262e-05, "weight_mean": 9.28615281736711e-06, "weight_min": 7.298507354036587e-10, "weight_std": 1.734380748530384e-05 }, { "epoch": 0.09346097679701444, "step": 288, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.261071681976318e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7292115444433875e-05 }, { "epoch": 0.09346097679701444, "step": 288, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920674655295443e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3915963791077957e-05 }, { "epoch": 0.09378549407755964, "step": 289, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0275733984599356e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9038134269067086e-05 }, { "epoch": 0.09378549407755964, "step": 289, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.8187325622420758e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9690445697051473e-05 }, { "epoch": 0.09378549407755964, "step": 289, "weight_max": 3.602553260861896e-05, "weight_mean": 7.932400876597967e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4862795978842769e-05 }, { "epoch": 0.09378549407755964, "step": 289, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4777939213672653e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0468078218982555e-05 }, { "epoch": 0.09411001135810482, "grad_norm": 0.561366617679596, "learning_rate": 9.616326530612246e-06, "loss": 0.1543, "step": 290 }, { "epoch": 0.09411001135810482, "step": 290, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5206624084385112e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.090461930492893e-05 }, { "epoch": 0.09411001135810482, "step": 290, "weight_max": 3.967146767536178e-05, "weight_mean": 4.95957829116378e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4025721611687914e-05 }, { "epoch": 0.09411001135810482, "step": 290, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7244212358491495e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.876701571745798e-05 }, { "epoch": 0.09411001135810482, "step": 290, "weight_max": 3.28377827827353e-05, "weight_mean": 4.10717348131584e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1608919521677308e-05 }, { "epoch": 0.09443452863865001, "step": 291, "weight_max": 4.252122016623616e-05, "weight_mean": 1.50936575664673e-05, "weight_min": 0.0, "weight_std": 2.0872323148068972e-05 }, { "epoch": 0.09443452863865001, "step": 291, "weight_max": 4.187939703115262e-05, "weight_mean": 1.524814433651045e-05, "weight_min": 0.0, "weight_std": 1.8633445506566204e-05 }, { "epoch": 0.09443452863865001, "step": 291, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3856477380613796e-05, "weight_min": 0.0, "weight_std": 1.947449527506251e-05 }, { "epoch": 0.09443452863865001, "step": 291, "weight_max": 2.595289316786875e-09, "weight_mean": 8.173965948543582e-10, "weight_min": 0.0, "weight_std": 7.559803338885729e-10 }, { "epoch": 0.09475904591919519, "step": 292, "weight_max": 4.187939703115262e-05, "weight_mean": 1.54787594510708e-05, "weight_min": 0.0, "weight_std": 1.8162971173296683e-05 }, { "epoch": 0.09475904591919519, "step": 292, "weight_max": 4.252122016623616e-05, "weight_mean": 7.651092346350197e-06, "weight_min": 0.0, "weight_std": 1.553263427922502e-05 }, { "epoch": 0.09475904591919519, "step": 292, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313412531104404e-05, "weight_min": 0.0, "weight_std": 1.9107890693703666e-05 }, { "epoch": 0.09475904591919519, "step": 292, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156022325536469e-06, "weight_min": 0.0, "weight_std": 1.4581122741219588e-05 }, { "epoch": 0.09508356319974039, "step": 293, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.258284767885925e-06, "weight_min": 0.0, "weight_std": 1.493851868872298e-05 }, { "epoch": 0.09508356319974039, "step": 293, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0314684004697483e-05, "weight_min": 0.0, "weight_std": 1.908904050651472e-05 }, { "epoch": 0.09508356319974039, "step": 293, "weight_max": 2.202445102739148e-05, "weight_mean": 2.7844414489663905e-06, "weight_min": 0.0, "weight_std": 7.774616278766189e-06 }, { "epoch": 0.09508356319974039, "step": 293, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5939047443680465e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7923086488735862e-05 }, { "epoch": 0.09540808048028558, "step": 294, "weight_max": 3.572963396436535e-05, "weight_mean": 7.798021215421613e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4640130757470615e-05 }, { "epoch": 0.09540808048028558, "step": 294, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0253233085677493e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8932740204036236e-05 }, { "epoch": 0.09540808048028558, "step": 294, "weight_max": 3.255531191825867e-05, "weight_mean": 7.419184839818627e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3578368452726863e-05 }, { "epoch": 0.09540808048028558, "step": 294, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4268292943597771e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9708884792635217e-05 }, { "epoch": 0.09573259776083076, "step": 295, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4970982192608062e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.068218236672692e-05 }, { "epoch": 0.09573259776083076, "step": 295, "weight_max": 3.255531191825867e-05, "weight_mean": 7.99969166109804e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4814140740782022e-05 }, { "epoch": 0.09573259776083076, "step": 295, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.4262535842135549e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9685186998685822e-05 }, { "epoch": 0.09573259776083076, "step": 295, "weight_max": 4.187939703115262e-05, "weight_mean": 1.7413374735042453e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.964531111298129e-05 }, { "epoch": 0.09605711504137596, "step": 296, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.857455754361581e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0018951545353048e-05 }, { "epoch": 0.09605711504137596, "step": 296, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5466397599084303e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.134460737579502e-05 }, { "epoch": 0.09605711504137596, "step": 296, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8496200937079266e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.023226443270687e-05 }, { "epoch": 0.09605711504137596, "step": 296, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4962282875785604e-05, "weight_min": 0.0, "weight_std": 1.927568155224435e-05 }, { "epoch": 0.09638163232192114, "step": 297, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.085762030736078e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7024585758917965e-05 }, { "epoch": 0.09638163232192114, "step": 297, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.572363524057437e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.791414433682803e-05 }, { "epoch": 0.09638163232192114, "step": 297, "weight_max": 3.28377827827353e-05, "weight_mean": 8.244835044024512e-06, "weight_min": 0.0, "weight_std": 1.5093361071194522e-05 }, { "epoch": 0.09638163232192114, "step": 297, "weight_max": 3.602553260861896e-05, "weight_mean": 7.740565706626512e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4583754818886518e-05 }, { "epoch": 0.09670614960246633, "step": 298, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3085648788546678e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8133661797037348e-05 }, { "epoch": 0.09670614960246633, "step": 298, "weight_max": 3.602553260861896e-05, "weight_mean": 1.3111231055518147e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8111728422809392e-05 }, { "epoch": 0.09670614960246633, "step": 298, "weight_max": 3.2345194966865165e-08, "weight_mean": 5.631651145421301e-09, "weight_min": 5.864964625068581e-10, "weight_std": 1.112108272138812e-08 }, { "epoch": 0.09670614960246633, "step": 298, "weight_max": 7.215334107968374e-08, "weight_mean": 9.616045737459444e-09, "weight_min": 0.0, "weight_std": 2.527505671423569e-08 }, { "epoch": 0.09703066688301153, "step": 299, "weight_max": 4.252122016623616e-05, "weight_mean": 9.563420462654904e-06, "weight_min": 0.0, "weight_std": 1.7852629753178917e-05 }, { "epoch": 0.09703066688301153, "step": 299, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.3316819099127315e-06, "weight_min": 0.0, "weight_std": 9.421855793334544e-06 }, { "epoch": 0.09703066688301153, "step": 299, "weight_max": 8.858740329742432e-06, "weight_mean": 1.139179971687554e-06, "weight_min": 5.958802340444436e-10, "weight_std": 3.120359451713739e-06 }, { "epoch": 0.09703066688301153, "step": 299, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.884595783660188e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8323073163628578e-05 }, { "epoch": 0.09735518416355671, "grad_norm": 0.63735032081604, "learning_rate": 9.595918367346939e-06, "loss": 0.4266, "step": 300 }, { "epoch": 0.09735518416355671, "eval_accuracy": 0.494, "eval_auc": 0.4921803739738523, "eval_loss": 2.160552740097046, "eval_runtime": 155.7681, "eval_samples_per_second": 12.84, "eval_steps_per_second": 1.605, "step": 300 }, { "epoch": 0.09735518416355671, "step": 300, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4730416296515614e-05, "weight_min": 0.0, "weight_std": 2.0365854652482085e-05 }, { "epoch": 0.09735518416355671, "step": 300, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5551297110505402e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1471367290359922e-05 }, { "epoch": 0.09735518416355671, "step": 300, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0768230721587315e-06, "weight_min": 0.0, "weight_std": 1.4357735381054226e-05 }, { "epoch": 0.09735518416355671, "step": 300, "weight_max": 4.9337902652268895e-09, "weight_mean": 1.4094403422149071e-09, "weight_min": 6.100955851628953e-10, "weight_std": 1.486052725319098e-09 }, { "epoch": 0.0976797014441019, "step": 301, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2147615052526817e-05, "weight_min": 0.0, "weight_std": 1.6952802980085835e-05 }, { "epoch": 0.0976797014441019, "step": 301, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0630701581249014e-05, "weight_min": 0.0, "weight_std": 1.9683255231939256e-05 }, { "epoch": 0.0976797014441019, "step": 301, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3873506759409793e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9416685972828418e-05 }, { "epoch": 0.0976797014441019, "step": 301, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2848259757447522e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8161665138904937e-05 }, { "epoch": 0.09800421872464708, "step": 302, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2218366211745888e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.871000131359324e-05 }, { "epoch": 0.09800421872464708, "step": 302, "weight_max": 3.312147964606993e-05, "weight_mean": 7.933144843264017e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4706830370414536e-05 }, { "epoch": 0.09800421872464708, "step": 302, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9891709598596208e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.1333677068469115e-05 }, { "epoch": 0.09800421872464708, "step": 302, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5198440451058559e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.1011606804677285e-05 }, { "epoch": 0.09832873600519228, "step": 303, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3181406757212244e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8361248294240795e-05 }, { "epoch": 0.09832873600519228, "step": 303, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.998693712987006e-06, "weight_min": 0.0, "weight_std": 1.8524919141782448e-05 }, { "epoch": 0.09832873600519228, "step": 303, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4728487258253153e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0354793377919123e-05 }, { "epoch": 0.09832873600519228, "step": 303, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8747488866210915e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.013798621192109e-05 }, { "epoch": 0.09865325328573747, "step": 304, "weight_max": 3.602553260861896e-05, "weight_mean": 4.503820946410997e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2736695680359844e-05 }, { "epoch": 0.09865325328573747, "step": 304, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2149924259574618e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8181732230004855e-05 }, { "epoch": 0.09865325328573747, "step": 304, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.51207312027691e-06, "weight_min": 0.0, "weight_std": 1.7694130292511545e-05 }, { "epoch": 0.09865325328573747, "step": 304, "weight_max": 3.998322426923551e-05, "weight_mean": 4.999816155759618e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.413543304806808e-05 }, { "epoch": 0.09897777056628265, "step": 305, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4837853086646646e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.064110231003724e-05 }, { "epoch": 0.09897777056628265, "step": 305, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4265206118579954e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.00185168068856e-05 }, { "epoch": 0.09897777056628265, "step": 305, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578368589136517e-06, "weight_min": 0.0, "weight_std": 1.294717367272824e-05 }, { "epoch": 0.09897777056628265, "step": 305, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430258286447497e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2527792023320217e-05 }, { "epoch": 0.09930228784682785, "step": 306, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155925919098081e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4581161849491764e-05 }, { "epoch": 0.09930228784682785, "step": 306, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.802245247352403e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6549798601772636e-05 }, { "epoch": 0.09930228784682785, "step": 306, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.765915142954327e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5373731002910063e-05 }, { "epoch": 0.09930228784682785, "step": 306, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3648987078340724e-05, "weight_min": 0.0, "weight_std": 1.8955148334498517e-05 }, { "epoch": 0.09962680512737303, "step": 307, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.359894758934388e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.907840123749338e-05 }, { "epoch": 0.09962680512737303, "step": 307, "weight_max": 3.998322426923551e-05, "weight_mean": 5.67948836760479e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3991269952384755e-05 }, { "epoch": 0.09962680512737303, "step": 307, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.9651908991800155e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1212700883334037e-05 }, { "epoch": 0.09962680512737303, "step": 307, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0237403330393136e-05, "weight_min": 0.0, "weight_std": 1.896952926472295e-05 }, { "epoch": 0.09995132240791822, "step": 308, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3465121810440905e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.885675374069251e-05 }, { "epoch": 0.09995132240791822, "step": 308, "weight_max": 3.843663216684945e-05, "weight_mean": 1.83866432053037e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6153964679688215e-05 }, { "epoch": 0.09995132240791822, "step": 308, "weight_max": 3.998322426923551e-05, "weight_mean": 9.918675459630322e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.836494993767701e-05 }, { "epoch": 0.09995132240791822, "step": 308, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076878096588189e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357713553181384e-05 }, { "epoch": 0.10027583968846342, "step": 309, "weight_max": 3.722132169059478e-05, "weight_mean": 8.900771717890166e-06, "weight_min": 0.0, "weight_std": 1.650224658078514e-05 }, { "epoch": 0.10027583968846342, "step": 309, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0470941560924985e-05, "weight_min": 0.0, "weight_std": 1.939023786690086e-05 }, { "epoch": 0.10027583968846342, "step": 309, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.430201609764481e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9905537556041963e-05 }, { "epoch": 0.10027583968846342, "step": 309, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3166274584364146e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9224837160436437e-05 }, { "epoch": 0.1006003569690086, "grad_norm": 0.6117706894874573, "learning_rate": 9.575510204081634e-06, "loss": 0.1479, "step": 310 }, { "epoch": 0.1006003569690086, "step": 310, "weight_max": 4.187939703115262e-05, "weight_mean": 1.523347782494966e-05, "weight_min": 0.0, "weight_std": 2.1042180378572084e-05 }, { "epoch": 0.1006003569690086, "step": 310, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.565749340050388e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.609939499758184e-05 }, { "epoch": 0.1006003569690086, "step": 310, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.15704869030742e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.458070801163558e-05 }, { "epoch": 0.1006003569690086, "step": 310, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235824119154131e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.480623905081302e-05 }, { "epoch": 0.10092487424955379, "step": 311, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3618584489449859e-05, "weight_min": 0.0, "weight_std": 1.813488415791653e-05 }, { "epoch": 0.10092487424955379, "step": 311, "weight_max": 3.632264997577295e-05, "weight_mean": 4.544362582237227e-06, "weight_min": 0.0, "weight_std": 1.2840371709899046e-05 }, { "epoch": 0.10092487424955379, "step": 311, "weight_max": 4.447596802492626e-05, "weight_mean": 1.0442108759889379e-05, "weight_min": 0.0, "weight_std": 1.938644163601566e-05 }, { "epoch": 0.10092487424955379, "step": 311, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.8703003661357798e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0062032490386628e-05 }, { "epoch": 0.10124939153009897, "step": 312, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844524028158048e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3697258509637322e-05 }, { "epoch": 0.10124939153009897, "step": 312, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.72890504472889e-05, "weight_min": 0.0, "weight_std": 1.8833563444786705e-05 }, { "epoch": 0.10124939153009897, "step": 312, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4631108570029028e-05, "weight_min": 0.0, "weight_std": 2.045348628598731e-05 }, { "epoch": 0.10124939153009897, "step": 312, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1560164138209075e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4581124560208991e-05 }, { "epoch": 0.10157390881064417, "step": 313, "weight_max": 3.998322426923551e-05, "weight_mean": 9.464710274187382e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.756058554747142e-05 }, { "epoch": 0.10157390881064417, "step": 313, "weight_max": 4.187939703115262e-05, "weight_mean": 1.467453967052279e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0376508473418653e-05 }, { "epoch": 0.10157390881064417, "step": 313, "weight_max": 2.846461757144425e-05, "weight_mean": 3.5586606372817187e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0063526133308187e-05 }, { "epoch": 0.10157390881064417, "step": 313, "weight_max": 2.639517333591357e-05, "weight_mean": 3.3002725103870034e-06, "weight_min": 0.0, "weight_std": 9.331749424745794e-06 }, { "epoch": 0.10189842609118936, "step": 314, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9285404050606303e-05, "weight_min": 0.0, "weight_std": 2.073346149700228e-05 }, { "epoch": 0.10189842609118936, "step": 314, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.7302907508565113e-05, "weight_min": 0.0, "weight_std": 1.8576540242065676e-05 }, { "epoch": 0.10189842609118936, "step": 314, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.023966251523234e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8995200662175193e-05 }, { "epoch": 0.10189842609118936, "step": 314, "weight_max": 3.967146767536178e-05, "weight_mean": 9.242990927305073e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7174272215925157e-05 }, { "epoch": 0.10222294337173454, "step": 315, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0044056580227334e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8638807887327857e-05 }, { "epoch": 0.10222294337173454, "step": 315, "weight_max": 3.843663216684945e-05, "weight_mean": 5.67891902392148e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3459802175930236e-05 }, { "epoch": 0.10222294337173454, "step": 315, "weight_max": 4.252122016623616e-05, "weight_mean": 6.189477971929591e-06, "weight_min": 0.0, "weight_std": 1.488223915657727e-05 }, { "epoch": 0.10222294337173454, "step": 315, "weight_max": 4.187939703115262e-05, "weight_mean": 1.7859943909570575e-05, "weight_min": 0.0, "weight_std": 2.000217318709474e-05 }, { "epoch": 0.10254746065227974, "step": 316, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1910493412869982e-05, "weight_min": 0.0, "weight_std": 1.852042623795569e-05 }, { "epoch": 0.10254746065227974, "step": 316, "weight_max": 3.116124207735993e-05, "weight_mean": 3.895869667758234e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.10168739411165e-05 }, { "epoch": 0.10254746065227974, "step": 316, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8480492144590244e-05, "weight_min": 0.0, "weight_std": 2.0156963728368282e-05 }, { "epoch": 0.10254746065227974, "step": 316, "weight_max": 3.813097282545641e-05, "weight_mean": 9.125837095780298e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6743493688409217e-05 }, { "epoch": 0.10287197793282492, "step": 317, "weight_max": 3.967146767536178e-05, "weight_mean": 1.5532401448581368e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.724656976875849e-05 }, { "epoch": 0.10287197793282492, "step": 317, "weight_max": 3.813097282545641e-05, "weight_mean": 9.19914964470081e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.704034184513148e-05 }, { "epoch": 0.10287197793282492, "step": 317, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882776011072565e-06, "weight_min": 0.0, "weight_std": 1.3806292372464668e-05 }, { "epoch": 0.10287197793282492, "step": 317, "weight_max": 3.4268457966391e-05, "weight_mean": 4.2842025322897825e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2115468962292653e-05 }, { "epoch": 0.10319649521337011, "step": 318, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4851384548819624e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0552048226818442e-05 }, { "epoch": 0.10319649521337011, "step": 318, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4686791473650374e-05, "weight_min": 0.0, "weight_std": 2.0282652258174494e-05 }, { "epoch": 0.10319649521337011, "step": 318, "weight_max": 4.187939703115262e-05, "weight_mean": 9.628414773033e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.7917102013598196e-05 }, { "epoch": 0.10319649521337011, "step": 318, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4731175724591594e-05, "weight_min": 0.0, "weight_std": 2.037026933976449e-05 }, { "epoch": 0.10352101249391531, "step": 319, "weight_max": 3.5434961318969727e-05, "weight_mean": 6.792405656597111e-06, "weight_min": 0.0, "weight_std": 1.3329285138752311e-05 }, { "epoch": 0.10352101249391531, "step": 319, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5006111425464042e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0758763639605604e-05 }, { "epoch": 0.10352101249391531, "step": 319, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998689746571472e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4135886885924265e-05 }, { "epoch": 0.10352101249391531, "step": 319, "weight_max": 4.252122016623616e-05, "weight_mean": 9.52780646912288e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.77964993781643e-05 }, { "epoch": 0.10384552977446049, "grad_norm": 2.4026412963867188, "learning_rate": 9.555102040816327e-06, "loss": 0.2666, "step": 320 }, { "epoch": 0.10384552977446049, "step": 320, "weight_max": 3.998322426923551e-05, "weight_mean": 4.999375505576609e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4135609490040224e-05 }, { "epoch": 0.10384552977446049, "step": 320, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5033088604686782e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0910225430270657e-05 }, { "epoch": 0.10384552977446049, "step": 320, "weight_max": 3.006794759130571e-05, "weight_mean": 7.4849644988717046e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.3855302313459106e-05 }, { "epoch": 0.10384552977446049, "step": 320, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3854785720468499e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.947025339177344e-05 }, { "epoch": 0.10417004705500568, "step": 321, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.881285222945735e-06, "weight_min": 0.0, "weight_std": 1.8304805053048767e-05 }, { "epoch": 0.10417004705500568, "step": 321, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9708115360117517e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1335092242225073e-05 }, { "epoch": 0.10417004705500568, "step": 321, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313875463907607e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9107606931356713e-05 }, { "epoch": 0.10417004705500568, "step": 321, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0325733455829322e-05, "weight_min": 0.0, "weight_std": 1.852124296419788e-05 }, { "epoch": 0.10449456433555086, "step": 322, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.808896720642224e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8192038623965345e-05 }, { "epoch": 0.10449456433555086, "step": 322, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745372153702192e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8141949112759903e-05 }, { "epoch": 0.10449456433555086, "step": 322, "weight_max": 4.187939703115262e-05, "weight_mean": 2.099993434967473e-05, "weight_min": 9.893572627817093e-10, "weight_std": 1.791316935850773e-05 }, { "epoch": 0.10449456433555086, "step": 322, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767008249473292e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3481077075994108e-05 }, { "epoch": 0.10481908161609606, "step": 323, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.78741128265392e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4434642253036145e-05 }, { "epoch": 0.10481908161609606, "step": 323, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0769635890901554e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357678082888015e-05 }, { "epoch": 0.10481908161609606, "step": 323, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.309007711824961e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.816003532439936e-05 }, { "epoch": 0.10481908161609606, "step": 323, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.093500921153463e-06, "weight_min": 0.0, "weight_std": 1.6897507521207444e-05 }, { "epoch": 0.10514359889664125, "step": 324, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5964476915542036e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.791106114978902e-05 }, { "epoch": 0.10514359889664125, "step": 324, "weight_max": 3.813097282545641e-05, "weight_mean": 9.495195627096109e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7580621715751477e-05 }, { "epoch": 0.10514359889664125, "step": 324, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8586892110761255e-05, "weight_min": 1.0888551171817085e-09, "weight_std": 2.061331360891927e-05 }, { "epoch": 0.10514359889664125, "step": 324, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793252744799247e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0726281288953032e-05 }, { "epoch": 0.10546811617718643, "step": 325, "weight_max": 4.187939703115262e-05, "weight_mean": 1.468455775466282e-05, "weight_min": 0.0, "weight_std": 1.9923067156923935e-05 }, { "epoch": 0.10546811617718643, "step": 325, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6199224774027243e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8004351659328677e-05 }, { "epoch": 0.10546811617718643, "step": 325, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3392219443630893e-05, "weight_min": 0.0, "weight_std": 1.8613265638123266e-05 }, { "epoch": 0.10546811617718643, "step": 325, "weight_max": 3.255531191825867e-05, "weight_mean": 8.070741387200542e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4939739230612759e-05 }, { "epoch": 0.10579263345773163, "step": 326, "weight_max": 4.252122016623616e-05, "weight_mean": 1.321285617450485e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.7906455468619242e-05 }, { "epoch": 0.10579263345773163, "step": 326, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.487690289854072e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0781526473001577e-05 }, { "epoch": 0.10579263345773163, "step": 326, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3562388378195465e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9658960809465498e-05 }, { "epoch": 0.10579263345773163, "step": 326, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0196959919994697e-05, "weight_min": 0.0, "weight_std": 1.8902979718404822e-05 }, { "epoch": 0.10611715073827681, "step": 327, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3756984117208049e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9236174921388738e-05 }, { "epoch": 0.10611715073827681, "step": 327, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.958994269254617e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.844211328716483e-05 }, { "epoch": 0.10611715073827681, "step": 327, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843915121455211e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.36975031637121e-05 }, { "epoch": 0.10611715073827681, "step": 327, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.89228101389017e-06, "weight_min": 0.0, "weight_std": 1.7467489669797942e-05 }, { "epoch": 0.106441668018822, "step": 328, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998380518372869e-06, "weight_min": 0.0, "weight_std": 1.4136011486698408e-05 }, { "epoch": 0.106441668018822, "step": 328, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4133780496194959e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.961044836207293e-05 }, { "epoch": 0.106441668018822, "step": 328, "weight_max": 3.9360926166409627e-05, "weight_mean": 7.124519470380619e-06, "weight_min": 0.0, "weight_std": 1.4412748896575067e-05 }, { "epoch": 0.106441668018822, "step": 328, "weight_max": 2.952862632810138e-05, "weight_mean": 3.6919172998750582e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0439606739964802e-05 }, { "epoch": 0.1067661852993672, "step": 329, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827134150924394e-06, "weight_min": 0.0, "weight_std": 1.082278413377935e-05 }, { "epoch": 0.1067661852993672, "step": 329, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4920688045094721e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0602932636393234e-05 }, { "epoch": 0.1067661852993672, "step": 329, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153818038816098e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.88027297554072e-05 }, { "epoch": 0.1067661852993672, "step": 329, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8276434275321662e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9857445295201614e-05 }, { "epoch": 0.10709070257991238, "grad_norm": 0.4135395586490631, "learning_rate": 9.53469387755102e-06, "loss": 0.0647, "step": 330 }, { "epoch": 0.10709070257991238, "step": 330, "weight_max": 4.187939703115262e-05, "weight_mean": 6.524349828396225e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.473409793106839e-05 }, { "epoch": 0.10709070257991238, "step": 330, "weight_max": 1.4281727999332361e-05, "weight_mean": 1.7868359236672404e-06, "weight_min": 5.771870759119224e-10, "weight_std": 5.048699222243158e-06 }, { "epoch": 0.10709070257991238, "step": 330, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4660665328847244e-05, "weight_min": 0.0, "weight_std": 2.0302299162722193e-05 }, { "epoch": 0.10709070257991238, "step": 330, "weight_max": 4.187939703115262e-05, "weight_mean": 1.990910641325172e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1421135897981003e-05 }, { "epoch": 0.10741521986045757, "step": 331, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9629864254966378e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.958646680577658e-05 }, { "epoch": 0.10741521986045757, "step": 331, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313616257917602e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9107767002424225e-05 }, { "epoch": 0.10741521986045757, "step": 331, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9985515033768024e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4135943274595775e-05 }, { "epoch": 0.10741521986045757, "step": 331, "weight_max": 3.843663216684945e-05, "weight_mean": 9.234749086317606e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7116268281824887e-05 }, { "epoch": 0.10773973714100275, "step": 332, "weight_max": 3.813097282545641e-05, "weight_mean": 1.125148446590174e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.6468056855956092e-05 }, { "epoch": 0.10773973714100275, "step": 332, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.015329053188907e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.879853880382143e-05 }, { "epoch": 0.10773973714100275, "step": 332, "weight_max": 4.187939703115262e-05, "weight_mean": 9.37573622650234e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7516036677989177e-05 }, { "epoch": 0.10773973714100275, "step": 332, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.34670060814824e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7320555343758315e-05 }, { "epoch": 0.10806425442154795, "step": 333, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.130581929639447e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.700379652902484e-05 }, { "epoch": 0.10806425442154795, "step": 333, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.511195457889698e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7651953385211527e-05 }, { "epoch": 0.10806425442154795, "step": 333, "weight_max": 3.752332122530788e-05, "weight_mean": 1.2866978977399413e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7826814655563794e-05 }, { "epoch": 0.10806425442154795, "step": 333, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.070479629561305e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9453831555438228e-05 }, { "epoch": 0.10838877170209314, "step": 334, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.572381713951472e-06, "weight_min": 0.0, "weight_std": 1.7914131603902206e-05 }, { "epoch": 0.10838877170209314, "step": 334, "weight_max": 4.187939703115262e-05, "weight_mean": 9.13113217393402e-06, "weight_min": 0.0, "weight_std": 1.714595600788016e-05 }, { "epoch": 0.10838877170209314, "step": 334, "weight_max": 3.782653220696375e-05, "weight_mean": 9.307830623583868e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7233100152225234e-05 }, { "epoch": 0.10838877170209314, "step": 334, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.884473001875449e-06, "weight_min": 0.0, "weight_std": 1.8323149561183527e-05 }, { "epoch": 0.10871328898263832, "step": 335, "weight_max": 3.967146767536178e-05, "weight_mean": 9.388888429384679e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7420234144083224e-05 }, { "epoch": 0.10871328898263832, "step": 335, "weight_max": 4.187939703115262e-05, "weight_mean": 1.977896499738563e-05, "weight_min": 7.829137893544669e-10, "weight_std": 2.0945821233908646e-05 }, { "epoch": 0.10871328898263832, "step": 335, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.263106216560118e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9218347006244585e-05 }, { "epoch": 0.10871328898263832, "step": 335, "weight_max": 4.187939703115262e-05, "weight_mean": 1.935175714606885e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0762548956554383e-05 }, { "epoch": 0.10903780626318352, "step": 336, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2763372069457546e-05, "weight_min": 0.0, "weight_std": 1.8167098460253328e-05 }, { "epoch": 0.10903780626318352, "step": 336, "weight_max": 3.572963396436535e-05, "weight_mean": 6.2916365095588844e-06, "weight_min": 0.0, "weight_std": 1.2761770449287724e-05 }, { "epoch": 0.10903780626318352, "step": 336, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274889064021409e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9038656319025904e-05 }, { "epoch": 0.10903780626318352, "step": 336, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.652543667471036e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.782138497219421e-05 }, { "epoch": 0.1093623235437287, "step": 337, "weight_max": 4.252122016623616e-05, "weight_mean": 7.926132639113348e-06, "weight_min": 0.0, "weight_std": 1.5772768165334128e-05 }, { "epoch": 0.1093623235437287, "step": 337, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.616501428245101e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.3052805115876254e-05 }, { "epoch": 0.1093623235437287, "step": 337, "weight_max": 4.252122016623616e-05, "weight_mean": 2.142875200661365e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1605601432383992e-05 }, { "epoch": 0.1093623235437287, "step": 337, "weight_max": 2.716206836339552e-05, "weight_mean": 3.395826979613048e-06, "weight_min": 0.0, "weight_std": 9.603011676517781e-06 }, { "epoch": 0.10968684082427389, "step": 338, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550662409514189e-05, "weight_min": 0.0, "weight_std": 1.9535338651621714e-05 }, { "epoch": 0.10968684082427389, "step": 338, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.903212801669724e-06, "weight_min": 0.0, "weight_std": 1.6699748812243342e-05 }, { "epoch": 0.10968684082427389, "step": 338, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.360932381241582e-06, "weight_min": 0.0, "weight_std": 1.759994847816415e-05 }, { "epoch": 0.10968684082427389, "step": 338, "weight_max": 4.187939703115262e-05, "weight_mean": 1.011699168884661e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.874674671853427e-05 }, { "epoch": 0.11001135810481909, "step": 339, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430693934409646e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2527615581348073e-05 }, { "epoch": 0.11001135810481909, "step": 339, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.023995537252631e-05, "weight_min": 0.0, "weight_std": 1.899501876323484e-05 }, { "epoch": 0.11001135810481909, "step": 339, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.990553735406138e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.6426178262918256e-05 }, { "epoch": 0.11001135810481909, "step": 339, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156515726412181e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4580922652385198e-05 }, { "epoch": 0.11033587538536427, "grad_norm": 4.209699630737305, "learning_rate": 9.514285714285715e-06, "loss": 0.0798, "step": 340 }, { "epoch": 0.11033587538536427, "step": 340, "weight_max": 2.3459858500274322e-09, "weight_mean": 1.1854686121637314e-09, "weight_min": 6.053385570581327e-10, "weight_std": 7.387423450744279e-10 }, { "epoch": 0.11033587538536427, "step": 340, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.5033372619654983e-05, "weight_min": 0.0, "weight_std": 2.0749079340021126e-05 }, { "epoch": 0.11033587538536427, "step": 340, "weight_max": 3.255531191825867e-05, "weight_mean": 4.080681264895247e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.150552725448506e-05 }, { "epoch": 0.11033587538536427, "step": 340, "weight_max": 1.3363533071242273e-05, "weight_mean": 1.671527570579201e-06, "weight_min": 0.0, "weight_std": 4.724284281110158e-06 }, { "epoch": 0.11066039266590946, "step": 341, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236293873167597e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.480604987591505e-05 }, { "epoch": 0.11066039266590946, "step": 341, "weight_max": 3.0886087188264355e-05, "weight_mean": 9.29096404433949e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3578733160102274e-05 }, { "epoch": 0.11066039266590946, "step": 341, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3639326425618492e-05, "weight_min": 0.0, "weight_std": 1.8900207578553818e-05 }, { "epoch": 0.11066039266590946, "step": 341, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3004188986087684e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8113669284502976e-05 }, { "epoch": 0.11098490994645464, "step": 342, "weight_max": 3.967146767536178e-05, "weight_mean": 1.664218507357873e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.841209268604871e-05 }, { "epoch": 0.11098490994645464, "step": 342, "weight_max": 4.252122016623616e-05, "weight_mean": 2.463935197738465e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0569554180838168e-05 }, { "epoch": 0.11098490994645464, "step": 342, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.77058027475141e-05, "weight_min": 0.0, "weight_std": 2.0027620848850347e-05 }, { "epoch": 0.11098490994645464, "step": 342, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8309814549866132e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9723975128727034e-05 }, { "epoch": 0.11130942722699984, "step": 343, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.506202331976965e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.765502202033531e-05 }, { "epoch": 0.11130942722699984, "step": 343, "weight_max": 4.187939703115262e-05, "weight_mean": 2.5214734705514275e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.664604860707186e-05 }, { "epoch": 0.11130942722699984, "step": 343, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5073641407070681e-05, "weight_min": 0.0, "weight_std": 2.080921331071295e-05 }, { "epoch": 0.11130942722699984, "step": 343, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.69201028055977e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7971608031075448e-05 }, { "epoch": 0.11163394450754503, "step": 344, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.34123773832107e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4514484973915387e-05 }, { "epoch": 0.11163394450754503, "step": 344, "weight_max": 4.381950566312298e-05, "weight_mean": 1.3586284694611095e-05, "weight_min": 0.0, "weight_std": 1.909388265630696e-05 }, { "epoch": 0.11163394450754503, "step": 344, "weight_max": 4.252122016623616e-05, "weight_mean": 8.128602530632634e-06, "weight_min": 0.0, "weight_std": 1.59699993673712e-05 }, { "epoch": 0.11163394450754503, "step": 344, "weight_max": 3.36925259034615e-05, "weight_mean": 4.224646545480937e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.1906868166988716e-05 }, { "epoch": 0.11195846178809021, "step": 345, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0194467904511839e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8884202290792018e-05 }, { "epoch": 0.11195846178809021, "step": 345, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.386916654970264e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4502194062515628e-05 }, { "epoch": 0.11195846178809021, "step": 345, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9803077520919032e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.119342389050871e-05 }, { "epoch": 0.11195846178809021, "step": 345, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.547074680274818e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1369474779930897e-05 }, { "epoch": 0.11228297906863541, "step": 346, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2963576409674715e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8067710698232986e-05 }, { "epoch": 0.11228297906863541, "step": 346, "weight_max": 3.813097282545641e-05, "weight_mean": 1.1521279702719767e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.726123082335107e-05 }, { "epoch": 0.11228297906863541, "step": 346, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5231175893859472e-05, "weight_min": 0.0, "weight_std": 2.102817961713299e-05 }, { "epoch": 0.11228297906863541, "step": 346, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7809845303418115e-05, "weight_min": 0.0, "weight_std": 1.927531229739543e-05 }, { "epoch": 0.11260749634918059, "step": 347, "weight_max": 4.187939703115262e-05, "weight_mean": 2.1480875147972256e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.923427225847263e-05 }, { "epoch": 0.11260749634918059, "step": 347, "weight_max": 1.3301527657105794e-09, "weight_mean": 8.478741597706119e-10, "weight_min": 6.686325937366178e-10, "weight_std": 2.375902252715889e-10 }, { "epoch": 0.11260749634918059, "step": 347, "weight_max": 3.061215829802677e-05, "weight_mean": 7.319430096686119e-06, "weight_min": 0.0, "weight_std": 1.3570724149758462e-05 }, { "epoch": 0.11260749634918059, "step": 347, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0314289283996914e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9107350453850813e-05 }, { "epoch": 0.11293201362972578, "step": 348, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467046892386861e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2631992831302341e-05 }, { "epoch": 0.11293201362972578, "step": 348, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.0596933861961588e-05, "weight_min": 0.0, "weight_std": 1.566175888001453e-05 }, { "epoch": 0.11293201362972578, "step": 348, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5000890925875865e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.0723691704915836e-05 }, { "epoch": 0.11293201362972578, "step": 348, "weight_max": 3.602553260861896e-05, "weight_mean": 7.057766197249293e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3715415661863517e-05 }, { "epoch": 0.11325653091027098, "step": 349, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.2832551621831954e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9314466044306755e-05 }, { "epoch": 0.11325653091027098, "step": 349, "weight_max": 2.639517333591357e-05, "weight_mean": 3.300303660580539e-06, "weight_min": 5.771870759119224e-10, "weight_std": 9.33173669181997e-06 }, { "epoch": 0.11325653091027098, "step": 349, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.1988667867844924e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.861319651652593e-05 }, { "epoch": 0.11325653091027098, "step": 349, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153463335882407e-05, "weight_min": 0.0, "weight_std": 1.880294985312503e-05 }, { "epoch": 0.11358104819081616, "grad_norm": 49.36016082763672, "learning_rate": 9.493877551020409e-06, "loss": 0.3802, "step": 350 }, { "epoch": 0.11358104819081616, "step": 350, "weight_max": 3.9360926166409627e-05, "weight_mean": 6.121919341239845e-06, "weight_min": 0.0, "weight_std": 1.3845035937265493e-05 }, { "epoch": 0.11358104819081616, "step": 350, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8791102775139734e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0353527361294255e-05 }, { "epoch": 0.11358104819081616, "step": 350, "weight_max": 3.843663216684945e-05, "weight_mean": 9.01674047781853e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6742367733968422e-05 }, { "epoch": 0.11358104819081616, "step": 350, "weight_max": 4.252122016623616e-05, "weight_mean": 8.275763320853002e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6127874914673157e-05 }, { "epoch": 0.11390556547136135, "step": 351, "weight_max": 3.813097282545641e-05, "weight_mean": 7.803866537869908e-06, "weight_min": 0.0, "weight_std": 1.482753395976033e-05 }, { "epoch": 0.11390556547136135, "step": 351, "weight_max": 3.998322426923551e-05, "weight_mean": 9.539040547679178e-06, "weight_min": 0.0, "weight_std": 1.768795664247591e-05 }, { "epoch": 0.11390556547136135, "step": 351, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691914000432007e-06, "weight_min": 0.0, "weight_std": 1.3265892448544037e-05 }, { "epoch": 0.11390556547136135, "step": 351, "weight_max": 3.9360926166409627e-05, "weight_mean": 5.343866632756544e-06, "weight_min": 0.0, "weight_std": 1.3795928680337965e-05 }, { "epoch": 0.11423008275190653, "step": 352, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6241850971709937e-05, "weight_min": 0.0, "weight_std": 1.9768927813856862e-05 }, { "epoch": 0.11423008275190653, "step": 352, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4706328329339158e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6756939658080228e-05 }, { "epoch": 0.11423008275190653, "step": 352, "weight_max": 3.813097282545641e-05, "weight_mean": 4.770908162754495e-06, "weight_min": 0.0, "weight_std": 1.3479506378644146e-05 }, { "epoch": 0.11423008275190653, "step": 352, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.066468368866481e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0530766050796956e-05 }, { "epoch": 0.11455460003245173, "step": 353, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.924945980950724e-06, "weight_min": 0.0, "weight_std": 1.4337963875732385e-05 }, { "epoch": 0.11455460003245173, "step": 353, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.547118154121563e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.136911280103959e-05 }, { "epoch": 0.11455460003245173, "step": 353, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4556278074451257e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9887625967385247e-05 }, { "epoch": 0.11455460003245173, "step": 353, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4174505849950947e-05, "weight_min": 0.0, "weight_std": 1.9683184291352518e-05 }, { "epoch": 0.11487911731299692, "step": 354, "weight_max": 4.252122016623616e-05, "weight_mean": 1.591773980180733e-05, "weight_min": 0.0, "weight_std": 1.9002947738044895e-05 }, { "epoch": 0.11487911731299692, "step": 354, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7951111658476293e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9398074073251337e-05 }, { "epoch": 0.11487911731299692, "step": 354, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.759603017300833e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5227909898385406e-05 }, { "epoch": 0.11487911731299692, "step": 354, "weight_max": 3.662098970380612e-05, "weight_mean": 1.5105084457900375e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.7579492123331875e-05 }, { "epoch": 0.1152036345935421, "step": 355, "weight_max": 4.252122016623616e-05, "weight_mean": 1.746630368870683e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.94994227058487e-05 }, { "epoch": 0.1152036345935421, "step": 355, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.595379762875382e-05, "weight_min": 8.267048712262692e-10, "weight_std": 1.9617315047071315e-05 }, { "epoch": 0.1152036345935421, "step": 355, "weight_max": 3.632264997577295e-05, "weight_mean": 1.3075130482320674e-05, "weight_min": 0.0, "weight_std": 1.8069782527163625e-05 }, { "epoch": 0.1152036345935421, "step": 355, "weight_max": 3.36925259034615e-05, "weight_mean": 4.213155534671387e-06, "weight_min": 0.0, "weight_std": 1.1911464753211476e-05 }, { "epoch": 0.1155281518740873, "step": 356, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.615759725216776e-06, "weight_min": 0.0, "weight_std": 1.635866101423744e-05 }, { "epoch": 0.1155281518740873, "step": 356, "weight_max": 4.252122016623616e-05, "weight_mean": 1.89364673133241e-05, "weight_min": 0.0, "weight_std": 2.049386966973543e-05 }, { "epoch": 0.1155281518740873, "step": 356, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4850860679871403e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0549863620544784e-05 }, { "epoch": 0.1155281518740873, "step": 356, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390898751211353e-05, "weight_min": 0.0, "weight_std": 1.9239312678109854e-05 }, { "epoch": 0.11585266915463248, "step": 357, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.295755646424368e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8031852960120887e-05 }, { "epoch": 0.11585266915463248, "step": 357, "weight_max": 3.199401908204891e-05, "weight_mean": 7.792034011799842e-06, "weight_min": 0.0, "weight_std": 1.4433951037062798e-05 }, { "epoch": 0.11585266915463248, "step": 357, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930508682969958e-06, "weight_min": 0.0, "weight_std": 1.1114549124613404e-05 }, { "epoch": 0.11585266915463248, "step": 357, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921019353787415e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3915823728893884e-05 }, { "epoch": 0.11617718643517767, "step": 358, "weight_max": 4.252122016623616e-05, "weight_mean": 9.563373168930411e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7852658857009374e-05 }, { "epoch": 0.11617718643517767, "step": 358, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.811875886749476e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.644105759623926e-05 }, { "epoch": 0.11617718643517767, "step": 358, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882368557446171e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.380645790050039e-05 }, { "epoch": 0.11617718643517767, "step": 358, "weight_max": 2.4786623864514468e-09, "weight_mean": 9.051347449329228e-10, "weight_min": 0.0, "weight_std": 7.406194546533129e-10 }, { "epoch": 0.11650170371572287, "step": 359, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470512279425748e-05, "weight_min": 0.0, "weight_std": 1.938598506967537e-05 }, { "epoch": 0.11650170371572287, "step": 359, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3968960047350265e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9637363948277198e-05 }, { "epoch": 0.11650170371572287, "step": 359, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4575063687516376e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0143510482739657e-05 }, { "epoch": 0.11650170371572287, "step": 359, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070031081937486e-06, "weight_min": 0.0, "weight_std": 1.1509792784636375e-05 }, { "epoch": 0.11682622099626805, "grad_norm": 0.9254018664360046, "learning_rate": 9.473469387755104e-06, "loss": 0.2167, "step": 360 }, { "epoch": 0.11682622099626805, "step": 360, "weight_max": 3.967146767536178e-05, "weight_mean": 1.359476664219983e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8961709429277107e-05 }, { "epoch": 0.11682622099626805, "step": 360, "weight_max": 4.187939703115262e-05, "weight_mean": 5.27439033248811e-06, "weight_min": 0.0, "weight_std": 1.4791058674745727e-05 }, { "epoch": 0.11682622099626805, "step": 360, "weight_max": 3.998322426923551e-05, "weight_mean": 5.003956175642088e-06, "weight_min": 0.0, "weight_std": 1.4133763215795625e-05 }, { "epoch": 0.11682622099626805, "step": 360, "weight_max": 1.7626041071139298e-08, "weight_mean": 3.0621236657424333e-09, "weight_min": 6.29309993005478e-10, "weight_std": 5.904403632683852e-09 }, { "epoch": 0.11715073827681324, "step": 361, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0237932656309567e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8857759641832672e-05 }, { "epoch": 0.11715073827681324, "step": 361, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.016850526677445e-06, "weight_min": 0.0, "weight_std": 1.6921931091928855e-05 }, { "epoch": 0.11715073827681324, "step": 361, "weight_max": 4.252122016623616e-05, "weight_mean": 7.07785829945351e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5146771147556137e-05 }, { "epoch": 0.11715073827681324, "step": 361, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.0081073924084194e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.4550953892467078e-05 }, { "epoch": 0.11747525555735842, "step": 362, "weight_max": 3.9051607018336654e-05, "weight_mean": 5.980215064482763e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3711866813537199e-05 }, { "epoch": 0.11747525555735842, "step": 362, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471128007338848e-05, "weight_min": 0.0, "weight_std": 1.939012508955784e-05 }, { "epoch": 0.11747525555735842, "step": 362, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.85639780992642e-06, "weight_min": 1.076157385426768e-09, "weight_std": 1.5946856365189888e-05 }, { "epoch": 0.11747525555735842, "step": 362, "weight_max": 2.9469142646121327e-06, "weight_mean": 3.6885725762658694e-07, "weight_min": 0.0, "weight_std": 1.041692371472891e-06 }, { "epoch": 0.11779977283790362, "step": 363, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4704191926284693e-05, "weight_min": 0.0, "weight_std": 2.0383144146762788e-05 }, { "epoch": 0.11779977283790362, "step": 363, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2809086658526212e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.948554017872084e-05 }, { "epoch": 0.11779977283790362, "step": 363, "weight_max": 3.006794759130571e-05, "weight_mean": 5.3434814617503434e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0926416507572867e-05 }, { "epoch": 0.11779977283790362, "step": 363, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0005551050417125e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.1435771486721933e-05 }, { "epoch": 0.11812429011844881, "step": 364, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3085729480953887e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9328010239405558e-05 }, { "epoch": 0.11812429011844881, "step": 364, "weight_max": 3.632264997577295e-05, "weight_mean": 8.099346814560704e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.5141117728489917e-05 }, { "epoch": 0.11812429011844881, "step": 364, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3598226360045373e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8823826394509524e-05 }, { "epoch": 0.11812429011844881, "step": 364, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.28559431686881e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8335742424824275e-05 }, { "epoch": 0.118448807398994, "step": 365, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.686602425063029e-06, "weight_min": 0.0, "weight_std": 1.7934311472345144e-05 }, { "epoch": 0.118448807398994, "step": 365, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998777967557544e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4135852325125597e-05 }, { "epoch": 0.118448807398994, "step": 365, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1560923566285055e-06, "weight_min": 0.0, "weight_std": 1.4581095456378534e-05 }, { "epoch": 0.118448807398994, "step": 365, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235688604443567e-06, "weight_min": 0.0, "weight_std": 1.4806293620495126e-05 }, { "epoch": 0.11877332467953919, "step": 366, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.750032975513022e-06, "weight_min": 7.508524357824342e-10, "weight_std": 1.6201433027163148e-05 }, { "epoch": 0.11877332467953919, "step": 366, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9624611013568938e-05, "weight_min": 0.0, "weight_std": 2.1361902327043936e-05 }, { "epoch": 0.11877332467953919, "step": 366, "weight_max": 3.752332122530788e-05, "weight_mean": 1.202035400638124e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7096002920879982e-05 }, { "epoch": 0.11877332467953919, "step": 366, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397219410951948e-06, "weight_min": 0.0, "weight_std": 1.5261668522725813e-05 }, { "epoch": 0.11909784196008437, "step": 367, "weight_max": 5.153043769468013e-09, "weight_mean": 1.1923539933178517e-09, "weight_min": 0.0, "weight_std": 1.630845125610847e-09 }, { "epoch": 0.11909784196008437, "step": 367, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0161016436759382e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1605183064821176e-05 }, { "epoch": 0.11909784196008437, "step": 367, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.0512630979064852e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.505245927546639e-05 }, { "epoch": 0.11909784196008437, "step": 367, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.1903912309207954e-05, "weight_min": 0.0, "weight_std": 1.633735519135371e-05 }, { "epoch": 0.11942235924062956, "step": 368, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729012744064676e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3373417459661141e-05 }, { "epoch": 0.11942235924062956, "step": 368, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.0145413145655766e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0986366507713683e-05 }, { "epoch": 0.11942235924062956, "step": 368, "weight_max": 3.9051607018336654e-05, "weight_mean": 5.37601863470627e-06, "weight_min": 0.0, "weight_std": 1.3677124115929473e-05 }, { "epoch": 0.11942235924062956, "step": 368, "weight_max": 3.967146767536178e-05, "weight_mean": 5.0038943300023675e-06, "weight_min": 0.0, "weight_std": 1.4008202015247662e-05 }, { "epoch": 0.11974687652117476, "step": 369, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.306519698118791e-05, "weight_min": 0.0, "weight_std": 1.7610980648896657e-05 }, { "epoch": 0.11974687652117476, "step": 369, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8663598893908784e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.030746327363886e-05 }, { "epoch": 0.11974687652117476, "step": 369, "weight_max": 2.3459858500274322e-09, "weight_mean": 1.1746569272830243e-09, "weight_min": 5.771870759119224e-10, "weight_std": 7.608962904193106e-10 }, { "epoch": 0.11974687652117476, "step": 369, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5709460058133118e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.169886283809319e-05 }, { "epoch": 0.12007139380171994, "grad_norm": 4.586213111877441, "learning_rate": 9.453061224489797e-06, "loss": 0.2737, "step": 370 }, { "epoch": 0.12007139380171994, "step": 370, "weight_max": 3.199401908204891e-05, "weight_mean": 3.999767159257317e-06, "weight_min": 0.0, "weight_std": 1.1311385605949908e-05 }, { "epoch": 0.12007139380171994, "step": 370, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.7599064449314028e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.907911428133957e-05 }, { "epoch": 0.12007139380171994, "step": 370, "weight_max": 3.602553260861896e-05, "weight_mean": 1.4753588402527384e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7233203834621236e-05 }, { "epoch": 0.12007139380171994, "step": 370, "weight_max": 4.203939862890138e-09, "weight_mean": 1.2256243797637012e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.2129212079159402e-09 }, { "epoch": 0.12039591108226513, "step": 371, "weight_max": 3.998322426923551e-05, "weight_mean": 5.456300186779117e-06, "weight_min": 0.0, "weight_std": 1.4009773622092325e-05 }, { "epoch": 0.12039591108226513, "step": 371, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998500571673503e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4135962373984512e-05 }, { "epoch": 0.12039591108226513, "step": 371, "weight_max": 3.312147964606993e-05, "weight_mean": 7.178899068094324e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.346150747849606e-05 }, { "epoch": 0.12039591108226513, "step": 371, "weight_max": 3.782653220696375e-05, "weight_mean": 4.739140877063619e-06, "weight_min": 0.0, "weight_std": 1.3369356565817725e-05 }, { "epoch": 0.12072042836281031, "step": 372, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.51990770973498e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.101108111673966e-05 }, { "epoch": 0.12072042836281031, "step": 372, "weight_max": 3.662098970380612e-05, "weight_mean": 9.201953616866376e-06, "weight_min": 0.0, "weight_std": 1.6400121239712462e-05 }, { "epoch": 0.12072042836281031, "step": 372, "weight_max": 3.752332122530788e-05, "weight_mean": 4.698210432252381e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3263363143778406e-05 }, { "epoch": 0.12072042836281031, "step": 372, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3831000615027733e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8058844943880104e-05 }, { "epoch": 0.12104494564335551, "step": 373, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.959157068806235e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.844201142375823e-05 }, { "epoch": 0.12104494564335551, "step": 373, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5231420547934249e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1027977709309198e-05 }, { "epoch": 0.12104494564335551, "step": 373, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.69617576629389e-06, "weight_min": 0.0, "weight_std": 1.8000589989242144e-05 }, { "epoch": 0.12104494564335551, "step": 373, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.018879422394093e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1586800357908942e-05 }, { "epoch": 0.1213694629239007, "step": 374, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393218205223093e-06, "weight_min": 0.0, "weight_std": 1.2424182386894245e-05 }, { "epoch": 0.1213694629239007, "step": 374, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2167550266894978e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7395443137502298e-05 }, { "epoch": 0.1213694629239007, "step": 374, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.843331099546049e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8236723917652853e-05 }, { "epoch": 0.1213694629239007, "step": 374, "weight_max": 4.252122016623616e-05, "weight_mean": 1.592954868101515e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.1509049474843778e-05 }, { "epoch": 0.12169398020444588, "step": 375, "weight_max": 4.252122016623616e-05, "weight_mean": 1.716170118015725e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.047204543487169e-05 }, { "epoch": 0.12169398020444588, "step": 375, "weight_max": 3.9360926166409627e-05, "weight_mean": 7.91048205428524e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5216327483358327e-05 }, { "epoch": 0.12169398020444588, "step": 375, "weight_max": 2.846461757144425e-05, "weight_mean": 3.730697244463954e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0005601325246971e-05 }, { "epoch": 0.12169398020444588, "step": 375, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4457034922088496e-05, "weight_min": 0.0, "weight_std": 1.997197250602767e-05 }, { "epoch": 0.12201849748499108, "step": 376, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0475332601345144e-05, "weight_min": 0.0, "weight_std": 1.938753302965779e-05 }, { "epoch": 0.12201849748499108, "step": 376, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8467340851202607e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.006808244914282e-05 }, { "epoch": 0.12201849748499108, "step": 376, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959818397765048e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4025624295754824e-05 }, { "epoch": 0.12201849748499108, "step": 376, "weight_max": 3.967146767536178e-05, "weight_mean": 1.623914977244567e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7918739104061387e-05 }, { "epoch": 0.12234301476553626, "step": 377, "weight_max": 4.252122016623616e-05, "weight_mean": 2.02469509531511e-05, "weight_min": 0.0, "weight_std": 1.8819675460690632e-05 }, { "epoch": 0.12234301476553626, "step": 377, "weight_max": 3.722132169059478e-05, "weight_mean": 4.980696303391596e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3059310731478035e-05 }, { "epoch": 0.12234301476553626, "step": 377, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861642653646413e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0919525266217533e-05 }, { "epoch": 0.12234301476553626, "step": 377, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3986875273985788e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.957815584319178e-05 }, { "epoch": 0.12266753204608145, "step": 378, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8522543541621417e-05, "weight_min": 0.0, "weight_std": 2.0178356862743385e-05 }, { "epoch": 0.12266753204608145, "step": 378, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2422295185388066e-05, "weight_min": 0.0, "weight_std": 1.8462549633113667e-05 }, { "epoch": 0.12266753204608145, "step": 378, "weight_max": 4.187939703115262e-05, "weight_mean": 9.375631634611636e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7516102161607705e-05 }, { "epoch": 0.12266753204608145, "step": 378, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2250296751735732e-05, "weight_min": 0.0, "weight_std": 2.009180752793327e-05 }, { "epoch": 0.12299204932662665, "step": 379, "weight_max": 2.202445102739148e-05, "weight_mean": 2.7540145310922526e-06, "weight_min": 5.864964625068581e-10, "weight_std": 7.786432433931623e-06 }, { "epoch": 0.12299204932662665, "step": 379, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076900833955733e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4357703548739664e-05 }, { "epoch": 0.12299204932662665, "step": 379, "weight_max": 3.998322426923551e-05, "weight_mean": 5.0017592911899555e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4134649063635152e-05 }, { "epoch": 0.12299204932662665, "step": 379, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.262155274162069e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.769781010807492e-05 }, { "epoch": 0.12331656660717183, "grad_norm": 6.781548500061035, "learning_rate": 9.43265306122449e-06, "loss": 0.2392, "step": 380 }, { "epoch": 0.12331656660717183, "step": 380, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077230525785126e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.4357570762513205e-05 }, { "epoch": 0.12331656660717183, "step": 380, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793227051573922e-06, "weight_min": 0.0, "weight_std": 1.0726292202889454e-05 }, { "epoch": 0.12331656660717183, "step": 380, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006686352426186e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8574111891211942e-05 }, { "epoch": 0.12331656660717183, "step": 380, "weight_max": 3.843663216684945e-05, "weight_mean": 1.41499094752362e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.953869650606066e-05 }, { "epoch": 0.12364108388771702, "step": 381, "weight_max": 3.397988257347606e-05, "weight_mean": 4.249330231687054e-06, "weight_min": 0.0, "weight_std": 1.201295708597172e-05 }, { "epoch": 0.12364108388771702, "step": 381, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233321518171579e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.895398600026965e-05 }, { "epoch": 0.12364108388771702, "step": 381, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.953265928023029e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7911397662828676e-05 }, { "epoch": 0.12364108388771702, "step": 381, "weight_max": 3.22740561387036e-05, "weight_mean": 7.981881935847923e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4491202819044702e-05 }, { "epoch": 0.1239656011682622, "step": 382, "weight_max": 4.447596802492626e-05, "weight_mean": 1.5470932339667343e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0376020984258503e-05 }, { "epoch": 0.1239656011682622, "step": 382, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.582505865371786e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.02441297005862e-05 }, { "epoch": 0.1239656011682622, "step": 382, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9107532352791168e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0733365090563893e-05 }, { "epoch": 0.1239656011682622, "step": 382, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767914106196258e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.3480712368618697e-05 }, { "epoch": 0.1242901184488074, "step": 383, "weight_max": 3.36925259034615e-05, "weight_mean": 1.1802327207988128e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.634357795410324e-05 }, { "epoch": 0.1242901184488074, "step": 383, "weight_max": 4.252122016623616e-05, "weight_mean": 1.055064785759896e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9535347746568732e-05 }, { "epoch": 0.1242901184488074, "step": 383, "weight_max": 3.998322426923551e-05, "weight_mean": 1.778922523953952e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9163780962117016e-05 }, { "epoch": 0.1242901184488074, "step": 383, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467583039513556e-06, "weight_min": 0.0, "weight_std": 1.2631776371563319e-05 }, { "epoch": 0.1246146357293526, "step": 384, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998527856514556e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4135952369542792e-05 }, { "epoch": 0.1246146357293526, "step": 384, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4102777388179675e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.013210360019002e-05 }, { "epoch": 0.1246146357293526, "step": 384, "weight_max": 3.843663216684945e-05, "weight_mean": 4.80530161439674e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3589109585154802e-05 }, { "epoch": 0.1246146357293526, "step": 384, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6347219570889138e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8035067114396952e-05 }, { "epoch": 0.12493915300989777, "step": 385, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0158566510654055e-05, "weight_min": 0.0, "weight_std": 1.883591539808549e-05 }, { "epoch": 0.12493915300989777, "step": 385, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3895952179154847e-05, "weight_min": 0.0, "weight_std": 1.9185688870493323e-05 }, { "epoch": 0.12493915300989777, "step": 385, "weight_max": 3.813097282545641e-05, "weight_mean": 1.1883689694514032e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.6671176126692444e-05 }, { "epoch": 0.12493915300989777, "step": 385, "weight_max": 3.602553260861896e-05, "weight_mean": 1.293231707677478e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7885085981106386e-05 }, { "epoch": 0.12526367029044297, "step": 386, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.378899014525814e-05, "weight_min": 0.0, "weight_std": 1.9264620277681388e-05 }, { "epoch": 0.12526367029044297, "step": 386, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6600590242887847e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.822985359467566e-05 }, { "epoch": 0.12526367029044297, "step": 386, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235973276285222e-06, "weight_min": 0.0, "weight_std": 1.4806179024162702e-05 }, { "epoch": 0.12526367029044297, "step": 386, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.438648541807197e-05, "weight_min": 0.0, "weight_std": 1.990157034015283e-05 }, { "epoch": 0.12558818757098816, "step": 387, "weight_max": 3.6920544516760856e-05, "weight_mean": 7.979276233527344e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5012691619631369e-05 }, { "epoch": 0.12558818757098816, "step": 387, "weight_max": 4.252122016623616e-05, "weight_mean": 9.635468813939951e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.796659125830047e-05 }, { "epoch": 0.12558818757098816, "step": 387, "weight_max": 3.632264997577295e-05, "weight_mean": 7.411288152070483e-06, "weight_min": 0.0, "weight_std": 1.4178205674397759e-05 }, { "epoch": 0.12558818757098816, "step": 387, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8626291421242058e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0132212739554234e-05 }, { "epoch": 0.12591270485153336, "step": 388, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729481133836089e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3373228284763172e-05 }, { "epoch": 0.12591270485153336, "step": 388, "weight_max": 2.407077559496429e-08, "weight_mean": 3.5839116119262826e-09, "weight_min": 5.679522963042416e-10, "weight_std": 8.278389529436936e-09 }, { "epoch": 0.12591270485153336, "step": 388, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4461214959737845e-05, "weight_min": 0.0, "weight_std": 1.999564301513601e-05 }, { "epoch": 0.12591270485153336, "step": 388, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3010749171371572e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8161057596444152e-05 }, { "epoch": 0.12623722213207852, "step": 389, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0760752047644928e-05, "weight_min": 8.378387983398738e-10, "weight_std": 1.9606157366069965e-05 }, { "epoch": 0.12623722213207852, "step": 389, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.6307942132698372e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8114376871380955e-05 }, { "epoch": 0.12623722213207852, "step": 389, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0394524906587321e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9264172806288116e-05 }, { "epoch": 0.12623722213207852, "step": 389, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.547931788081769e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.012959521380253e-05 }, { "epoch": 0.12656173941262372, "grad_norm": 0.703646183013916, "learning_rate": 9.412244897959183e-06, "loss": 0.236, "step": 390 }, { "epoch": 0.12656173941262372, "step": 390, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7926926375366747e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.942551534739323e-05 }, { "epoch": 0.12656173941262372, "step": 390, "weight_max": 4.252122016623616e-05, "weight_mean": 9.14216616365593e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7223426766577177e-05 }, { "epoch": 0.12656173941262372, "step": 390, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861592176690465e-06, "weight_min": 0.0, "weight_std": 1.0919545275100973e-05 }, { "epoch": 0.12656173941262372, "step": 390, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1686011930578388e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6817089999676682e-05 }, { "epoch": 0.12688625669316891, "step": 391, "weight_max": 3.813097282545641e-05, "weight_mean": 8.226951649703551e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5485753465327434e-05 }, { "epoch": 0.12688625669316891, "step": 391, "weight_max": 4.187939703115262e-05, "weight_mean": 1.894204251584597e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0368859622976743e-05 }, { "epoch": 0.12688625669316891, "step": 391, "weight_max": 4.187939703115262e-05, "weight_mean": 9.555121323501226e-06, "weight_min": 0.0, "weight_std": 1.779949525371194e-05 }, { "epoch": 0.12688625669316891, "step": 391, "weight_max": 2.538973967602942e-05, "weight_mean": 3.455985279288143e-06, "weight_min": 5.679522963042416e-10, "weight_std": 8.897494808479678e-06 }, { "epoch": 0.1272107739737141, "step": 392, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728894055006094e-06, "weight_min": 0.0, "weight_std": 1.3373465662880335e-05 }, { "epoch": 0.1272107739737141, "step": 392, "weight_max": 5.2767168199352454e-06, "weight_mean": 7.312721095331653e-07, "weight_min": 5.864964625068581e-10, "weight_std": 1.846519808168523e-06 }, { "epoch": 0.1272107739737141, "step": 392, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2121916370233521e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7970774933928624e-05 }, { "epoch": 0.1272107739737141, "step": 392, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8588765669846907e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0230247173458338e-05 }, { "epoch": 0.1275352912542593, "step": 393, "weight_max": 4.252122016623616e-05, "weight_mean": 1.81147079274524e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9530240024323575e-05 }, { "epoch": 0.1275352912542593, "step": 393, "weight_max": 7.67519914290915e-09, "weight_mean": 1.946926619922351e-09, "weight_min": 5.864964625068581e-10, "weight_std": 2.3776707269718145e-09 }, { "epoch": 0.1275352912542593, "step": 393, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.400487225415418e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5260351574397646e-05 }, { "epoch": 0.1275352912542593, "step": 393, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4923567505320534e-05, "weight_min": 0.0, "weight_std": 2.0628824131563306e-05 }, { "epoch": 0.12785980853480447, "step": 394, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5158384485403076e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0952385966666043e-05 }, { "epoch": 0.12785980853480447, "step": 394, "weight_max": 3.662098970380612e-05, "weight_mean": 1.0421283150208183e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.604567478352692e-05 }, { "epoch": 0.12785980853480447, "step": 394, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.551083187223412e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1426112652989104e-05 }, { "epoch": 0.12785980853480447, "step": 394, "weight_max": 3.662098970380612e-05, "weight_mean": 4.5783772293361835e-06, "weight_min": 0.0, "weight_std": 1.2947170944244135e-05 }, { "epoch": 0.12818432581534966, "step": 395, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5409633457893506e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2841740499425214e-05 }, { "epoch": 0.12818432581534966, "step": 395, "weight_max": 4.252122016623616e-05, "weight_mean": 7.101011306076543e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.515928397566313e-05 }, { "epoch": 0.12818432581534966, "step": 395, "weight_max": 4.187939703115262e-05, "weight_mean": 9.963971024262719e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8479660866432823e-05 }, { "epoch": 0.12818432581534966, "step": 395, "weight_max": 4.252122016623616e-05, "weight_mean": 1.89344154932769e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0487448637140915e-05 }, { "epoch": 0.12850884309589486, "step": 396, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3550866242439952e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9035071090911515e-05 }, { "epoch": 0.12850884309589486, "step": 396, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0531494808674324e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.935358341143001e-05 }, { "epoch": 0.12850884309589486, "step": 396, "weight_max": 3.967146767536178e-05, "weight_mean": 9.612196663510986e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7808895790949464e-05 }, { "epoch": 0.12850884309589486, "step": 396, "weight_max": 4.187939703115262e-05, "weight_mean": 5.8739615269587375e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4657747669843957e-05 }, { "epoch": 0.12883336037644005, "step": 397, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653281848732149e-06, "weight_min": 0.0, "weight_std": 1.3159475201973692e-05 }, { "epoch": 0.12883336037644005, "step": 397, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.916314072848763e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6607387806288898e-05 }, { "epoch": 0.12883336037644005, "step": 397, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315805992722744e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5033257113827858e-05 }, { "epoch": 0.12883336037644005, "step": 397, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430728040460963e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2527601938927546e-05 }, { "epoch": 0.12915787765698525, "step": 398, "weight_max": 3.632264997577295e-05, "weight_mean": 8.716546290088445e-06, "weight_min": 0.0, "weight_std": 1.6157689969986677e-05 }, { "epoch": 0.12915787765698525, "step": 398, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9985269470198546e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4135952369542792e-05 }, { "epoch": 0.12915787765698525, "step": 398, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5375821021734737e-05, "weight_min": 9.062063877074422e-10, "weight_std": 2.1206496967351995e-05 }, { "epoch": 0.12915787765698525, "step": 398, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6628238881821744e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8544202248449437e-05 }, { "epoch": 0.12948239493753042, "step": 399, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4324399671750143e-05, "weight_min": 0.0, "weight_std": 1.987979885598179e-05 }, { "epoch": 0.12948239493753042, "step": 399, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.2960215826751664e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7887912690639496e-05 }, { "epoch": 0.12948239493753042, "step": 399, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.9474873523868155e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.511819391453173e-05 }, { "epoch": 0.12948239493753042, "step": 399, "weight_max": 3.998322426923551e-05, "weight_mean": 9.996407243306749e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8508266293792985e-05 }, { "epoch": 0.1298069122180756, "grad_norm": 0.2726770043373108, "learning_rate": 9.391836734693878e-06, "loss": 0.1151, "step": 400 }, { "epoch": 0.1298069122180756, "eval_accuracy": 0.494, "eval_auc": 0.44802351538621565, "eval_loss": 2.9344186782836914, "eval_runtime": 155.6489, "eval_samples_per_second": 12.849, "eval_steps_per_second": 1.606, "step": 400 }, { "epoch": 0.1298069122180756, "step": 400, "weight_max": 4.252122016623616e-05, "weight_mean": 5.434992999653332e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.498879191785818e-05 }, { "epoch": 0.1298069122180756, "step": 400, "weight_max": 2.2718091852169664e-09, "weight_mean": 9.208604989652258e-10, "weight_min": 6.053385570581327e-10, "weight_std": 5.542986070139477e-10 }, { "epoch": 0.1298069122180756, "step": 400, "weight_max": 1.6735920427279893e-09, "weight_mean": 1.0268784711442436e-09, "weight_min": 6.100955851628953e-10, "weight_std": 4.702460643102313e-10 }, { "epoch": 0.1298069122180756, "step": 400, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2366794989211485e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.4805895261815749e-05 }, { "epoch": 0.1301314294986208, "step": 401, "weight_max": 3.5141503758495674e-05, "weight_mean": 9.685518307378516e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5492976672248915e-05 }, { "epoch": 0.1301314294986208, "step": 401, "weight_max": 3.813097282545641e-05, "weight_mean": 9.050383596331812e-06, "weight_min": 0.0, "weight_std": 1.6788697394076735e-05 }, { "epoch": 0.1301314294986208, "step": 401, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4116487363935448e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.951149351953063e-05 }, { "epoch": 0.1301314294986208, "step": 401, "weight_max": 3.998322426923551e-05, "weight_mean": 9.841398423304781e-06, "weight_min": 0.0, "weight_std": 1.8224365703645162e-05 }, { "epoch": 0.130455946779166, "step": 402, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0392512194812298e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9238317690906115e-05 }, { "epoch": 0.130455946779166, "step": 402, "weight_max": 2.2256894226302393e-05, "weight_mean": 2.787076027743751e-06, "weight_min": 0.0, "weight_std": 7.867002750572283e-06 }, { "epoch": 0.130455946779166, "step": 402, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541202997643268e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2841645002481528e-05 }, { "epoch": 0.130455946779166, "step": 402, "weight_max": 2.820166446326766e-05, "weight_mean": 6.762013981642667e-06, "weight_min": 0.0, "weight_std": 1.2534684174170252e-05 }, { "epoch": 0.1307804640597112, "step": 403, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5355162759078667e-05, "weight_min": 0.0, "weight_std": 2.1223519070190378e-05 }, { "epoch": 0.1307804640597112, "step": 403, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8166461813962087e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9594164768932387e-05 }, { "epoch": 0.1307804640597112, "step": 403, "weight_max": 4.381950566312298e-05, "weight_mean": 5.478067123476649e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5492279999307357e-05 }, { "epoch": 0.1307804640597112, "step": 403, "weight_max": 3.22740561387036e-05, "weight_mean": 1.0319457942387089e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.4649047443526797e-05 }, { "epoch": 0.13110498134025636, "step": 404, "weight_max": 2.626842672270868e-07, "weight_mean": 3.367942369436605e-08, "weight_min": 5.958802340444436e-10, "weight_std": 9.253282229337856e-08 }, { "epoch": 0.13110498134025636, "step": 404, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3171625040704384e-06, "weight_min": 0.0, "weight_std": 1.503270959801739e-05 }, { "epoch": 0.13110498134025636, "step": 404, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.237149813794531e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6200134268729016e-05 }, { "epoch": 0.13110498134025636, "step": 404, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7806651158025488e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0221026716171764e-05 }, { "epoch": 0.13142949862080155, "step": 405, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9887238522642292e-05, "weight_min": 0.0, "weight_std": 2.1310044758138247e-05 }, { "epoch": 0.13142949862080155, "step": 405, "weight_max": 2.7549629244560947e-09, "weight_mean": 8.894217029542517e-10, "weight_min": 0.0, "weight_std": 8.150350394586781e-10 }, { "epoch": 0.13142949862080155, "step": 405, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4129033843346406e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9572218661778606e-05 }, { "epoch": 0.13142949862080155, "step": 405, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.936929018294904e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8488854038878344e-05 }, { "epoch": 0.13175401590134675, "step": 406, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4065533832763322e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9544531824067235e-05 }, { "epoch": 0.13175401590134675, "step": 406, "weight_max": 4.187939703115262e-05, "weight_mean": 9.6648363978602e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.797715958673507e-05 }, { "epoch": 0.13175401590134675, "step": 406, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4126848074956797e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0095943909836933e-05 }, { "epoch": 0.13175401590134675, "step": 406, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.5390885891974904e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8578555682324804e-05 }, { "epoch": 0.13207853318189194, "step": 407, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392051080998499e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9247638192609884e-05 }, { "epoch": 0.13207853318189194, "step": 407, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9536348190740682e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.09276895475341e-05 }, { "epoch": 0.13207853318189194, "step": 407, "weight_max": 3.5141503758495674e-05, "weight_mean": 7.788537004671525e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4576952708011959e-05 }, { "epoch": 0.13207853318189194, "step": 407, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4333757462736685e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.008187584578991e-05 }, { "epoch": 0.13240305046243714, "step": 408, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5201942005660385e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.081256388919428e-05 }, { "epoch": 0.13240305046243714, "step": 408, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4899995221639983e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.065818989649415e-05 }, { "epoch": 0.13240305046243714, "step": 408, "weight_max": 3.998322426923551e-05, "weight_mean": 9.803439752431586e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8154783901991323e-05 }, { "epoch": 0.13240305046243714, "step": 408, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.958245755115058e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8442573491483927e-05 }, { "epoch": 0.1327275677429823, "step": 409, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3815697457175702e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.909976708702743e-05 }, { "epoch": 0.1327275677429823, "step": 409, "weight_max": 3.602553260861896e-05, "weight_mean": 8.028801858017687e-06, "weight_min": 0.0, "weight_std": 1.5011803952802438e-05 }, { "epoch": 0.1327275677429823, "step": 409, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0865718650165945e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.954894651134964e-05 }, { "epoch": 0.1327275677429823, "step": 409, "weight_max": 5.130903559802391e-07, "weight_mean": 6.480426151256324e-08, "weight_min": 6.053385570581327e-10, "weight_std": 1.8113512112449826e-07 }, { "epoch": 0.1330520850235275, "grad_norm": 11.879359245300293, "learning_rate": 9.371428571428572e-06, "loss": 0.3124, "step": 410 }, { "epoch": 0.1330520850235275, "step": 410, "weight_max": 3.967146767536178e-05, "weight_mean": 9.537100595480297e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.767674802977126e-05 }, { "epoch": 0.1330520850235275, "step": 410, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.6071540863486e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.593490014784038e-05 }, { "epoch": 0.1330520850235275, "step": 410, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.890674507711083e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6952695659711026e-05 }, { "epoch": 0.1330520850235275, "step": 410, "weight_max": 9.15524742595153e-06, "weight_mean": 1.145429791904462e-06, "weight_min": 0.0, "weight_std": 3.2364550861530006e-06 }, { "epoch": 0.1333766023040727, "step": 411, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.400841928349109e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5260206055245362e-05 }, { "epoch": 0.1333766023040727, "step": 411, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0551043487794232e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9535104001988657e-05 }, { "epoch": 0.1333766023040727, "step": 411, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2375645383144729e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7336218661512248e-05 }, { "epoch": 0.1333766023040727, "step": 411, "weight_max": 4.187939703115262e-05, "weight_mean": 1.489183705416508e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0476187273743562e-05 }, { "epoch": 0.1337011195846179, "step": 412, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.128890956664691e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.4696432117489167e-05 }, { "epoch": 0.1337011195846179, "step": 412, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3847846275893971e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9119090211461298e-05 }, { "epoch": 0.1337011195846179, "step": 412, "weight_max": 4.187939703115262e-05, "weight_mean": 1.894357592391316e-05, "weight_min": 0.0, "weight_std": 2.034586577792652e-05 }, { "epoch": 0.1337011195846179, "step": 412, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3676792150363326e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.894815795822069e-05 }, { "epoch": 0.13402563686516308, "step": 413, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393586550577311e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2424035048752557e-05 }, { "epoch": 0.13402563686516308, "step": 413, "weight_max": 3.998322426923551e-05, "weight_mean": 8.689539754414e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.63293479999993e-05 }, { "epoch": 0.13402563686516308, "step": 413, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4624753021053039e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0262574253138155e-05 }, { "epoch": 0.13402563686516308, "step": 413, "weight_max": 3.602553260861896e-05, "weight_mean": 4.503819582168944e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2736695680359844e-05 }, { "epoch": 0.13435015414570825, "step": 414, "weight_max": 4.187939703115262e-05, "weight_mean": 1.485232132836245e-05, "weight_min": 0.0, "weight_std": 2.0555475202854723e-05 }, { "epoch": 0.13435015414570825, "step": 414, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0201221812167205e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8929660654976033e-05 }, { "epoch": 0.13435015414570825, "step": 414, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.22482014275738e-06, "weight_min": 0.0, "weight_std": 1.4303896023193374e-05 }, { "epoch": 0.13435015414570825, "step": 414, "weight_max": 2.0739674422287635e-09, "weight_mean": 1.0592865473668667e-09, "weight_min": 5.958802340444436e-10, "weight_std": 6.027212617887301e-10 }, { "epoch": 0.13467467142625345, "step": 415, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0781995772267692e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.2222744519240223e-05 }, { "epoch": 0.13467467142625345, "step": 415, "weight_max": 1.7219337678398006e-05, "weight_mean": 2.1541006844927324e-06, "weight_min": 6.100955851628953e-10, "weight_std": 6.087275323807262e-06 }, { "epoch": 0.13467467142625345, "step": 415, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4611156984756235e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0347481040516868e-05 }, { "epoch": 0.13467467142625345, "step": 415, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.958353646448813e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7036465578712523e-05 }, { "epoch": 0.13499918870679864, "step": 416, "weight_max": 4.279248599914354e-08, "weight_mean": 5.969678085193664e-09, "weight_min": 5.864964625068581e-10, "weight_std": 1.4879850240845371e-08 }, { "epoch": 0.13499918870679864, "step": 416, "weight_max": 5.010648251868588e-08, "weight_mean": 6.915239936233775e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.7453235301445602e-08 }, { "epoch": 0.13499918870679864, "step": 416, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2737646102323197e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.863878787844442e-05 }, { "epoch": 0.13499918870679864, "step": 416, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4968501091061626e-05, "weight_min": 0.0, "weight_std": 2.071593189612031e-05 }, { "epoch": 0.13532370598734383, "step": 417, "weight_max": 3.813097282545641e-05, "weight_mean": 1.4523081517836545e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.6662303096381947e-05 }, { "epoch": 0.13532370598734383, "step": 417, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6995061741909012e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.875734596978873e-05 }, { "epoch": 0.13532370598734383, "step": 417, "weight_max": 2.919402941614635e-09, "weight_mean": 1.1014646972284936e-09, "weight_min": 0.0, "weight_std": 8.60491899956628e-10 }, { "epoch": 0.13532370598734383, "step": 417, "weight_max": 3.199401908204891e-05, "weight_mean": 7.492231816286221e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3913965631218161e-05 }, { "epoch": 0.13564822326788903, "step": 418, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.309932667936664e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8209342670161277e-05 }, { "epoch": 0.13564822326788903, "step": 418, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.511739901616238e-05, "weight_min": 0.0, "weight_std": 2.0769433831446804e-05 }, { "epoch": 0.13564822326788903, "step": 418, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2748465451295488e-05, "weight_min": 0.0, "weight_std": 1.7780508642317727e-05 }, { "epoch": 0.13564822326788903, "step": 418, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3785677765554283e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8634327716426924e-05 }, { "epoch": 0.1359727405484342, "step": 419, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3708318622084334e-05, "weight_min": 0.0, "weight_std": 2.020571264438331e-05 }, { "epoch": 0.1359727405484342, "step": 419, "weight_max": 2.1780790575576248e-06, "weight_mean": 2.949088298009883e-07, "weight_min": 5.864964625068581e-10, "weight_std": 7.633900622749934e-07 }, { "epoch": 0.1359727405484342, "step": 419, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691589765570825e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3266021596791688e-05 }, { "epoch": 0.1359727405484342, "step": 419, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4547448699886445e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8277794879395515e-05 }, { "epoch": 0.1362972578289794, "grad_norm": 1.1475224494934082, "learning_rate": 9.351020408163267e-06, "loss": 0.303, "step": 420 }, { "epoch": 0.1362972578289794, "step": 420, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0075996215164196e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8662482034415007e-05 }, { "epoch": 0.1362972578289794, "step": 420, "weight_max": 3.813097282545641e-05, "weight_mean": 1.0802641554619186e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.722722481645178e-05 }, { "epoch": 0.1362972578289794, "step": 420, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4038830158824567e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9396384232095443e-05 }, { "epoch": 0.1362972578289794, "step": 420, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8707532944972627e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.008665251196362e-05 }, { "epoch": 0.13662177510952458, "step": 421, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8998120140167885e-05, "weight_min": 0.0, "weight_std": 2.0505993234110065e-05 }, { "epoch": 0.13662177510952458, "step": 421, "weight_max": 4.252122016623616e-05, "weight_mean": 9.68061613093596e-06, "weight_min": 0.0, "weight_std": 1.8019089111476205e-05 }, { "epoch": 0.13662177510952458, "step": 421, "weight_max": 1.4499302203319075e-08, "weight_mean": 3.970793915897275e-09, "weight_min": 6.053385570581327e-10, "weight_std": 6.013752162914443e-09 }, { "epoch": 0.13662177510952458, "step": 421, "weight_max": 3.116124207735993e-05, "weight_mean": 3.895791451213881e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1016906682925764e-05 }, { "epoch": 0.13694629239006978, "step": 422, "weight_max": 3.998322426923551e-05, "weight_mean": 7.213948265416548e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4605368960474152e-05 }, { "epoch": 0.13694629239006978, "step": 422, "weight_max": 7.508524357824342e-10, "weight_mean": 6.230261861972508e-10, "weight_min": 5.771870759119224e-10, "weight_std": 5.404417885435109e-11 }, { "epoch": 0.13694629239006978, "step": 422, "weight_max": 3.4268457966391e-05, "weight_mean": 9.025740837387275e-06, "weight_min": 0.0, "weight_std": 1.5551197066088207e-05 }, { "epoch": 0.13694629239006978, "step": 422, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920705123367952e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.391595014865743e-05 }, { "epoch": 0.13727080967061497, "step": 423, "weight_max": 1.783980223990511e-05, "weight_mean": 2.230747440989944e-06, "weight_min": 5.864964625068581e-10, "weight_std": 6.307010153250303e-06 }, { "epoch": 0.13727080967061497, "step": 423, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.823646238743095e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0714166819525417e-05 }, { "epoch": 0.13727080967061497, "step": 423, "weight_max": 1.2354433920336305e-08, "weight_mean": 2.3387940473185154e-09, "weight_min": 6.196655966128617e-10, "weight_std": 4.078675175378521e-09 }, { "epoch": 0.13727080967061497, "step": 423, "weight_max": 3.5141503758495674e-05, "weight_mean": 7.100447419361444e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3320728612598032e-05 }, { "epoch": 0.13759532695116014, "step": 424, "weight_max": 9.295914038531805e-10, "weight_mean": 7.184094985568379e-10, "weight_min": 6.196655966128617e-10, "weight_std": 1.1978806835344358e-10 }, { "epoch": 0.13759532695116014, "step": 424, "weight_max": 4.252122016623616e-05, "weight_mean": 9.563355888531078e-06, "weight_min": 0.0, "weight_std": 1.785266795195639e-05 }, { "epoch": 0.13759532695116014, "step": 424, "weight_max": 3.843663216684945e-05, "weight_mean": 5.287063686409965e-06, "weight_min": 0.0, "weight_std": 1.3462215974868741e-05 }, { "epoch": 0.13759532695116014, "step": 424, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5093786714714952e-05, "weight_min": 7.721521200210191e-10, "weight_std": 2.0884799596387893e-05 }, { "epoch": 0.13791984423170534, "step": 425, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3098553608870134e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.83758711500559e-05 }, { "epoch": 0.13791984423170534, "step": 425, "weight_max": 3.662098970380612e-05, "weight_mean": 1.0804000339703634e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.6571009837207384e-05 }, { "epoch": 0.13791984423170534, "step": 425, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0711852155509405e-05, "weight_min": 0.0, "weight_std": 1.983349648071453e-05 }, { "epoch": 0.13791984423170534, "step": 425, "weight_max": 3.22740561387036e-05, "weight_mean": 7.659148650418501e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4207526874088217e-05 }, { "epoch": 0.13824436151225053, "step": 426, "weight_max": 3.664386440505041e-06, "weight_mean": 4.5871001930208877e-07, "weight_min": 0.0, "weight_std": 1.295289052904991e-06 }, { "epoch": 0.13824436151225053, "step": 426, "weight_max": 4.252122016623616e-05, "weight_mean": 9.968702215701342e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8510540030547418e-05 }, { "epoch": 0.13824436151225053, "step": 426, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4392129489569925e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.0450373995117843e-05 }, { "epoch": 0.13824436151225053, "step": 426, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1840818842756562e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.90828232007334e-05 }, { "epoch": 0.13856887879279572, "step": 427, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8628239558893256e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9985971448477358e-05 }, { "epoch": 0.13856887879279572, "step": 427, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3676919479621574e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9128554413327947e-05 }, { "epoch": 0.13856887879279572, "step": 427, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4969653420848772e-05, "weight_min": 0.0, "weight_std": 2.072139068332035e-05 }, { "epoch": 0.13856887879279572, "step": 427, "weight_max": 1.763176260283217e-05, "weight_mean": 2.204881411671522e-06, "weight_min": 0.0, "weight_std": 6.233401109057013e-06 }, { "epoch": 0.13889339607334092, "step": 428, "weight_max": 3.752332122530788e-05, "weight_mean": 6.817695066274609e-06, "weight_min": 0.0, "weight_std": 1.3761543414148036e-05 }, { "epoch": 0.13889339607334092, "step": 428, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5625369744375348e-05, "weight_min": 9.772551656794803e-10, "weight_std": 2.1246398318908177e-05 }, { "epoch": 0.13889339607334092, "step": 428, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3987841157359071e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9429400708759204e-05 }, { "epoch": 0.13889339607334092, "step": 428, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2353934734128416e-06, "weight_min": 0.0, "weight_std": 1.4806412764301058e-05 }, { "epoch": 0.13921791335388609, "step": 429, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8532380636315793e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.021956788667012e-05 }, { "epoch": 0.13921791335388609, "step": 429, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8570190150057897e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.999982850975357e-05 }, { "epoch": 0.13921791335388609, "step": 429, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1886354513990227e-05, "weight_min": 0.0, "weight_std": 1.7793758161133155e-05 }, { "epoch": 0.13921791335388609, "step": 429, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.717072316561826e-06, "weight_min": 0.0, "weight_std": 1.8135877326130867e-05 }, { "epoch": 0.13954243063443128, "grad_norm": 2.608097791671753, "learning_rate": 9.33061224489796e-06, "loss": 0.2506, "step": 430 }, { "epoch": 0.13954243063443128, "step": 430, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.386670742766e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6356994819943793e-05 }, { "epoch": 0.13954243063443128, "step": 430, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3122862330637872e-05, "weight_min": 0.0, "weight_std": 1.9353494280949235e-05 }, { "epoch": 0.13954243063443128, "step": 430, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.110057372803567e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4344605915539432e-05 }, { "epoch": 0.13954243063443128, "step": 430, "weight_max": 3.967146767536178e-05, "weight_mean": 1.498402161814738e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8430169802741148e-05 }, { "epoch": 0.13986694791497647, "step": 431, "weight_max": 4.252122016623616e-05, "weight_mean": 6.126922926341649e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4878720321576111e-05 }, { "epoch": 0.13986694791497647, "step": 431, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793122914430569e-06, "weight_min": 0.0, "weight_std": 1.0726333130151033e-05 }, { "epoch": 0.13986694791497647, "step": 431, "weight_max": 1.247039249641091e-09, "weight_mean": 6.873769331505741e-10, "weight_min": 0.0, "weight_std": 4.555880117607103e-10 }, { "epoch": 0.13986694791497647, "step": 431, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0924682101176586e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6997639249893837e-05 }, { "epoch": 0.14019146519552167, "step": 432, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467683083930751e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.263173544430174e-05 }, { "epoch": 0.14019146519552167, "step": 432, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0001822374761105e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8545555576565675e-05 }, { "epoch": 0.14019146519552167, "step": 432, "weight_max": 3.632264997577295e-05, "weight_mean": 1.3089316780678928e-05, "weight_min": 0.0, "weight_std": 1.6065274394350126e-05 }, { "epoch": 0.14019146519552167, "step": 432, "weight_max": 4.252122016623616e-05, "weight_mean": 9.819637853070162e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.826194602472242e-05 }, { "epoch": 0.14051598247606686, "step": 433, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920741048408672e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.39159346872475e-05 }, { "epoch": 0.14051598247606686, "step": 433, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5390580301755108e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1253876184346154e-05 }, { "epoch": 0.14051598247606686, "step": 433, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.808482900552917e-06, "weight_min": 0.0, "weight_std": 1.8192293282481842e-05 }, { "epoch": 0.14051598247606686, "step": 433, "weight_max": 3.662098970380612e-05, "weight_mean": 6.0098341236880515e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3001510524190962e-05 }, { "epoch": 0.14084049975661203, "step": 434, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4296105291577987e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9879591491189785e-05 }, { "epoch": 0.14084049975661203, "step": 434, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.612179383111652e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7801005014916882e-05 }, { "epoch": 0.14084049975661203, "step": 434, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467235612537479e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2631916433747392e-05 }, { "epoch": 0.14084049975661203, "step": 434, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105604602955282e-06, "weight_min": 0.0, "weight_std": 1.1609553439484444e-05 }, { "epoch": 0.14116501703715723, "step": 435, "weight_max": 3.813097282545641e-05, "weight_mean": 9.233446689904667e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7106915038311854e-05 }, { "epoch": 0.14116501703715723, "step": 435, "weight_max": 3.397988257347606e-05, "weight_mean": 4.2480787669774145e-06, "weight_min": 0.0, "weight_std": 1.2013462765025906e-05 }, { "epoch": 0.14116501703715723, "step": 435, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074878446175717e-05, "weight_min": 0.0, "weight_std": 1.8654136511031538e-05 }, { "epoch": 0.14116501703715723, "step": 435, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.442394386685919e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.994668855331838e-05 }, { "epoch": 0.14148953431770242, "step": 436, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5388184692710638e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.123986996593885e-05 }, { "epoch": 0.14148953431770242, "step": 436, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4621567061112728e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0244429833837785e-05 }, { "epoch": 0.14148953431770242, "step": 436, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235702701611444e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4806288163526915e-05 }, { "epoch": 0.14148953431770242, "step": 436, "weight_max": 4.252122016623616e-05, "weight_mean": 1.375394822389353e-05, "weight_min": 0.0, "weight_std": 1.923439049278386e-05 }, { "epoch": 0.14181405159824761, "step": 437, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0577243301668204e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.6548443454666995e-05 }, { "epoch": 0.14181405159824761, "step": 437, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745264833327383e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.814201459637843e-05 }, { "epoch": 0.14181405159824761, "step": 437, "weight_max": 4.252122016623616e-05, "weight_mean": 9.670380677562207e-06, "weight_min": 0.0, "weight_std": 1.7638953067944385e-05 }, { "epoch": 0.14181405159824761, "step": 437, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.724899427965283e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.800594327505678e-05 }, { "epoch": 0.1421385688787928, "step": 438, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233246030111331e-05, "weight_min": 0.0, "weight_std": 1.8954033293994144e-05 }, { "epoch": 0.1421385688787928, "step": 438, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.6155669224390294e-06, "weight_min": 0.0, "weight_std": 1.3053181646682788e-05 }, { "epoch": 0.1421385688787928, "step": 438, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.3077376934234053e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7765809388947673e-05 }, { "epoch": 0.1421385688787928, "step": 438, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4455868949880823e-05, "weight_min": 0.0, "weight_std": 1.9963117665611207e-05 }, { "epoch": 0.14246308615933798, "step": 439, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5158447240537498e-05, "weight_min": 0.0, "weight_std": 2.0952335034962744e-05 }, { "epoch": 0.14246308615933798, "step": 439, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156511633686023e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4580924471374601e-05 }, { "epoch": 0.14246308615933798, "step": 439, "weight_max": 4.252122016623616e-05, "weight_mean": 5.559322744375095e-06, "weight_min": 0.0, "weight_std": 1.4950387594581116e-05 }, { "epoch": 0.14246308615933798, "step": 439, "weight_max": 3.143761568935588e-05, "weight_mean": 7.135296982596628e-06, "weight_min": 0.0, "weight_std": 1.330089071416296e-05 }, { "epoch": 0.14278760343988317, "grad_norm": 0.32711660861968994, "learning_rate": 9.310204081632653e-06, "loss": 0.0986, "step": 440 }, { "epoch": 0.14278760343988317, "step": 440, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.511208392912522e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.085745654767379e-05 }, { "epoch": 0.14278760343988317, "step": 440, "weight_max": 1.7424939869670197e-05, "weight_mean": 3.964089955843519e-06, "weight_min": 5.958802340444436e-10, "weight_std": 7.386127890640637e-06 }, { "epoch": 0.14278760343988317, "step": 440, "weight_max": 3.998322426923551e-05, "weight_mean": 2.300785854458809e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.92421066458337e-05 }, { "epoch": 0.14278760343988317, "step": 440, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0476704119355418e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9422803234192543e-05 }, { "epoch": 0.14311212072042837, "step": 441, "weight_max": 4.381950566312298e-05, "weight_mean": 2.11045680771349e-05, "weight_min": 7.091466858177853e-10, "weight_std": 2.2577074560103938e-05 }, { "epoch": 0.14311212072042837, "step": 441, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921059826301644e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3915807357989252e-05 }, { "epoch": 0.14311212072042837, "step": 441, "weight_max": 3.782653220696375e-05, "weight_mean": 7.453001217072597e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4448990441451315e-05 }, { "epoch": 0.14311212072042837, "step": 441, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.918206054251641e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.471993437007768e-05 }, { "epoch": 0.14343663800097356, "step": 442, "weight_max": 4.187939703115262e-05, "weight_mean": 1.047046225721715e-05, "weight_min": 0.0, "weight_std": 1.938601599249523e-05 }, { "epoch": 0.14343663800097356, "step": 442, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2386448361212388e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.7398904674337246e-05 }, { "epoch": 0.14343663800097356, "step": 442, "weight_max": 2.952862632810138e-05, "weight_mean": 3.6919507238053484e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0439593097544275e-05 }, { "epoch": 0.14343663800097356, "step": 442, "weight_max": 2.319888153579086e-05, "weight_mean": 2.9008399451413425e-06, "weight_min": 0.0, "weight_std": 8.201647688110825e-06 }, { "epoch": 0.14376115528151875, "step": 443, "weight_max": 3.782653220696375e-05, "weight_mean": 9.38141965889372e-06, "weight_min": 0.0, "weight_std": 1.7370684872730635e-05 }, { "epoch": 0.14376115528151875, "step": 443, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4070056749915238e-05, "weight_min": 0.0, "weight_std": 1.9724669982679188e-05 }, { "epoch": 0.14376115528151875, "step": 443, "weight_max": 3.967146767536178e-05, "weight_mean": 1.87523546628654e-05, "weight_min": 0.0, "weight_std": 1.6676563973305747e-05 }, { "epoch": 0.14376115528151875, "step": 443, "weight_max": 4.772478412462533e-09, "weight_mean": 1.4846106566324124e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.4970497064226151e-09 }, { "epoch": 0.14408567256206392, "step": 444, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2976690413779579e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.798361336113885e-05 }, { "epoch": 0.14408567256206392, "step": 444, "weight_max": 4.187939703115262e-05, "weight_mean": 9.818591934163123e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8219867342850193e-05 }, { "epoch": 0.14408567256206392, "step": 444, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1597810953389853e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6680989574524574e-05 }, { "epoch": 0.14408567256206392, "step": 444, "weight_max": 2.4160392058547586e-05, "weight_mean": 3.1101949389267247e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.509241524734534e-06 }, { "epoch": 0.14441018984260912, "step": 445, "weight_max": 4.187939703115262e-05, "weight_mean": 8.866762073012069e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6757061530370265e-05 }, { "epoch": 0.14441018984260912, "step": 445, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077080913906684e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4357630789163522e-05 }, { "epoch": 0.14441018984260912, "step": 445, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.572327144269366e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.791416434571147e-05 }, { "epoch": 0.14441018984260912, "step": 445, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3311229849932715e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8617984096636064e-05 }, { "epoch": 0.1447347071231543, "step": 446, "weight_max": 3.843663216684945e-05, "weight_mean": 1.8615884982864372e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9920194972655736e-05 }, { "epoch": 0.1447347071231543, "step": 446, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235786248813383e-05, "weight_min": 0.0, "weight_std": 1.8970526070916094e-05 }, { "epoch": 0.1447347071231543, "step": 446, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827178716164781e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0822765943885315e-05 }, { "epoch": 0.1447347071231543, "step": 446, "weight_max": 3.843663216684945e-05, "weight_mean": 1.6551814042031765e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7950036635738797e-05 }, { "epoch": 0.1450592244036995, "step": 447, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3237251550890505e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.83653028216213e-05 }, { "epoch": 0.1450592244036995, "step": 447, "weight_max": 9.013804813662318e-09, "weight_mean": 1.669298366202554e-09, "weight_min": 5.771870759119224e-10, "weight_std": 2.967804890730008e-09 }, { "epoch": 0.1450592244036995, "step": 447, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.410343070863746e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.566496393934358e-05 }, { "epoch": 0.1450592244036995, "step": 447, "weight_max": 6.855033007013844e-06, "weight_mean": 8.575087804274517e-07, "weight_min": 5.958802340444436e-10, "weight_std": 2.4233656858996255e-06 }, { "epoch": 0.1453837416842447, "step": 448, "weight_max": 3.998322426923551e-05, "weight_mean": 9.247146408597473e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.719307510938961e-05 }, { "epoch": 0.1453837416842447, "step": 448, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921164872939698e-06, "weight_min": 0.0, "weight_std": 1.391576552123297e-05 }, { "epoch": 0.1453837416842447, "step": 448, "weight_max": 4.187939703115262e-05, "weight_mean": 9.888048225548118e-06, "weight_min": 0.0, "weight_std": 1.8350254322285764e-05 }, { "epoch": 0.1453837416842447, "step": 448, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.964269126299769e-06, "weight_min": 0.0, "weight_std": 1.0459711120347492e-05 }, { "epoch": 0.14570825896478987, "step": 449, "weight_max": 3.813097282545641e-05, "weight_mean": 9.160110494121909e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.697744482953567e-05 }, { "epoch": 0.14570825896478987, "step": 449, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8051260819192976e-06, "weight_min": 0.0, "weight_std": 1.358918052574154e-05 }, { "epoch": 0.14570825896478987, "step": 449, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0166116883046925e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1628899048664607e-05 }, { "epoch": 0.14570825896478987, "step": 449, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.4394799816655e-06, "weight_min": 0.0, "weight_std": 1.7576110622030683e-05 }, { "epoch": 0.14603277624533506, "grad_norm": 0.22939084470272064, "learning_rate": 9.289795918367348e-06, "loss": 0.1933, "step": 450 }, { "epoch": 0.14603277624533506, "step": 450, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4656045095762238e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.043058157141786e-05 }, { "epoch": 0.14603277624533506, "step": 450, "weight_max": 4.252122016623616e-05, "weight_mean": 8.011087629711255e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.585472455190029e-05 }, { "epoch": 0.14603277624533506, "step": 450, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3603681509266607e-05, "weight_min": 0.0, "weight_std": 1.775891905708704e-05 }, { "epoch": 0.14603277624533506, "step": 450, "weight_max": 4.187939703115262e-05, "weight_mean": 1.488518864789512e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0574458176270127e-05 }, { "epoch": 0.14635729352588026, "step": 451, "weight_max": 3.4268457966391e-05, "weight_mean": 8.353625162271783e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.547306965221651e-05 }, { "epoch": 0.14635729352588026, "step": 451, "weight_max": 3.998322426923551e-05, "weight_mean": 9.765832146513276e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.808573870221153e-05 }, { "epoch": 0.14635729352588026, "step": 451, "weight_max": 3.662098970380612e-05, "weight_mean": 4.57824717159383e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2947223694936838e-05 }, { "epoch": 0.14635729352588026, "step": 451, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232139175059274e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8945682313642465e-05 }, { "epoch": 0.14668181080642545, "step": 452, "weight_max": 4.6178138290997595e-06, "weight_mean": 5.777511091764609e-07, "weight_min": 0.0, "weight_std": 1.6324319176419522e-06 }, { "epoch": 0.14668181080642545, "step": 452, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1483388334454503e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7690568711259402e-05 }, { "epoch": 0.14668181080642545, "step": 452, "weight_max": 4.252122016623616e-05, "weight_mean": 1.031359988701297e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9107776097371243e-05 }, { "epoch": 0.14668181080642545, "step": 452, "weight_max": 4.252122016623616e-05, "weight_mean": 5.31588102603564e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.50332271005027e-05 }, { "epoch": 0.14700632808697064, "step": 453, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.361114280181937e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7599837519810535e-05 }, { "epoch": 0.14700632808697064, "step": 453, "weight_max": 1.076157385426768e-09, "weight_mean": 6.972947774741556e-10, "weight_min": 5.864964625068581e-10, "weight_std": 1.7076126623827292e-10 }, { "epoch": 0.14700632808697064, "step": 453, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6672551282681525e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.936663466040045e-05 }, { "epoch": 0.14700632808697064, "step": 453, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2638477528525982e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7760638002073392e-05 }, { "epoch": 0.1473308453675158, "step": 454, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4243265468394384e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9900004190276377e-05 }, { "epoch": 0.1473308453675158, "step": 454, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805447588296374e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3589050467999186e-05 }, { "epoch": 0.1473308453675158, "step": 454, "weight_max": 1.8287995544596924e-06, "weight_mean": 2.3018640149530256e-07, "weight_min": 0.0, "weight_std": 6.459403039116296e-07 }, { "epoch": 0.1473308453675158, "step": 454, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1518215615069494e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.895476452773437e-05 }, { "epoch": 0.147655362648061, "step": 455, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.78788523550611e-06, "weight_min": 0.0, "weight_std": 1.6266934835584834e-05 }, { "epoch": 0.147655362648061, "step": 455, "weight_max": 3.9360926166409627e-05, "weight_mean": 5.885851805942366e-06, "weight_min": 0.0, "weight_std": 1.3793157449981663e-05 }, { "epoch": 0.147655362648061, "step": 455, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.007499304250814e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.86540655704448e-05 }, { "epoch": 0.147655362648061, "step": 455, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0354906155262142e-05, "weight_min": 0.0, "weight_std": 2.180239971494302e-05 }, { "epoch": 0.1479798799286062, "step": 456, "weight_max": 1.4305072681963793e-09, "weight_mean": 8.759010738934592e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.675017152231618e-10 }, { "epoch": 0.1479798799286062, "step": 456, "weight_max": 4.187939703115262e-05, "weight_mean": 1.620873990759719e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.7962158381124027e-05 }, { "epoch": 0.1479798799286062, "step": 456, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6094061720650643e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1077727069496177e-05 }, { "epoch": 0.1479798799286062, "step": 456, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2895005966129247e-05, "weight_min": 0.0, "weight_std": 1.7970589397009462e-05 }, { "epoch": 0.1483043972091514, "step": 457, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390942406957038e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.92392853932688e-05 }, { "epoch": 0.1483043972091514, "step": 457, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5707713828305714e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1688378183171153e-05 }, { "epoch": 0.1483043972091514, "step": 457, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4093274330662098e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9694571165018715e-05 }, { "epoch": 0.1483043972091514, "step": 457, "weight_max": 2.2534512034155796e-09, "weight_mean": 1.0926675120259688e-09, "weight_min": 6.100955851628953e-10, "weight_std": 5.798852509286689e-10 }, { "epoch": 0.1486289144896966, "step": 458, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921066192764556e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3915805538999848e-05 }, { "epoch": 0.1486289144896966, "step": 458, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070599516126094e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.1509561772982124e-05 }, { "epoch": 0.1486289144896966, "step": 458, "weight_max": 3.998322426923551e-05, "weight_mean": 9.65115032158792e-06, "weight_min": 0.0, "weight_std": 1.7884260159917176e-05 }, { "epoch": 0.1486289144896966, "step": 458, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155883627594449e-06, "weight_min": 0.0, "weight_std": 1.4581179129891098e-05 }, { "epoch": 0.14895343177024176, "step": 459, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9550656361388974e-05, "weight_min": 0.0, "weight_std": 2.115395363944117e-05 }, { "epoch": 0.14895343177024176, "step": 459, "weight_max": 4.187939703115262e-05, "weight_mean": 9.628387488191947e-06, "weight_min": 0.0, "weight_std": 1.791712020349223e-05 }, { "epoch": 0.14895343177024176, "step": 459, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827435193670681e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0822662261489313e-05 }, { "epoch": 0.14895343177024176, "step": 459, "weight_max": 4.187939703115262e-05, "weight_mean": 1.071338374458719e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.861926466517616e-05 }, { "epoch": 0.14927794905078695, "grad_norm": 0.34643033146858215, "learning_rate": 9.269387755102042e-06, "loss": 0.3146, "step": 460 }, { "epoch": 0.14927794905078695, "step": 460, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076748493593186e-06, "weight_min": 0.0, "weight_std": 1.4357765394379385e-05 }, { "epoch": 0.14927794905078695, "step": 460, "weight_max": 3.843663216684945e-05, "weight_mean": 1.0070532880490646e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.560810414957814e-05 }, { "epoch": 0.14927794905078695, "step": 460, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.8518509023124352e-05, "weight_min": 0.0, "weight_std": 1.987874202313833e-05 }, { "epoch": 0.14927794905078695, "step": 460, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.9683797038160264e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.835845068853814e-05 }, { "epoch": 0.14960246633133215, "step": 461, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.369268809678033e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7386166291544214e-05 }, { "epoch": 0.14960246633133215, "step": 461, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397232143877773e-06, "weight_min": 0.0, "weight_std": 1.5261663065757602e-05 }, { "epoch": 0.14960246633133215, "step": 461, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.431045868434012e-05, "weight_min": 0.0, "weight_std": 2.03637591766892e-05 }, { "epoch": 0.14960246633133215, "step": 461, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1688747690641321e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.7988140825764276e-05 }, { "epoch": 0.14992698361187734, "step": 462, "weight_max": 4.187939703115262e-05, "weight_mean": 9.2348482212401e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.730087387841195e-05 }, { "epoch": 0.14992698361187734, "step": 462, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.724853953230195e-06, "weight_min": 0.0, "weight_std": 1.8005972378887236e-05 }, { "epoch": 0.14992698361187734, "step": 462, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2877495464636013e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8056269254884683e-05 }, { "epoch": 0.14992698361187734, "step": 462, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3050877896603197e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8384618670097552e-05 }, { "epoch": 0.15025150089242253, "step": 463, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.164928997051902e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7235814084415324e-05 }, { "epoch": 0.15025150089242253, "step": 463, "weight_max": 1.252662062967147e-08, "weight_mean": 2.1108101933009493e-09, "weight_min": 0.0, "weight_std": 4.217682203488948e-09 }, { "epoch": 0.15025150089242253, "step": 463, "weight_max": 4.0610404539620504e-05, "weight_mean": 6.042022505425848e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4226628081814852e-05 }, { "epoch": 0.15025150089242253, "step": 463, "weight_max": 1.868417864869798e-09, "weight_mean": 9.296624581267565e-10, "weight_min": 5.771870759119224e-10, "weight_std": 4.15466439029899e-10 }, { "epoch": 0.1505760181729677, "step": 464, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.783177867764607e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.966508716577664e-05 }, { "epoch": 0.1505760181729677, "step": 464, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965982159570558e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1212380741199013e-05 }, { "epoch": 0.1505760181729677, "step": 464, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.333210199547466e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.2212811270728707e-05 }, { "epoch": 0.1505760181729677, "step": 464, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5435231186274905e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2840708222938702e-05 }, { "epoch": 0.1509005354535129, "step": 465, "weight_max": 4.187939703115262e-05, "weight_mean": 2.237171065644361e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9721697753993794e-05 }, { "epoch": 0.1509005354535129, "step": 465, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4691822798340581e-05, "weight_min": 0.0, "weight_std": 2.0475230485317297e-05 }, { "epoch": 0.1509005354535129, "step": 465, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311174264643341e-05, "weight_min": 0.0, "weight_std": 1.9091203284915537e-05 }, { "epoch": 0.1509005354535129, "step": 465, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998629265173804e-06, "weight_min": 0.0, "weight_std": 1.4135910532786511e-05 }, { "epoch": 0.1512250527340581, "step": 466, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0131761630182154e-05, "weight_min": 0.0, "weight_std": 2.161449083359912e-05 }, { "epoch": 0.1512250527340581, "step": 466, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4883776202623267e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0569737898767926e-05 }, { "epoch": 0.1512250527340581, "step": 466, "weight_max": 1.3183556802687235e-05, "weight_mean": 1.6497990600328194e-06, "weight_min": 6.196655966128617e-10, "weight_std": 4.6603422561020125e-06 }, { "epoch": 0.1512250527340581, "step": 466, "weight_max": 3.722132169059478e-05, "weight_mean": 4.694013114203699e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3143510841473471e-05 }, { "epoch": 0.15154957001460329, "step": 467, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4081916560826357e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9785858967225067e-05 }, { "epoch": 0.15154957001460329, "step": 467, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3947126717539504e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.980623710551299e-05 }, { "epoch": 0.15154957001460329, "step": 467, "weight_max": 3.22740561387036e-05, "weight_mean": 4.034926405438455e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1410331353545189e-05 }, { "epoch": 0.15154957001460329, "step": 467, "weight_max": 4.381950566312298e-05, "weight_mean": 1.3901044439990073e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9851524484693073e-05 }, { "epoch": 0.15187408729514848, "step": 468, "weight_max": 3.9051607018336654e-05, "weight_mean": 5.277327545627486e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3691652384295594e-05 }, { "epoch": 0.15187408729514848, "step": 468, "weight_max": 3.662098970380612e-05, "weight_mean": 4.588600404531462e-06, "weight_min": 0.0, "weight_std": 1.2943072761117946e-05 }, { "epoch": 0.15187408729514848, "step": 468, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3868848327547312e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9240504116169177e-05 }, { "epoch": 0.15187408729514848, "step": 468, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.743748028588016e-06, "weight_min": 0.0, "weight_std": 1.5234537386277225e-05 }, { "epoch": 0.15219860457569365, "step": 469, "weight_max": 4.252122016623616e-05, "weight_mean": 1.047189834935125e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.938964669534471e-05 }, { "epoch": 0.15219860457569365, "step": 469, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767062364408048e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3481055248121265e-05 }, { "epoch": 0.15219860457569365, "step": 469, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.883030669589061e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3806189599563368e-05 }, { "epoch": 0.15219860457569365, "step": 469, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390967872808687e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9239270841353573e-05 }, { "epoch": 0.15252312185623884, "grad_norm": 0.5971319079399109, "learning_rate": 9.248979591836735e-06, "loss": 0.3087, "step": 470 }, { "epoch": 0.15252312185623884, "step": 470, "weight_max": 4.252122016623616e-05, "weight_mean": 1.910430182761047e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.058662175841164e-05 }, { "epoch": 0.15252312185623884, "step": 470, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745056559040677e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8142143744626082e-05 }, { "epoch": 0.15252312185623884, "step": 470, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2389717994665261e-05, "weight_min": 0.0, "weight_std": 1.5387209714390337e-05 }, { "epoch": 0.15252312185623884, "step": 470, "weight_max": 4.3167921830900013e-05, "weight_mean": 3.0014885851414874e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.865946251200512e-05 }, { "epoch": 0.15284763913678404, "step": 471, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745010174810886e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.814217284845654e-05 }, { "epoch": 0.15284763913678404, "step": 471, "weight_max": 2.846461757144425e-05, "weight_mean": 6.763766123185633e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2544879609777126e-05 }, { "epoch": 0.15284763913678404, "step": 471, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0864326870650984e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.2320184143609367e-05 }, { "epoch": 0.15284763913678404, "step": 471, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0240036317554768e-05, "weight_min": 0.0, "weight_std": 1.8994969650520943e-05 }, { "epoch": 0.15317215641732923, "step": 472, "weight_max": 3.967146767536178e-05, "weight_mean": 7.137638021958992e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4490819012280554e-05 }, { "epoch": 0.15317215641732923, "step": 472, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3413187843980268e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9022230844711885e-05 }, { "epoch": 0.15317215641732923, "step": 472, "weight_max": 3.1715208024252206e-05, "weight_mean": 7.328111678361893e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3628411579702515e-05 }, { "epoch": 0.15317215641732923, "step": 472, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843831902689999e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3697536815016065e-05 }, { "epoch": 0.15349667369787443, "step": 473, "weight_max": 3.998322426923551e-05, "weight_mean": 9.103045158553869e-06, "weight_min": 0.0, "weight_std": 1.696237268333789e-05 }, { "epoch": 0.15349667369787443, "step": 473, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.217490176321007e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5251251170411706e-05 }, { "epoch": 0.15349667369787443, "step": 473, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0967383786919527e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.6900083210202865e-05 }, { "epoch": 0.15349667369787443, "step": 473, "weight_max": 3.602553260861896e-05, "weight_mean": 8.933320714277215e-06, "weight_min": 0.0, "weight_std": 1.654017614782788e-05 }, { "epoch": 0.1538211909784196, "step": 474, "weight_max": 2.4160392058547586e-05, "weight_mean": 3.0211285775294527e-06, "weight_min": 0.0, "weight_std": 8.54155223350972e-06 }, { "epoch": 0.1538211909784196, "step": 474, "weight_max": 1.3727149417164242e-09, "weight_mean": 9.188042549013176e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.9658825395628696e-10 }, { "epoch": 0.1538211909784196, "step": 474, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2929431250086054e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8177377569372766e-05 }, { "epoch": 0.1538211909784196, "step": 474, "weight_max": 2.926079287135508e-05, "weight_mean": 3.658324430944049e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0344958354835398e-05 }, { "epoch": 0.1541457082589648, "step": 475, "weight_max": 3.843663216684945e-05, "weight_mean": 4.80528615298681e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3589115951617714e-05 }, { "epoch": 0.1541457082589648, "step": 475, "weight_max": 4.252122016623616e-05, "weight_mean": 8.71090560394805e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6642428818158805e-05 }, { "epoch": 0.1541457082589648, "step": 475, "weight_max": 2.4894337911973707e-05, "weight_mean": 3.112832473561866e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.801057447271887e-06 }, { "epoch": 0.1541457082589648, "step": 475, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.359435585385654e-06, "weight_min": 0.0, "weight_std": 1.4877239664201625e-05 }, { "epoch": 0.15447022553950998, "step": 476, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.023208551487187e-05, "weight_min": 0.0, "weight_std": 1.894571505545173e-05 }, { "epoch": 0.15447022553950998, "step": 476, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0470970664755441e-05, "weight_min": 0.0, "weight_std": 1.939022149599623e-05 }, { "epoch": 0.15447022553950998, "step": 476, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0172032236587256e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.1660358470398933e-05 }, { "epoch": 0.15447022553950998, "step": 476, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.626234385796124e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.304889974562684e-05 }, { "epoch": 0.15479474282005518, "step": 477, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2357746527413838e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8473017917131074e-05 }, { "epoch": 0.15479474282005518, "step": 477, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3399299859884195e-05, "weight_min": 0.0, "weight_std": 1.881712523754686e-05 }, { "epoch": 0.15479474282005518, "step": 477, "weight_max": 4.252122016623616e-05, "weight_mean": 2.8889437089674175e-05, "weight_min": 0.0, "weight_std": 1.8283904864802025e-05 }, { "epoch": 0.15479474282005518, "step": 477, "weight_max": 3.967146767536178e-05, "weight_mean": 9.80244658421725e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8150880350731313e-05 }, { "epoch": 0.15511926010060037, "step": 478, "weight_max": 4.252122016623616e-05, "weight_mean": 5.317481736710761e-06, "weight_min": 0.0, "weight_std": 1.5032580449769739e-05 }, { "epoch": 0.15511926010060037, "step": 478, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2974142009625211e-05, "weight_min": 0.0, "weight_std": 1.8122420442523435e-05 }, { "epoch": 0.15511926010060037, "step": 478, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3917578144173604e-05, "weight_min": 0.0, "weight_std": 1.9512654034770094e-05 }, { "epoch": 0.15511926010060037, "step": 478, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.862396857177373e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9500359485391527e-05 }, { "epoch": 0.15544377738114554, "step": 479, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8276969058206305e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0001583834527992e-05 }, { "epoch": 0.15544377738114554, "step": 479, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311747246305458e-05, "weight_min": 0.0, "weight_std": 1.9095368770649657e-05 }, { "epoch": 0.15544377738114554, "step": 479, "weight_max": 3.722132169059478e-05, "weight_mean": 8.865107702149544e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6439898899989203e-05 }, { "epoch": 0.15544377738114554, "step": 479, "weight_max": 9.413956281179026e-10, "weight_mean": 6.947393771383759e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.3340051285837262e-10 }, { "epoch": 0.15576829466169073, "grad_norm": 5.009500026702881, "learning_rate": 9.22857142857143e-06, "loss": 0.2173, "step": 480 }, { "epoch": 0.15576829466169073, "step": 480, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.257335477741435e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7451295207138173e-05 }, { "epoch": 0.15576829466169073, "step": 480, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5785961295478046e-05, "weight_min": 0.0, "weight_std": 2.1789499442093074e-05 }, { "epoch": 0.15576829466169073, "step": 480, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0296813343302347e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.8467251720721833e-05 }, { "epoch": 0.15576829466169073, "step": 480, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0279672096658032e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6027132005547173e-05 }, { "epoch": 0.15609281194223593, "step": 481, "weight_max": 3.782653220696375e-05, "weight_mean": 8.905477443477139e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.6528647392988205e-05 }, { "epoch": 0.15609281194223593, "step": 481, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3780405424768105e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9216875443817116e-05 }, { "epoch": 0.15609281194223593, "step": 481, "weight_max": 4.252122016623616e-05, "weight_mean": 8.489508218190167e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6371224774047732e-05 }, { "epoch": 0.15609281194223593, "step": 481, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3289974958752282e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8471228031557985e-05 }, { "epoch": 0.15641732922278112, "step": 482, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2821226846426725e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8040573195321485e-05 }, { "epoch": 0.15641732922278112, "step": 482, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4276400179369375e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.974455881281756e-05 }, { "epoch": 0.15641732922278112, "step": 482, "weight_max": 3.602553260861896e-05, "weight_mean": 4.503849595494103e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.273668294743402e-05 }, { "epoch": 0.15641732922278112, "step": 482, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1325448213028722e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.5116233043954708e-05 }, { "epoch": 0.15674184650332632, "step": 483, "weight_max": 3.602553260861896e-05, "weight_mean": 8.40100346977124e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.560332930239383e-05 }, { "epoch": 0.15674184650332632, "step": 483, "weight_max": 4.252122016623616e-05, "weight_mean": 9.818932085181586e-06, "weight_min": 0.0, "weight_std": 1.8262380763189867e-05 }, { "epoch": 0.15674184650332632, "step": 483, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.090819730772637e-06, "weight_min": 0.0, "weight_std": 1.6880030671018176e-05 }, { "epoch": 0.15674184650332632, "step": 483, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6954400052782148e-05, "weight_min": 0.0, "weight_std": 1.934514511958696e-05 }, { "epoch": 0.15706636378387148, "step": 484, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0713729352573864e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8016984540736303e-05 }, { "epoch": 0.15706636378387148, "step": 484, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5080515368026681e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0844006940023974e-05 }, { "epoch": 0.15706636378387148, "step": 484, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156198312761262e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.458105180063285e-05 }, { "epoch": 0.15706636378387148, "step": 484, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4353750884765759e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.988038820854854e-05 }, { "epoch": 0.15739088106441668, "step": 485, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5259668038634118e-06, "weight_min": 0.0, "weight_std": 9.970487553800922e-06 }, { "epoch": 0.15739088106441668, "step": 485, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3610102541861124e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.904022428789176e-05 }, { "epoch": 0.15739088106441668, "step": 485, "weight_max": 3.662098970380612e-05, "weight_mean": 1.3073841728328262e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8063201423501596e-05 }, { "epoch": 0.15739088106441668, "step": 485, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2175023584859446e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7131442291429266e-05 }, { "epoch": 0.15771539834496187, "step": 486, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5865703971940093e-05, "weight_min": 0.0, "weight_std": 2.189659971918445e-05 }, { "epoch": 0.15771539834496187, "step": 486, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393656581669347e-06, "weight_min": 0.0, "weight_std": 1.2424006854416803e-05 }, { "epoch": 0.15771539834496187, "step": 486, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.948484719439875e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6821835743030533e-05 }, { "epoch": 0.15771539834496187, "step": 486, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.733452316140756e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8063841707771644e-05 }, { "epoch": 0.15803991562550707, "step": 487, "weight_max": 3.572963396436535e-05, "weight_mean": 9.203705303661991e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5525263734161854e-05 }, { "epoch": 0.15803991562550707, "step": 487, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8438202459947206e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9933464500354603e-05 }, { "epoch": 0.15803991562550707, "step": 487, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.885866731754504e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6553516616113484e-05 }, { "epoch": 0.15803991562550707, "step": 487, "weight_max": 4.187939703115262e-05, "weight_mean": 5.248813522484852e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.480103765061358e-05 }, { "epoch": 0.15836443290605226, "step": 488, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076798515801784e-06, "weight_min": 0.0, "weight_std": 1.4357745385495946e-05 }, { "epoch": 0.15836443290605226, "step": 488, "weight_max": 1.0539043415747074e-08, "weight_mean": 2.612340566088278e-09, "weight_min": 6.053385570581327e-10, "weight_std": 3.558583872020904e-09 }, { "epoch": 0.15836443290605226, "step": 488, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4610899597755633e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0179326384095475e-05 }, { "epoch": 0.15836443290605226, "step": 488, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.616254955180921e-06, "weight_min": 0.0, "weight_std": 1.3052903341304045e-05 }, { "epoch": 0.15868895018659743, "step": 489, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0194387868978083e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8884253222495317e-05 }, { "epoch": 0.15868895018659743, "step": 489, "weight_max": 3.998322426923551e-05, "weight_mean": 9.689001672086306e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7950902474694885e-05 }, { "epoch": 0.15868895018659743, "step": 489, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3883065548725426e-05, "weight_min": 0.0, "weight_std": 1.9758161215577275e-05 }, { "epoch": 0.15868895018659743, "step": 489, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396893357101362e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.526179948996287e-05 }, { "epoch": 0.15901346746714262, "grad_norm": 0.7405959367752075, "learning_rate": 9.208163265306123e-06, "loss": 0.1558, "step": 490 }, { "epoch": 0.15901346746714262, "step": 490, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8362834453000687e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.061522172880359e-05 }, { "epoch": 0.15901346746714262, "step": 490, "weight_max": 3.998322426923551e-05, "weight_mean": 9.880002835416235e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8294304027222097e-05 }, { "epoch": 0.15901346746714262, "step": 490, "weight_max": 4.252122016623616e-05, "weight_mean": 9.527469956083223e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.77967049239669e-05 }, { "epoch": 0.15901346746714262, "step": 490, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5281671949196607e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0986701201763935e-05 }, { "epoch": 0.15933798474768782, "step": 491, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.455287176417187e-05, "weight_min": 8.156453290553145e-10, "weight_std": 2.0460156520130113e-05 }, { "epoch": 0.15933798474768782, "step": 491, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.078519734524889e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.435704962204909e-05 }, { "epoch": 0.15933798474768782, "step": 491, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.873663131846115e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.024037530645728e-05 }, { "epoch": 0.15933798474768782, "step": 491, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959622401656816e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.402570433128858e-05 }, { "epoch": 0.159662502028233, "step": 492, "weight_max": 3.572963396436535e-05, "weight_mean": 1.1627597814367618e-05, "weight_min": 0.0, "weight_std": 1.636963861528784e-05 }, { "epoch": 0.159662502028233, "step": 492, "weight_max": 3.722132169059478e-05, "weight_mean": 9.0460680439719e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.675751991569996e-05 }, { "epoch": 0.159662502028233, "step": 492, "weight_max": 4.252122016623616e-05, "weight_mean": 8.841126145853195e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.681013236520812e-05 }, { "epoch": 0.159662502028233, "step": 492, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3710940038436092e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.89884831343079e-05 }, { "epoch": 0.1599870193087782, "step": 493, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6715435776859522e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8479497157386504e-05 }, { "epoch": 0.1599870193087782, "step": 493, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5260245567769744e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.970464816433378e-06 }, { "epoch": 0.1599870193087782, "step": 493, "weight_max": 4.187939703115262e-05, "weight_mean": 1.530932422610931e-05, "weight_min": 0.0, "weight_std": 2.113475238729734e-05 }, { "epoch": 0.1599870193087782, "step": 493, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4684852430946194e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0272264009690844e-05 }, { "epoch": 0.16031153658932337, "step": 494, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.431911914551165e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9829594748443924e-05 }, { "epoch": 0.16031153658932337, "step": 494, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691854883276392e-06, "weight_min": 0.0, "weight_std": 1.3265916095406283e-05 }, { "epoch": 0.16031153658932337, "step": 494, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3724567907047458e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.936076660058461e-05 }, { "epoch": 0.16031153658932337, "step": 494, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0077626029669773e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8652443031896837e-05 }, { "epoch": 0.16063605386986857, "step": 495, "weight_max": 4.252122016623616e-05, "weight_mean": 1.800957033992745e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9575063561205752e-05 }, { "epoch": 0.16063605386986857, "step": 495, "weight_max": 1.753466261078529e-09, "weight_mean": 9.746671247867766e-10, "weight_min": 6.29309993005478e-10, "weight_std": 4.235450878908864e-10 }, { "epoch": 0.16063605386986857, "step": 495, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5233536032610573e-05, "weight_min": 0.0, "weight_std": 2.1042507796664722e-05 }, { "epoch": 0.16063605386986857, "step": 495, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.937672646425199e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6750689610489644e-05 }, { "epoch": 0.16096057115041376, "step": 496, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3655077054863796e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9156756025040522e-05 }, { "epoch": 0.16096057115041376, "step": 496, "weight_max": 2.7939937353949063e-05, "weight_mean": 3.4931526897707954e-06, "weight_min": 0.0, "weight_std": 9.877992852125317e-06 }, { "epoch": 0.16096057115041376, "step": 496, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3323053281055763e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.852189961937256e-05 }, { "epoch": 0.16096057115041376, "step": 496, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.98595351120457e-06, "weight_min": 0.0, "weight_std": 1.7669270164333284e-05 }, { "epoch": 0.16128508843095896, "step": 497, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0116932571690995e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.874678309832234e-05 }, { "epoch": 0.16128508843095896, "step": 497, "weight_max": 3.998322426923551e-05, "weight_mean": 5.004168542654952e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4133681361272465e-05 }, { "epoch": 0.16128508843095896, "step": 497, "weight_max": 4.3167921830900013e-05, "weight_mean": 7.892967914813198e-06, "weight_min": 0.0, "weight_std": 1.583783159730956e-05 }, { "epoch": 0.16128508843095896, "step": 497, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070140221301699e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1509747309901286e-05 }, { "epoch": 0.16160960571150415, "step": 498, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1881124009960331e-05, "weight_min": 0.0, "weight_std": 1.7693238987703808e-05 }, { "epoch": 0.16160960571150415, "step": 498, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.065851392922923e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.836475894378964e-05 }, { "epoch": 0.16160960571150415, "step": 498, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0116893463418819e-05, "weight_min": 0.0, "weight_std": 1.874680856417399e-05 }, { "epoch": 0.16160960571150415, "step": 498, "weight_max": 4.187939703115262e-05, "weight_mean": 7.900368473201524e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4705628018418793e-05 }, { "epoch": 0.16193412299204932, "step": 499, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235649496171391e-06, "weight_min": 0.0, "weight_std": 1.4806309991399758e-05 }, { "epoch": 0.16193412299204932, "step": 499, "weight_max": 4.252122016623616e-05, "weight_mean": 1.791645263438113e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9379178411327302e-05 }, { "epoch": 0.16193412299204932, "step": 499, "weight_max": 1.3301527657105794e-09, "weight_mean": 5.885107401404355e-10, "weight_min": 0.0, "weight_std": 4.368375938756941e-10 }, { "epoch": 0.16193412299204932, "step": 499, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076984052720945e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357669897435699e-05 }, { "epoch": 0.1622586402725945, "grad_norm": 5.241079330444336, "learning_rate": 9.187755102040818e-06, "loss": 0.2291, "step": 500 }, { "epoch": 0.1622586402725945, "eval_accuracy": 0.494, "eval_auc": 0.5509873421772735, "eval_loss": 2.557065010070801, "eval_runtime": 155.337, "eval_samples_per_second": 12.875, "eval_steps_per_second": 1.609, "step": 500 }, { "epoch": 0.1622586402725945, "step": 500, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8068925783154555e-05, "weight_min": 0.0, "weight_std": 1.955018160515465e-05 }, { "epoch": 0.1622586402725945, "step": 500, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0002106137108058e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.8545380953582935e-05 }, { "epoch": 0.1622586402725945, "step": 500, "weight_max": 1.901931279135738e-09, "weight_mean": 7.594275208688828e-10, "weight_min": 0.0, "weight_std": 5.278641967976228e-10 }, { "epoch": 0.1622586402725945, "step": 500, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.41017118646414e-06, "weight_min": 0.0, "weight_std": 1.241742393176537e-05 }, { "epoch": 0.1625831575531397, "step": 501, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9756800611503422e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1277845007716678e-05 }, { "epoch": 0.1625831575531397, "step": 501, "weight_max": 2.5141427613561973e-05, "weight_mean": 3.1435854452865897e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.888470802048687e-06 }, { "epoch": 0.1625831575531397, "step": 501, "weight_max": 3.4413204730299185e-07, "weight_mean": 4.464438418949612e-08, "weight_min": 5.958802340444436e-10, "weight_std": 1.2104446511784772e-07 }, { "epoch": 0.1625831575531397, "step": 501, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8199010810349137e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9758566850214265e-05 }, { "epoch": 0.1629076748336849, "step": 502, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2328565389907453e-05, "weight_min": 0.0, "weight_std": 1.8963002730743028e-05 }, { "epoch": 0.1629076748336849, "step": 502, "weight_max": 3.199401908204891e-05, "weight_mean": 6.025839866197202e-06, "weight_min": 0.0, "weight_std": 1.1927659215871245e-05 }, { "epoch": 0.1629076748336849, "step": 502, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471603673067875e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9389830413274467e-05 }, { "epoch": 0.1629076748336849, "step": 502, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9038350728806108e-05, "weight_min": 0.0, "weight_std": 2.069604852295015e-05 }, { "epoch": 0.1632321921142301, "step": 503, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315810540196253e-06, "weight_min": 0.0, "weight_std": 1.5033256204333156e-05 }, { "epoch": 0.1632321921142301, "step": 503, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.260429578716867e-06, "weight_min": 0.0, "weight_std": 1.729250834614504e-05 }, { "epoch": 0.1632321921142301, "step": 503, "weight_max": 4.187939703115262e-05, "weight_mean": 8.017822437977884e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5743340554763563e-05 }, { "epoch": 0.1632321921142301, "step": 503, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843471742788097e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3697684153157752e-05 }, { "epoch": 0.16355670939477526, "step": 504, "weight_max": 2.836586743271141e-09, "weight_mean": 1.048908959688788e-09, "weight_min": 6.100955851628953e-10, "weight_std": 7.623193187811239e-10 }, { "epoch": 0.16355670939477526, "step": 504, "weight_max": 3.998322426923551e-05, "weight_mean": 1.335377419309225e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8555554561316967e-05 }, { "epoch": 0.16355670939477526, "step": 504, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.288275239465293e-06, "weight_min": 0.0, "weight_std": 1.5382664059870876e-05 }, { "epoch": 0.16355670939477526, "step": 504, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.825138138490729e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.687038275122177e-05 }, { "epoch": 0.16388122667532046, "step": 505, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827621640084544e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0822586773429066e-05 }, { "epoch": 0.16388122667532046, "step": 505, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390811439719982e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.923936724779196e-05 }, { "epoch": 0.16388122667532046, "step": 505, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.20410457183607e-06, "weight_min": 0.0, "weight_std": 1.709588832454756e-05 }, { "epoch": 0.16388122667532046, "step": 505, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2463439816201571e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7787780961953104e-05 }, { "epoch": 0.16420574395586565, "step": 506, "weight_max": 4.561566679939233e-09, "weight_mean": 1.1757926854372158e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.3714391844388274e-09 }, { "epoch": 0.16420574395586565, "step": 506, "weight_max": 3.36925259034615e-05, "weight_mean": 4.2122633203689475e-06, "weight_min": 0.0, "weight_std": 1.1911823094123974e-05 }, { "epoch": 0.16420574395586565, "step": 506, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1246032954659313e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.750909723341465e-05 }, { "epoch": 0.16420574395586565, "step": 506, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3157841648499016e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033267118269578e-05 }, { "epoch": 0.16453026123641085, "step": 507, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236303422861965e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4806045328441542e-05 }, { "epoch": 0.16453026123641085, "step": 507, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6452268027933314e-05, "weight_min": 0.0, "weight_std": 1.80156494025141e-05 }, { "epoch": 0.16453026123641085, "step": 507, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0335736078559421e-05, "weight_min": 0.0, "weight_std": 1.7391752408002503e-05 }, { "epoch": 0.16453026123641085, "step": 507, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.713798706594389e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.6310917999362573e-05 }, { "epoch": 0.16485477851695604, "step": 508, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156129191163927e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4581079994968604e-05 }, { "epoch": 0.16485477851695604, "step": 508, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.610822437214665e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9386790881981142e-05 }, { "epoch": 0.16485477851695604, "step": 508, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882543635176262e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.380638605041895e-05 }, { "epoch": 0.16485477851695604, "step": 508, "weight_max": 4.187939703115262e-05, "weight_mean": 9.926368875312619e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8414362784824334e-05 }, { "epoch": 0.1651792957975012, "step": 509, "weight_max": 3.752332122530788e-05, "weight_mean": 6.02444833930349e-06, "weight_min": 0.0, "weight_std": 1.3241200576885603e-05 }, { "epoch": 0.1651792957975012, "step": 509, "weight_max": 4.381950566312298e-05, "weight_mean": 5.478382718138164e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.549215267004911e-05 }, { "epoch": 0.1651792957975012, "step": 509, "weight_max": 5.153043769468013e-09, "weight_mean": 2.030353440929389e-09, "weight_min": 0.0, "weight_std": 1.8708152804691736e-09 }, { "epoch": 0.1651792957975012, "step": 509, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9946983229601756e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1298892534105107e-05 }, { "epoch": 0.1655038130780464, "grad_norm": 1.7345168590545654, "learning_rate": 9.167346938775511e-06, "loss": 0.2505, "step": 510 }, { "epoch": 0.1655038130780464, "step": 510, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.5071877896843944e-05, "weight_min": 0.0, "weight_std": 1.711692129902076e-05 }, { "epoch": 0.1655038130780464, "step": 510, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.001191594696138e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8516800992074423e-05 }, { "epoch": 0.1655038130780464, "step": 510, "weight_max": 1.661616799708554e-08, "weight_mean": 2.8333579926709263e-09, "weight_min": 5.771870759119224e-10, "weight_std": 5.585071072289338e-09 }, { "epoch": 0.1655038130780464, "step": 510, "weight_max": 4.252122016623616e-05, "weight_mean": 9.641450560593512e-06, "weight_min": 0.0, "weight_std": 1.563468686072156e-05 }, { "epoch": 0.1658283303585916, "step": 511, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.094169409858296e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4350805031426717e-05 }, { "epoch": 0.1658283303585916, "step": 511, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4057428415981121e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.94997301150579e-05 }, { "epoch": 0.1658283303585916, "step": 511, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0178141565120313e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.878777584352065e-05 }, { "epoch": 0.1658283303585916, "step": 511, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.2144828360760584e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7350223060930148e-05 }, { "epoch": 0.1661528476391368, "step": 512, "weight_max": 3.6920544516760856e-05, "weight_mean": 7.225842637126334e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.4047979675524402e-05 }, { "epoch": 0.1661528476391368, "step": 512, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5272022210410796e-05, "weight_min": 0.0, "weight_std": 2.1091551388963126e-05 }, { "epoch": 0.1661528476391368, "step": 512, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156535280548269e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4580915376427583e-05 }, { "epoch": 0.1661528476391368, "step": 512, "weight_max": 3.998322426923551e-05, "weight_mean": 5.410489393398166e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4016999557497911e-05 }, { "epoch": 0.16647736491968199, "step": 513, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4144759916234761e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.967225944099482e-05 }, { "epoch": 0.16647736491968199, "step": 513, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.003804209176451e-05, "weight_min": 0.0, "weight_std": 1.859289659478236e-05 }, { "epoch": 0.16647736491968199, "step": 513, "weight_max": 3.998322426923551e-05, "weight_mean": 9.541736289975233e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7686297724139877e-05 }, { "epoch": 0.16647736491968199, "step": 513, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.1355801689205691e-05, "weight_min": 0.0, "weight_std": 1.6536718248971738e-05 }, { "epoch": 0.16680188220022715, "step": 514, "weight_max": 3.998322426923551e-05, "weight_mean": 2.282440436829347e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9106000763713382e-05 }, { "epoch": 0.16680188220022715, "step": 514, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7932909435767215e-06, "weight_min": 0.0, "weight_std": 1.0726265827543102e-05 }, { "epoch": 0.16680188220022715, "step": 514, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.8160284525947645e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9559714928618632e-05 }, { "epoch": 0.16680188220022715, "step": 514, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.784105714061297e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8092590835294686e-05 }, { "epoch": 0.16712639948077235, "step": 515, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0712426046666224e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9833143596770242e-05 }, { "epoch": 0.16712639948077235, "step": 515, "weight_max": 3.813097282545641e-05, "weight_mean": 8.871778845787048e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6486459571751766e-05 }, { "epoch": 0.16712639948077235, "step": 515, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4025681593921036e-05, "weight_min": 0.0, "weight_std": 1.9636941942735575e-05 }, { "epoch": 0.16712639948077235, "step": 515, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0355803169659339e-05, "weight_min": 0.0, "weight_std": 1.762708052410744e-05 }, { "epoch": 0.16745091676131754, "step": 516, "weight_max": 3.0339444492710754e-05, "weight_mean": 6.813001164118759e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2721582606900483e-05 }, { "epoch": 0.16745091676131754, "step": 516, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0530588042456657e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7495678548584692e-05 }, { "epoch": 0.16745091676131754, "step": 516, "weight_max": 4.187939703115262e-05, "weight_mean": 5.8495811572356615e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.4653101061412599e-05 }, { "epoch": 0.16745091676131754, "step": 516, "weight_max": 2.846461757144425e-05, "weight_mean": 3.559404603947769e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.0063226000056602e-05 }, { "epoch": 0.16777543404186274, "step": 517, "weight_max": 3.843663216684945e-05, "weight_mean": 2.4590994144091383e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.5666269973735325e-05 }, { "epoch": 0.16777543404186274, "step": 517, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3156522881181445e-06, "weight_min": 0.0, "weight_std": 1.5033318959467579e-05 }, { "epoch": 0.16777543404186274, "step": 517, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.5262240746524185e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2491906090872362e-05 }, { "epoch": 0.16777543404186274, "step": 517, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2188970686111134e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7375385141349398e-05 }, { "epoch": 0.16809995132240793, "step": 518, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315906037139939e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5033216186566278e-05 }, { "epoch": 0.16809995132240793, "step": 518, "weight_max": 4.187939703115262e-05, "weight_mean": 1.503919338574633e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0772573407157324e-05 }, { "epoch": 0.16809995132240793, "step": 518, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4128045222605579e-05, "weight_min": 0.0, "weight_std": 1.972798963834066e-05 }, { "epoch": 0.16809995132240793, "step": 518, "weight_max": 4.187939703115262e-05, "weight_mean": 3.0578827136196196e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.888663973659277e-05 }, { "epoch": 0.1684244686029531, "step": 519, "weight_max": 4.252122016623616e-05, "weight_mean": 9.930824489856604e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8447650290909223e-05 }, { "epoch": 0.1684244686029531, "step": 519, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0062864021165296e-05, "weight_min": 0.0, "weight_std": 1.7142250726465136e-05 }, { "epoch": 0.1684244686029531, "step": 519, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.90259026264539e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.3798390682495665e-05 }, { "epoch": 0.1684244686029531, "step": 519, "weight_max": 3.752332122530788e-05, "weight_mean": 8.551738574169576e-06, "weight_min": 0.0, "weight_std": 1.593241177033633e-05 }, { "epoch": 0.1687489858834983, "grad_norm": 1.3970149755477905, "learning_rate": 9.146938775510205e-06, "loss": 0.1757, "step": 520 }, { "epoch": 0.1687489858834983, "step": 520, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3432643754640594e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8835917217074893e-05 }, { "epoch": 0.1687489858834983, "step": 520, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4857063433737494e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.0582838260452263e-05 }, { "epoch": 0.1687489858834983, "step": 520, "weight_max": 1.3161142176087992e-09, "weight_mean": 7.303068150221748e-10, "weight_min": 0.0, "weight_std": 3.6754393972771027e-10 }, { "epoch": 0.1687489858834983, "step": 520, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.817316145519726e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5142756637942512e-05 }, { "epoch": 0.1690735031640435, "step": 521, "weight_max": 1.9527599537383367e-09, "weight_mean": 8.329804623841142e-10, "weight_min": 0.0, "weight_std": 5.514411149931675e-10 }, { "epoch": 0.1690735031640435, "step": 521, "weight_max": 1.3477009588314104e-06, "weight_mean": 1.6941754665822373e-07, "weight_min": 0.0, "weight_std": 4.7610038222956064e-07 }, { "epoch": 0.1690735031640435, "step": 521, "weight_max": 3.312147964606993e-05, "weight_mean": 4.14087708122679e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1709931641235016e-05 }, { "epoch": 0.1690735031640435, "step": 521, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.155772775353398e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.540885386930313e-05 }, { "epoch": 0.16939802044458868, "step": 522, "weight_max": 3.782653220696375e-05, "weight_mean": 4.7293738134612795e-06, "weight_min": 0.0, "weight_std": 1.3373271940508857e-05 }, { "epoch": 0.16939802044458868, "step": 522, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.8393813661532477e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9689470718731172e-05 }, { "epoch": 0.16939802044458868, "step": 522, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9966044419561513e-05, "weight_min": 0.0, "weight_std": 2.1399051547632553e-05 }, { "epoch": 0.16939802044458868, "step": 522, "weight_max": 3.722132169059478e-05, "weight_mean": 7.764785550534725e-06, "weight_min": 0.0, "weight_std": 1.4749428373761475e-05 }, { "epoch": 0.16972253772513388, "step": 523, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.648347258917056e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.786566645023413e-05 }, { "epoch": 0.16972253772513388, "step": 523, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3079404197924305e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8694267055252567e-05 }, { "epoch": 0.16972253772513388, "step": 523, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6670483091729693e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.923486706800759e-05 }, { "epoch": 0.16972253772513388, "step": 523, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.88433021120727e-06, "weight_min": 0.0, "weight_std": 1.83232386916643e-05 }, { "epoch": 0.17004705500567904, "step": 524, "weight_max": 3.967146767536178e-05, "weight_mean": 9.934760782925878e-06, "weight_min": 0.0, "weight_std": 1.6460426195408218e-05 }, { "epoch": 0.17004705500567904, "step": 524, "weight_max": 4.187939703115262e-05, "weight_mean": 2.258101085317321e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.926550794451032e-05 }, { "epoch": 0.17004705500567904, "step": 524, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070077011419926e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1509773685247637e-05 }, { "epoch": 0.17004705500567904, "step": 524, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.181517841876484e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5520656234002672e-05 }, { "epoch": 0.17037157228622424, "step": 525, "weight_max": 4.187939703115262e-05, "weight_mean": 5.26639269082807e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.4794092749070842e-05 }, { "epoch": 0.17037157228622424, "step": 525, "weight_max": 3.967146767536178e-05, "weight_mean": 9.802544809645042e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8150818505091593e-05 }, { "epoch": 0.17037157228622424, "step": 525, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615647412720136e-06, "weight_min": 0.0, "weight_std": 1.3053149814368226e-05 }, { "epoch": 0.17037157228622424, "step": 525, "weight_max": 4.252122016623616e-05, "weight_mean": 8.873868864611723e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.68536753335502e-05 }, { "epoch": 0.17069608956676943, "step": 526, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5081070159794763e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.085087908199057e-05 }, { "epoch": 0.17069608956676943, "step": 526, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.585514817445073e-06, "weight_min": 0.0, "weight_std": 1.7814600141718984e-05 }, { "epoch": 0.17069608956676943, "step": 526, "weight_max": 3.602553260861896e-05, "weight_mean": 1.275168324355036e-05, "weight_min": 0.0, "weight_std": 1.7635909898672253e-05 }, { "epoch": 0.17069608956676943, "step": 526, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1630254347401205e-06, "weight_min": 0.0, "weight_std": 1.4578305126633495e-05 }, { "epoch": 0.17102060684731463, "step": 527, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3681311429536436e-05, "weight_min": 0.0, "weight_std": 1.8980781533173285e-05 }, { "epoch": 0.17102060684731463, "step": 527, "weight_max": 2.349726173633826e-07, "weight_mean": 5.4278320504863586e-08, "weight_min": 5.771870759119224e-10, "weight_std": 9.687126123480994e-08 }, { "epoch": 0.17102060684731463, "step": 527, "weight_max": 3.602553260861896e-05, "weight_mean": 1.1120895578642376e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.5931571397231892e-05 }, { "epoch": 0.17102060684731463, "step": 527, "weight_max": 4.187939703115262e-05, "weight_mean": 9.775890248420183e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.816069197957404e-05 }, { "epoch": 0.17134512412785982, "step": 528, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.729571502248291e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.803692794055678e-05 }, { "epoch": 0.17134512412785982, "step": 528, "weight_max": 1.0015337448265882e-09, "weight_mean": 7.26756432811726e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.5346274251371028e-10 }, { "epoch": 0.17134512412785982, "step": 528, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4299804358161055e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2527903891168535e-05 }, { "epoch": 0.17134512412785982, "step": 528, "weight_max": 4.187939703115262e-05, "weight_mean": 2.550261706346646e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1129466404090635e-05 }, { "epoch": 0.171669641408405, "step": 529, "weight_max": 3.998322426923551e-05, "weight_mean": 9.803075954550877e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8155009456677362e-05 }, { "epoch": 0.171669641408405, "step": 529, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4768286746402737e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0414583559613675e-05 }, { "epoch": 0.171669641408405, "step": 529, "weight_max": 3.061215829802677e-05, "weight_mean": 3.828524768323405e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.0822222975548357e-05 }, { "epoch": 0.171669641408405, "step": 529, "weight_max": 3.967146767536178e-05, "weight_mean": 4.960059413861018e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4025526979821734e-05 }, { "epoch": 0.17199415868895018, "grad_norm": 0.5647338628768921, "learning_rate": 9.126530612244898e-06, "loss": 0.3536, "step": 530 }, { "epoch": 0.17199415868895018, "step": 530, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805154731002403e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3589169611805119e-05 }, { "epoch": 0.17199415868895018, "step": 530, "weight_max": 3.998322426923551e-05, "weight_mean": 9.957460861187428e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.843628160713706e-05 }, { "epoch": 0.17199415868895018, "step": 530, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0000858057755977e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1251080397632904e-05 }, { "epoch": 0.17199415868895018, "step": 530, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155608833883889e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8815173461916856e-05 }, { "epoch": 0.17231867596949538, "step": 531, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4995098354120273e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0863593817921355e-05 }, { "epoch": 0.17231867596949538, "step": 531, "weight_max": 2.626842672270868e-07, "weight_mean": 3.353248345661086e-08, "weight_min": 5.864964625068581e-10, "weight_std": 9.259200339784002e-08 }, { "epoch": 0.17231867596949538, "step": 531, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5946538042044267e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.2009966414771043e-05 }, { "epoch": 0.17231867596949538, "step": 531, "weight_max": 3.632264997577295e-05, "weight_mean": 5.398086159402737e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2723496183753014e-05 }, { "epoch": 0.17264319325004057, "step": 532, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0315827239537612e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9106402760371566e-05 }, { "epoch": 0.17264319325004057, "step": 532, "weight_max": 3.998322426923551e-05, "weight_mean": 8.329609045176767e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5828543837415054e-05 }, { "epoch": 0.17264319325004057, "step": 532, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5314415577449836e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1163175915717147e-05 }, { "epoch": 0.17264319325004057, "step": 532, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5408860387397e-06, "weight_min": 0.0, "weight_std": 1.2841771422245074e-05 }, { "epoch": 0.17296771053058577, "step": 533, "weight_max": 4.252122016623616e-05, "weight_mean": 9.7453248599777e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8141980035579763e-05 }, { "epoch": 0.17296771053058577, "step": 533, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.301396514463704e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.819492354115937e-05 }, { "epoch": 0.17296771053058577, "step": 533, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0258308066113386e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.829180655477103e-05 }, { "epoch": 0.17296771053058577, "step": 533, "weight_max": 1.178765705844853e-05, "weight_mean": 1.5186571999947773e-06, "weight_min": 0.0, "weight_std": 4.150381755607668e-06 }, { "epoch": 0.17329222781113093, "step": 534, "weight_max": 3.998322426923551e-05, "weight_mean": 8.756796887610108e-06, "weight_min": 0.0, "weight_std": 1.6428579328930937e-05 }, { "epoch": 0.17329222781113093, "step": 534, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3159378542914055e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9535516912583262e-05 }, { "epoch": 0.17329222781113093, "step": 534, "weight_max": 3.662098970380612e-05, "weight_mean": 4.79658092444879e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2873498235421721e-05 }, { "epoch": 0.17329222781113093, "step": 534, "weight_max": 4.006134979306353e-09, "weight_mean": 1.070712518647099e-09, "weight_min": 0.0, "weight_std": 1.2432037621579184e-09 }, { "epoch": 0.17361674509167613, "step": 535, "weight_max": 4.252122016623616e-05, "weight_mean": 5.31591149410815e-06, "weight_min": 0.0, "weight_std": 1.5033215277071577e-05 }, { "epoch": 0.17361674509167613, "step": 535, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235829576122342e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806237231823616e-05 }, { "epoch": 0.17361674509167613, "step": 535, "weight_max": 3.632264997577295e-05, "weight_mean": 8.033473932300694e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5041283404571004e-05 }, { "epoch": 0.17361674509167613, "step": 535, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.044059627020033e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8386037481832318e-05 }, { "epoch": 0.17394126237222132, "step": 536, "weight_max": 3.340639523230493e-05, "weight_mean": 4.1810858419921715e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1808815543190576e-05 }, { "epoch": 0.17394126237222132, "step": 536, "weight_max": 5.309356652105635e-07, "weight_mean": 6.758884296687029e-08, "weight_min": 0.0, "weight_std": 1.872257655577414e-07 }, { "epoch": 0.17394126237222132, "step": 536, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355484846513718e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.919600435940083e-05 }, { "epoch": 0.17394126237222132, "step": 536, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0789853149617556e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8664975868887268e-05 }, { "epoch": 0.17426577965276652, "step": 537, "weight_max": 1.1019851697824379e-08, "weight_mean": 1.969365559517655e-09, "weight_min": 5.864964625068581e-10, "weight_std": 3.6583998053174582e-09 }, { "epoch": 0.17426577965276652, "step": 537, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0075445970869623e-05, "weight_min": 0.0, "weight_std": 1.8653785446076654e-05 }, { "epoch": 0.17426577965276652, "step": 537, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4541063137585297e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.011918149946723e-05 }, { "epoch": 0.17426577965276652, "step": 537, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.3804150259820744e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.975585473701358e-05 }, { "epoch": 0.1745902969333117, "step": 538, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.905128536047414e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.593568413227331e-05 }, { "epoch": 0.1745902969333117, "step": 538, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998723852622788e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4135872334009036e-05 }, { "epoch": 0.1745902969333117, "step": 538, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.063555469561834e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.540503762953449e-05 }, { "epoch": 0.1745902969333117, "step": 538, "weight_max": 5.3206110628423176e-09, "weight_mean": 2.0338442041634153e-09, "weight_min": 8.267048712262692e-10, "weight_std": 1.4939935955027295e-09 }, { "epoch": 0.17491481421385688, "step": 539, "weight_max": 3.340639523230493e-05, "weight_mean": 1.5932824680930935e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7054748241207562e-05 }, { "epoch": 0.17491481421385688, "step": 539, "weight_max": 1.5802011921195458e-09, "weight_mean": 7.236250487707707e-10, "weight_min": 0.0, "weight_std": 4.48955872478507e-10 }, { "epoch": 0.17491481421385688, "step": 539, "weight_max": 4.187939703115262e-05, "weight_mean": 5.329255600372562e-06, "weight_min": 0.0, "weight_std": 1.4769908375455998e-05 }, { "epoch": 0.17491481421385688, "step": 539, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4377204024640378e-05, "weight_min": 0.0, "weight_std": 1.984807022381574e-05 }, { "epoch": 0.17523933149440207, "grad_norm": 14.282212257385254, "learning_rate": 9.106122448979593e-06, "loss": 0.3293, "step": 540 }, { "epoch": 0.17523933149440207, "step": 540, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.403528120950796e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5259131032507867e-05 }, { "epoch": 0.17523933149440207, "step": 540, "weight_max": 3.340639523230493e-05, "weight_mean": 7.90099238656694e-06, "weight_min": 0.0, "weight_std": 1.4660362467111554e-05 }, { "epoch": 0.17523933149440207, "step": 540, "weight_max": 3.722132169059478e-05, "weight_mean": 4.656296368921176e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.315825829806272e-05 }, { "epoch": 0.17523933149440207, "step": 540, "weight_max": 4.252122016623616e-05, "weight_mean": 8.973429430625401e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.698767846392002e-05 }, { "epoch": 0.17556384877494727, "step": 541, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.370608156837989e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5140008144953754e-05 }, { "epoch": 0.17556384877494727, "step": 541, "weight_max": 3.752332122530788e-05, "weight_mean": 1.2378347491903696e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7209908037330024e-05 }, { "epoch": 0.17556384877494727, "step": 541, "weight_max": 4.187939703115262e-05, "weight_mean": 9.664834578870796e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7977161405724473e-05 }, { "epoch": 0.17556384877494727, "step": 541, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155994131200714e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581133655156009e-05 }, { "epoch": 0.17588836605549246, "step": 542, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2044872164551634e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.816093390516471e-05 }, { "epoch": 0.17588836605549246, "step": 542, "weight_max": 3.5434961318969727e-05, "weight_mean": 7.857714081183076e-06, "weight_min": 0.0, "weight_std": 1.4704472960147541e-05 }, { "epoch": 0.17588836605549246, "step": 542, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4766214917472098e-05, "weight_min": 0.0, "weight_std": 2.0402709196787328e-05 }, { "epoch": 0.17588836605549246, "step": 542, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1856034689117223e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7380551071255468e-05 }, { "epoch": 0.17621288333603766, "step": 543, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3709329323319253e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.89724778465461e-05 }, { "epoch": 0.17621288333603766, "step": 543, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965132691519102e-06, "weight_min": 0.0, "weight_std": 1.1212723620701581e-05 }, { "epoch": 0.17621288333603766, "step": 543, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.49660716287326e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7586526155355386e-05 }, { "epoch": 0.17621288333603766, "step": 543, "weight_max": 3.967146767536178e-05, "weight_mean": 9.462786692893133e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.754711956891697e-05 }, { "epoch": 0.17653740061658282, "step": 544, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0117213605553843e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8730806914390996e-05 }, { "epoch": 0.17653740061658282, "step": 544, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4297362213255838e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0034098270116374e-05 }, { "epoch": 0.17653740061658282, "step": 544, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.085391866392456e-06, "weight_min": 0.0, "weight_std": 1.7024811313604005e-05 }, { "epoch": 0.17653740061658282, "step": 544, "weight_max": 3.632264997577295e-05, "weight_mean": 7.562836799479555e-06, "weight_min": 0.0, "weight_std": 1.4370047210832126e-05 }, { "epoch": 0.17686191789712802, "step": 545, "weight_max": 3.752332122530788e-05, "weight_mean": 8.802649972494692e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6328895071637817e-05 }, { "epoch": 0.17686191789712802, "step": 545, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6896188753889874e-05, "weight_min": 0.0, "weight_std": 1.9530189092620276e-05 }, { "epoch": 0.17686191789712802, "step": 545, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.026328573061619e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.599476127012167e-05 }, { "epoch": 0.17686191789712802, "step": 545, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.44085513765458e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7575266610947438e-05 }, { "epoch": 0.1771864351776732, "step": 546, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4808005289523862e-05, "weight_min": 0.0, "weight_std": 2.047429370577447e-05 }, { "epoch": 0.1771864351776732, "step": 546, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.8032394766341895e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7501719412393868e-05 }, { "epoch": 0.1771864351776732, "step": 546, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.920102456817403e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8373104467173107e-05 }, { "epoch": 0.1771864351776732, "step": 546, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.747122592467349e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.636324122955557e-05 }, { "epoch": 0.1775109524582184, "step": 547, "weight_max": 3.967146767536178e-05, "weight_mean": 9.691030754765961e-06, "weight_min": 8.378387983398738e-10, "weight_std": 1.7941745682037435e-05 }, { "epoch": 0.1775109524582184, "step": 547, "weight_max": 4.187939703115262e-05, "weight_mean": 6.709241915814346e-06, "weight_min": 0.0, "weight_std": 1.47973032653681e-05 }, { "epoch": 0.1775109524582184, "step": 547, "weight_max": 4.252122016623616e-05, "weight_mean": 5.349644652596908e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5019868442323059e-05 }, { "epoch": 0.1775109524582184, "step": 547, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.02176270756172e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7125563317677006e-05 }, { "epoch": 0.1778354697387636, "step": 548, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844834620598704e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3697131180379074e-05 }, { "epoch": 0.1778354697387636, "step": 548, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.337509487115312e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4852910680929199e-05 }, { "epoch": 0.1778354697387636, "step": 548, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315631824487355e-06, "weight_min": 0.0, "weight_std": 1.5033327144919895e-05 }, { "epoch": 0.1778354697387636, "step": 548, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0078670129587408e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.867888931883499e-05 }, { "epoch": 0.17815998701930877, "step": 549, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0158889381273184e-05, "weight_min": 0.0, "weight_std": 1.88357153092511e-05 }, { "epoch": 0.17815998701930877, "step": 549, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9993564061878715e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.413561767549254e-05 }, { "epoch": 0.17815998701930877, "step": 549, "weight_max": 3.782653220696375e-05, "weight_mean": 9.269628208130598e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.716630140435882e-05 }, { "epoch": 0.17815998701930877, "step": 549, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.081495146441739e-06, "weight_min": 0.0, "weight_std": 1.4355849998537451e-05 }, { "epoch": 0.17848450429985396, "grad_norm": 0.40499064326286316, "learning_rate": 9.085714285714286e-06, "loss": 0.15, "step": 550 }, { "epoch": 0.17848450429985396, "step": 550, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.971981747483369e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.474268992751604e-05 }, { "epoch": 0.17848450429985396, "step": 550, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430499302543467e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2527694707387127e-05 }, { "epoch": 0.17848450429985396, "step": 550, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.035682882706169e-05, "weight_min": 7.721521200210191e-10, "weight_std": 2.1810696125612594e-05 }, { "epoch": 0.17848450429985396, "step": 550, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.498080205754377e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5737874491605908e-05 }, { "epoch": 0.17880902158039916, "step": 551, "weight_max": 3.632264997577295e-05, "weight_mean": 8.970244380179793e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.6610018064966425e-05 }, { "epoch": 0.17880902158039916, "step": 551, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2742065337079111e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7741360352374613e-05 }, { "epoch": 0.17880902158039916, "step": 551, "weight_max": 4.252122016623616e-05, "weight_mean": 5.347070327843539e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5020870705484413e-05 }, { "epoch": 0.17880902158039916, "step": 551, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0037076208391227e-05, "weight_min": 0.0, "weight_std": 1.859349140431732e-05 }, { "epoch": 0.17913353886094435, "step": 552, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2319023880991153e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7895747078000568e-05 }, { "epoch": 0.17913353886094435, "step": 552, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.7251663798087975e-06, "weight_min": 0.0, "weight_std": 1.0534886314417236e-05 }, { "epoch": 0.17913353886094435, "step": 552, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.080300979898311e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.435633203072939e-05 }, { "epoch": 0.17913353886094435, "step": 552, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4515873772324994e-05, "weight_min": 0.0, "weight_std": 2.029918323387392e-05 }, { "epoch": 0.17945805614148955, "step": 553, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.937165486917365e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.848870851972606e-05 }, { "epoch": 0.17945805614148955, "step": 553, "weight_max": 4.252122016623616e-05, "weight_mean": 9.563254934619181e-06, "weight_min": 0.0, "weight_std": 1.7852729797596112e-05 }, { "epoch": 0.17945805614148955, "step": 553, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5165525837801397e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.099360244756099e-05 }, { "epoch": 0.17945805614148955, "step": 553, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9445900761638768e-05, "weight_min": 0.0, "weight_std": 2.093881084874738e-05 }, { "epoch": 0.1797825734220347, "step": 554, "weight_max": 4.187939703115262e-05, "weight_mean": 5.238595804257784e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.480512128182454e-05 }, { "epoch": 0.1797825734220347, "step": 554, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4672022846061736e-05, "weight_min": 1.0509849657225345e-09, "weight_std": 1.76154662767658e-05 }, { "epoch": 0.1797825734220347, "step": 554, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4077469131734688e-05, "weight_min": 0.0, "weight_std": 1.961471571121365e-05 }, { "epoch": 0.1797825734220347, "step": 554, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2826588317693677e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.821998375817202e-05 }, { "epoch": 0.1801070907025799, "step": 555, "weight_max": 3.998322426923551e-05, "weight_mean": 9.765009963302873e-06, "weight_min": 0.0, "weight_std": 1.808624620025512e-05 }, { "epoch": 0.1801070907025799, "step": 555, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4443876352743246e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.005822716455441e-05 }, { "epoch": 0.1801070907025799, "step": 555, "weight_max": 4.252122016623616e-05, "weight_mean": 9.968375707103405e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8510743757360615e-05 }, { "epoch": 0.1801070907025799, "step": 555, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4151204595691524e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.970668563444633e-05 }, { "epoch": 0.1804316079831251, "step": 556, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2357941058289725e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806251783738844e-05 }, { "epoch": 0.1804316079831251, "step": 556, "weight_max": 3.340639523230493e-05, "weight_mean": 7.900986020104028e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4660366105090361e-05 }, { "epoch": 0.1804316079831251, "step": 556, "weight_max": 5.153043769468013e-09, "weight_mean": 1.2159845352854859e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.5918183438046185e-09 }, { "epoch": 0.1804316079831251, "step": 556, "weight_max": 4.252122016623616e-05, "weight_mean": 9.708375728223473e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.808297929528635e-05 }, { "epoch": 0.1807561252636703, "step": 557, "weight_max": 3.752332122530788e-05, "weight_mean": 8.517473361280281e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5877771147643216e-05 }, { "epoch": 0.1807561252636703, "step": 557, "weight_max": 4.187939703115262e-05, "weight_mean": 2.056842640740797e-05, "weight_min": 0.0, "weight_std": 1.7827738702180795e-05 }, { "epoch": 0.1807561252636703, "step": 557, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396629603637848e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5261906810337678e-05 }, { "epoch": 0.1807561252636703, "step": 557, "weight_max": 4.381950566312298e-05, "weight_mean": 1.9074688680120744e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0749183022417128e-05 }, { "epoch": 0.1810806425442155, "step": 558, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.076322385226376e-06, "weight_min": 0.0, "weight_std": 1.696123945293948e-05 }, { "epoch": 0.1810806425442155, "step": 558, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.1213540902826935e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4445931810769252e-05 }, { "epoch": 0.1810806425442155, "step": 558, "weight_max": 2.846461757144425e-05, "weight_mean": 3.558733169484185e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0063496119983029e-05 }, { "epoch": 0.1810806425442155, "step": 558, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0156554455752485e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1580006432486698e-05 }, { "epoch": 0.18140515982476066, "step": 559, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0082114386023022e-05, "weight_min": 0.0, "weight_std": 1.870383130153641e-05 }, { "epoch": 0.18140515982476066, "step": 559, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.998724635806866e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8524902770877816e-05 }, { "epoch": 0.18140515982476066, "step": 559, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5310748494812287e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1141546312719584e-05 }, { "epoch": 0.18140515982476066, "step": 559, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.6157947508618236e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.305309069721261e-05 }, { "epoch": 0.18172967710530585, "grad_norm": 7.001326084136963, "learning_rate": 9.065306122448981e-06, "loss": 0.1773, "step": 560 }, { "epoch": 0.18172967710530585, "step": 560, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.0816462286748e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5023168089101091e-05 }, { "epoch": 0.18172967710530585, "step": 560, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1467674994491972e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7221731468453072e-05 }, { "epoch": 0.18172967710530585, "step": 560, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.47611260926351e-06, "weight_min": 0.0, "weight_std": 1.763454383763019e-05 }, { "epoch": 0.18172967710530585, "step": 560, "weight_max": 1.56184469233267e-05, "weight_mean": 1.953534138010582e-06, "weight_min": 0.0, "weight_std": 5.521459115698235e-06 }, { "epoch": 0.18205419438585105, "step": 561, "weight_max": 4.252122016623616e-05, "weight_mean": 9.893270544125699e-06, "weight_min": 0.0, "weight_std": 1.8385308067081496e-05 }, { "epoch": 0.18205419438585105, "step": 561, "weight_max": 4.187939703115262e-05, "weight_mean": 9.977703484764788e-06, "weight_min": 0.0, "weight_std": 1.8471238945494406e-05 }, { "epoch": 0.18205419438585105, "step": 561, "weight_max": 3.572963396436535e-05, "weight_mean": 8.30877797852736e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.54061199282296e-05 }, { "epoch": 0.18205419438585105, "step": 561, "weight_max": 1.8022843217835316e-09, "weight_mean": 9.032565806421644e-10, "weight_min": 5.864964625068581e-10, "weight_std": 4.209220194617558e-10 }, { "epoch": 0.18237871166639624, "step": 562, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.938236194604542e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5548013834632002e-05 }, { "epoch": 0.18237871166639624, "step": 562, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.504174729750957e-06, "weight_min": 0.0, "weight_std": 1.6135589248733595e-05 }, { "epoch": 0.18237871166639624, "step": 562, "weight_max": 3.199401908204891e-05, "weight_mean": 3.999924501840724e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1311322850815486e-05 }, { "epoch": 0.18237871166639624, "step": 562, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.15623833052814e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581035429728217e-05 }, { "epoch": 0.18270322894694144, "step": 563, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.604387403465807e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.59298306243727e-05 }, { "epoch": 0.18270322894694144, "step": 563, "weight_max": 4.252122016623616e-05, "weight_mean": 7.678575457248371e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5554129277006723e-05 }, { "epoch": 0.18270322894694144, "step": 563, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.84426345792599e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3697364011022728e-05 }, { "epoch": 0.18270322894694144, "step": 563, "weight_max": 3.116124207735993e-05, "weight_mean": 3.89925844501704e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1015507880074438e-05 }, { "epoch": 0.1830277462274866, "step": 564, "weight_max": 3.967146767536178e-05, "weight_mean": 1.8946589989354834e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8682461814023554e-05 }, { "epoch": 0.1830277462274866, "step": 564, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7965654478757642e-05, "weight_min": 0.0, "weight_std": 1.947419877978973e-05 }, { "epoch": 0.1830277462274866, "step": 564, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5089935914147645e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7698808733257465e-05 }, { "epoch": 0.1830277462274866, "step": 564, "weight_max": 3.752332122530788e-05, "weight_mean": 1.2774673450621776e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7765642041922547e-05 }, { "epoch": 0.1833522635080318, "step": 565, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397434961196268e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5261581211234443e-05 }, { "epoch": 0.1833522635080318, "step": 565, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9985674195340835e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.413593599863816e-05 }, { "epoch": 0.1833522635080318, "step": 565, "weight_max": 1.0261104188558079e-09, "weight_mean": 6.220755022212643e-10, "weight_min": 0.0, "weight_std": 3.1415786638788745e-10 }, { "epoch": 0.1833522635080318, "step": 565, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.324260645371396e-06, "weight_min": 0.0, "weight_std": 1.7354714145767502e-05 }, { "epoch": 0.183676780788577, "step": 566, "weight_max": 3.722132169059478e-05, "weight_mean": 9.193647201755084e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.702341432974208e-05 }, { "epoch": 0.183676780788577, "step": 566, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5273206372512504e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1092893803142942e-05 }, { "epoch": 0.183676780788577, "step": 566, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729904958367115e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3373056390264537e-05 }, { "epoch": 0.183676780788577, "step": 566, "weight_max": 3.752332122530788e-05, "weight_mean": 8.349103154614568e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5613792129443027e-05 }, { "epoch": 0.1840012980691222, "step": 567, "weight_max": 3.36925259034615e-05, "weight_mean": 7.3575392889324576e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.380585217702901e-05 }, { "epoch": 0.1840012980691222, "step": 567, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.283705028588884e-06, "weight_min": 0.0, "weight_std": 1.56944715854479e-05 }, { "epoch": 0.1840012980691222, "step": 567, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1817764971056022e-05, "weight_min": 0.0, "weight_std": 1.8363392882747576e-05 }, { "epoch": 0.1840012980691222, "step": 567, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4045794159756042e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.960297413461376e-05 }, { "epoch": 0.18432581534966738, "step": 568, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0985447918064892e-05, "weight_min": 0.0, "weight_std": 2.2457232262240723e-05 }, { "epoch": 0.18432581534966738, "step": 568, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.452011929359287e-05, "weight_min": 0.0, "weight_std": 1.8927852579508908e-05 }, { "epoch": 0.18432581534966738, "step": 568, "weight_max": 3.967146767536178e-05, "weight_mean": 9.774909813131671e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.536814852443058e-05 }, { "epoch": 0.18432581534966738, "step": 568, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390770512458403e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.923939271364361e-05 }, { "epoch": 0.18465033263021255, "step": 569, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1683388695237227e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8336895664106123e-05 }, { "epoch": 0.18465033263021255, "step": 569, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311905498383567e-05, "weight_min": 0.0, "weight_std": 1.9095270545221865e-05 }, { "epoch": 0.18465033263021255, "step": 569, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4595710126741324e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.042033156612888e-05 }, { "epoch": 0.18465033263021255, "step": 569, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.696104825707152e-06, "weight_min": 0.0, "weight_std": 1.800063364498783e-05 }, { "epoch": 0.18497484991075774, "grad_norm": 0.2829243838787079, "learning_rate": 9.044897959183674e-06, "loss": 0.4003, "step": 570 }, { "epoch": 0.18497484991075774, "step": 570, "weight_max": 2.0612175077872052e-08, "weight_mean": 3.4937446269367456e-09, "weight_min": 5.771870759119224e-10, "weight_std": 6.942315611269123e-09 }, { "epoch": 0.18497484991075774, "step": 570, "weight_max": 3.998322426923551e-05, "weight_mean": 5.2804775805270765e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.404411523253657e-05 }, { "epoch": 0.18497484991075774, "step": 570, "weight_max": 3.662098970380612e-05, "weight_mean": 1.0965045476041269e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.5773330233059824e-05 }, { "epoch": 0.18497484991075774, "step": 570, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.404436170560075e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.988214353332296e-05 }, { "epoch": 0.18529936719130294, "step": 571, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0035810191766359e-05, "weight_min": 0.0, "weight_std": 1.858298128354363e-05 }, { "epoch": 0.18529936719130294, "step": 571, "weight_max": 1.6811793102533557e-05, "weight_mean": 2.1019923224230297e-06, "weight_min": 0.0, "weight_std": 5.943657015450299e-06 }, { "epoch": 0.18529936719130294, "step": 571, "weight_max": 1.504423474685268e-09, "weight_mean": 9.48647604914754e-10, "weight_min": 6.390288853630466e-10, "weight_std": 2.52544984924441e-10 }, { "epoch": 0.18529936719130294, "step": 571, "weight_max": 4.252122016623616e-05, "weight_mean": 1.524095205240883e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.108365879394114e-05 }, { "epoch": 0.18562388447184813, "step": 572, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5315719792852178e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1169937099330127e-05 }, { "epoch": 0.18562388447184813, "step": 572, "weight_max": 4.252122016623616e-05, "weight_mean": 5.535508535103872e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4956932318455074e-05 }, { "epoch": 0.18562388447184813, "step": 572, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.0728519302792847e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.873666224128101e-05 }, { "epoch": 0.18562388447184813, "step": 572, "weight_max": 1.1639718877631822e-07, "weight_mean": 1.5233455386010064e-08, "weight_min": 6.196655966128617e-10, "weight_std": 4.087657856644e-08 }, { "epoch": 0.18594840175239333, "step": 573, "weight_max": 3.998322426923551e-05, "weight_mean": 9.689413673186209e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7950645997188985e-05 }, { "epoch": 0.18594840175239333, "step": 573, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4608122000936419e-05, "weight_min": 0.0, "weight_std": 2.0168416085653007e-05 }, { "epoch": 0.18594840175239333, "step": 573, "weight_max": 3.782653220696375e-05, "weight_mean": 4.72886631541769e-06, "weight_min": 0.0, "weight_std": 1.3373477486311458e-05 }, { "epoch": 0.18594840175239333, "step": 573, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4751920389244333e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8643464500200935e-05 }, { "epoch": 0.1862729190329385, "step": 574, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.802001841308083e-06, "weight_min": 0.0, "weight_std": 1.814889401430264e-05 }, { "epoch": 0.1862729190329385, "step": 574, "weight_max": 3.843663216684945e-05, "weight_mean": 1.561801036586985e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7499762179795653e-05 }, { "epoch": 0.1862729190329385, "step": 574, "weight_max": 4.252122016623616e-05, "weight_mean": 9.211078577209264e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.732185046421364e-05 }, { "epoch": 0.1862729190329385, "step": 574, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.49986517499201e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0717279767268337e-05 }, { "epoch": 0.1865974363134837, "step": 575, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3225180737208575e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.84590771823423e-05 }, { "epoch": 0.1865974363134837, "step": 575, "weight_max": 4.187939703115262e-05, "weight_mean": 9.246429726772476e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7293839846388437e-05 }, { "epoch": 0.1865974363134837, "step": 575, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.996748536243103e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.561044882691931e-05 }, { "epoch": 0.1865974363134837, "step": 575, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4286182704381645e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.982166759262327e-05 }, { "epoch": 0.18692195359402888, "step": 576, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.86954330781009e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6649051758577116e-05 }, { "epoch": 0.18692195359402888, "step": 576, "weight_max": 4.187939703115262e-05, "weight_mean": 6.2008075474295765e-06, "weight_min": 0.0, "weight_std": 1.4667290088254958e-05 }, { "epoch": 0.18692195359402888, "step": 576, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0082085282192566e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8703849491430447e-05 }, { "epoch": 0.18692195359402888, "step": 576, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728933618025621e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3373449291975703e-05 }, { "epoch": 0.18724647087457408, "step": 577, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3183514965930954e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8662613001652062e-05 }, { "epoch": 0.18724647087457408, "step": 577, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.686214980320074e-06, "weight_min": 0.0, "weight_std": 1.7934547940967605e-05 }, { "epoch": 0.18724647087457408, "step": 577, "weight_max": 4.187939703115262e-05, "weight_mean": 6.8835820457024965e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4769316294405144e-05 }, { "epoch": 0.18724647087457408, "step": 577, "weight_max": 3.0886087188264355e-05, "weight_mean": 1.2844055163441226e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.4089703654462937e-05 }, { "epoch": 0.18757098815511927, "step": 578, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.2327219994622283e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0989184122299775e-05 }, { "epoch": 0.18757098815511927, "step": 578, "weight_max": 3.9360926166409627e-05, "weight_mean": 6.037101684341906e-06, "weight_min": 0.0, "weight_std": 1.3822846995026339e-05 }, { "epoch": 0.18757098815511927, "step": 578, "weight_max": 4.187939703115262e-05, "weight_mean": 9.067368409887422e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.612135201867204e-05 }, { "epoch": 0.18757098815511927, "step": 578, "weight_max": 3.045860896833119e-09, "weight_mean": 9.565954695034407e-10, "weight_min": 5.958802340444436e-10, "weight_std": 8.480630087071006e-10 }, { "epoch": 0.18789550543566444, "step": 579, "weight_max": 4.187939703115262e-05, "weight_mean": 8.770530257606879e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6558391507714987e-05 }, { "epoch": 0.18789550543566444, "step": 579, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0175030183745548e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6922187569434755e-05 }, { "epoch": 0.18789550543566444, "step": 579, "weight_max": 3.782653220696375e-05, "weight_mean": 1.380948378937319e-05, "weight_min": 0.0, "weight_std": 1.906443139887415e-05 }, { "epoch": 0.18789550543566444, "step": 579, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5432337022502907e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.1319707229849882e-05 }, { "epoch": 0.18822002271620963, "grad_norm": 0.30080103874206543, "learning_rate": 9.02448979591837e-06, "loss": 0.2479, "step": 580 }, { "epoch": 0.18822002271620963, "step": 580, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.360398507851642e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7413442037650384e-05 }, { "epoch": 0.18822002271620963, "step": 580, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.928174424392637e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.086627318931278e-05 }, { "epoch": 0.18822002271620963, "step": 580, "weight_max": 1.0888551171817085e-09, "weight_mean": 6.978277955482781e-10, "weight_min": 0.0, "weight_std": 3.2335883970446844e-10 }, { "epoch": 0.18822002271620963, "step": 580, "weight_max": 3.199401908204891e-05, "weight_mean": 7.363151325989747e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3699738701689057e-05 }, { "epoch": 0.18854453999675483, "step": 581, "weight_max": 5.874315434084565e-08, "weight_mean": 8.551614527618767e-09, "weight_min": 5.864964625068581e-10, "weight_std": 2.0299347980312632e-08 }, { "epoch": 0.18854453999675483, "step": 581, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4790492059546523e-05, "weight_min": 0.0, "weight_std": 2.0539851902867667e-05 }, { "epoch": 0.18854453999675483, "step": 581, "weight_max": 4.187939703115262e-05, "weight_mean": 9.665366633271333e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.797683216864243e-05 }, { "epoch": 0.18854453999675483, "step": 581, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.358195731882006e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.80554579856107e-05 }, { "epoch": 0.18886905727730002, "step": 582, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.958341252058744e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8442515283823013e-05 }, { "epoch": 0.18886905727730002, "step": 582, "weight_max": 3.998322426923551e-05, "weight_mean": 1.0306252079317346e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8336979337618686e-05 }, { "epoch": 0.18886905727730002, "step": 582, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3155778106010985e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.879982119135093e-05 }, { "epoch": 0.18886905727730002, "step": 582, "weight_max": 1.504423474685268e-09, "weight_mean": 7.283286196368977e-10, "weight_min": 0.0, "weight_std": 4.163215605590409e-10 }, { "epoch": 0.18919357455784522, "step": 583, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5173947758739814e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7829432181315497e-05 }, { "epoch": 0.18919357455784522, "step": 583, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.674715895205736e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9620094462879933e-05 }, { "epoch": 0.18919357455784522, "step": 583, "weight_max": 3.998322426923551e-05, "weight_mean": 9.688799764262512e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7951026165974326e-05 }, { "epoch": 0.18919357455784522, "step": 583, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3447387573251035e-05, "weight_min": 0.0, "weight_std": 1.8911779989139177e-05 }, { "epoch": 0.18951809183839038, "step": 584, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.8433394870953634e-05, "weight_min": 0.0, "weight_std": 1.9897308447980322e-05 }, { "epoch": 0.18951809183839038, "step": 584, "weight_max": 4.252122016623616e-05, "weight_mean": 9.599293662176933e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7909334928845055e-05 }, { "epoch": 0.18951809183839038, "step": 584, "weight_max": 3.813097282545641e-05, "weight_mean": 9.30724127101712e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7239108274225146e-05 }, { "epoch": 0.18951809183839038, "step": 584, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4597129595349543e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.0518336896202527e-05 }, { "epoch": 0.18984260911893558, "step": 585, "weight_max": 3.752332122530788e-05, "weight_mean": 7.897142495494336e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4959243344492279e-05 }, { "epoch": 0.18984260911893558, "step": 585, "weight_max": 4.252122016623616e-05, "weight_mean": 9.594825314707123e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7529640899738297e-05 }, { "epoch": 0.18984260911893558, "step": 585, "weight_max": 4.252122016623616e-05, "weight_mean": 8.743229045649059e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6683461581123993e-05 }, { "epoch": 0.18984260911893558, "step": 585, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8093007383868098e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9664888895931654e-05 }, { "epoch": 0.19016712639948077, "step": 586, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.3942286538367625e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.242377493326785e-05 }, { "epoch": 0.19016712639948077, "step": 586, "weight_max": 3.967146767536178e-05, "weight_mean": 9.099948329094332e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6938483895501122e-05 }, { "epoch": 0.19016712639948077, "step": 586, "weight_max": 3.397988257347606e-05, "weight_mean": 7.93985145719489e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4746633496542927e-05 }, { "epoch": 0.19016712639948077, "step": 586, "weight_max": 2.820166446326766e-05, "weight_mean": 6.891295925015584e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.268120377062587e-05 }, { "epoch": 0.19049164368002597, "step": 587, "weight_max": 3.752332122530788e-05, "weight_mean": 7.02644138073083e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3949556887382641e-05 }, { "epoch": 0.19049164368002597, "step": 587, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.8430362615617923e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.973304824787192e-05 }, { "epoch": 0.19049164368002597, "step": 587, "weight_max": 9.995806067308877e-06, "weight_mean": 1.2501834589784266e-06, "weight_min": 6.196655966128617e-10, "weight_std": 3.5337652661837637e-06 }, { "epoch": 0.19049164368002597, "step": 587, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.93754531716695e-06, "weight_min": 0.0, "weight_std": 1.6750767827033997e-05 }, { "epoch": 0.19081616096057116, "step": 588, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5264361031295266e-06, "weight_min": 5.958802340444436e-10, "weight_std": 9.970297469408251e-06 }, { "epoch": 0.19081616096057116, "step": 588, "weight_max": 8.22571433189978e-09, "weight_mean": 1.5033957412313725e-09, "weight_min": 0.0, "weight_std": 2.7253503898094777e-09 }, { "epoch": 0.19081616096057116, "step": 588, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2032622180413455e-05, "weight_min": 0.0, "weight_std": 1.721842636470683e-05 }, { "epoch": 0.19081616096057116, "step": 588, "weight_max": 3.813097282545641e-05, "weight_mean": 9.086628779186867e-06, "weight_min": 0.0, "weight_std": 1.685106872173492e-05 }, { "epoch": 0.19114067824111633, "step": 589, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998487838747678e-06, "weight_min": 0.0, "weight_std": 1.4135968740447424e-05 }, { "epoch": 0.19114067824111633, "step": 589, "weight_max": 4.252122016623616e-05, "weight_mean": 9.921713171934243e-06, "weight_min": 0.0, "weight_std": 1.8201448256149888e-05 }, { "epoch": 0.19114067824111633, "step": 589, "weight_max": 3.967146767536178e-05, "weight_mean": 9.247306479664985e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.717162194836419e-05 }, { "epoch": 0.19114067824111633, "step": 589, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.6893139875028282e-05, "weight_min": 1.0509849657225345e-09, "weight_std": 1.7221749658347107e-05 }, { "epoch": 0.19146519552166152, "grad_norm": 1.601848840713501, "learning_rate": 9.004081632653063e-06, "loss": 0.293, "step": 590 }, { "epoch": 0.19146519552166152, "step": 590, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5503781469305977e-05, "weight_min": 0.0, "weight_std": 2.1146486687939614e-05 }, { "epoch": 0.19146519552166152, "step": 590, "weight_max": 3.998322426923551e-05, "weight_mean": 9.714189218357205e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7935501091415063e-05 }, { "epoch": 0.19146519552166152, "step": 590, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4889761587255634e-05, "weight_min": 0.0, "weight_std": 2.0600089555955492e-05 }, { "epoch": 0.19146519552166152, "step": 590, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2035130566800945e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6768784917076118e-05 }, { "epoch": 0.19178971280220672, "step": 591, "weight_max": 4.187939703115262e-05, "weight_mean": 9.775853868632112e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8160713807446882e-05 }, { "epoch": 0.19178971280220672, "step": 591, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9927232642658055e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1359481252147816e-05 }, { "epoch": 0.19178971280220672, "step": 591, "weight_max": 3.813097282545641e-05, "weight_mean": 8.731258276384324e-06, "weight_min": 0.0, "weight_std": 1.625663207960315e-05 }, { "epoch": 0.19178971280220672, "step": 591, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965407813666388e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1212612662347965e-05 }, { "epoch": 0.1921142300827519, "step": 592, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4753986761206761e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.049618160526734e-05 }, { "epoch": 0.1921142300827519, "step": 592, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.635324775241315e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6313397281919606e-05 }, { "epoch": 0.1921142300827519, "step": 592, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5863853807095438e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.144921018043533e-05 }, { "epoch": 0.1921142300827519, "step": 592, "weight_max": 2.639517333591357e-05, "weight_mean": 3.301729293525568e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.331160981673747e-06 }, { "epoch": 0.1924387473632971, "step": 593, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0312149242963642e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9095121388090774e-05 }, { "epoch": 0.1924387473632971, "step": 593, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7704995116218925e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9436896764091216e-05 }, { "epoch": 0.1924387473632971, "step": 593, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5023830201243982e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0857714844169095e-05 }, { "epoch": 0.1924387473632971, "step": 593, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235401204117807e-06, "weight_min": 0.0, "weight_std": 1.4806410035816953e-05 }, { "epoch": 0.19276326464384227, "step": 594, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793164523813175e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0726316759246401e-05 }, { "epoch": 0.19276326464384227, "step": 594, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155865437700413e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4581186405848712e-05 }, { "epoch": 0.19276326464384227, "step": 594, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.8492970563529525e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.369533856632188e-05 }, { "epoch": 0.19276326464384227, "step": 594, "weight_max": 1.0385104998178463e-09, "weight_mean": 5.350147547211748e-10, "weight_min": 0.0, "weight_std": 3.6311836870694947e-10 }, { "epoch": 0.19308778192438747, "step": 595, "weight_max": 5.799720383947715e-06, "weight_mean": 1.2196696843602695e-06, "weight_min": 5.771870759119224e-10, "weight_std": 2.3104339561541565e-06 }, { "epoch": 0.19308778192438747, "step": 595, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9187988073099405e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.054421929642558e-05 }, { "epoch": 0.19308778192438747, "step": 595, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.431501004844904e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9822356989607215e-05 }, { "epoch": 0.19308778192438747, "step": 595, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4227486644813325e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9915414668503217e-05 }, { "epoch": 0.19341229920493266, "step": 596, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4917592125129886e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.059350117633585e-05 }, { "epoch": 0.19341229920493266, "step": 596, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0552159437793307e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.954345134436153e-05 }, { "epoch": 0.19341229920493266, "step": 596, "weight_max": 3.632264997577295e-05, "weight_mean": 8.978990081232041e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.660463931330014e-05 }, { "epoch": 0.19341229920493266, "step": 596, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504443950281711e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2736443750327453e-05 }, { "epoch": 0.19373681648547786, "step": 597, "weight_max": 3.199401908204891e-05, "weight_mean": 4.000384706159821e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1311136404401623e-05 }, { "epoch": 0.19373681648547786, "step": 597, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2907738891954068e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8371107216808014e-05 }, { "epoch": 0.19373681648547786, "step": 597, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.671240897849202e-05, "weight_min": 0.0, "weight_std": 1.8846469174604863e-05 }, { "epoch": 0.19373681648547786, "step": 597, "weight_max": 3.998322426923551e-05, "weight_mean": 8.859281479089987e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6581891031819396e-05 }, { "epoch": 0.19406133376602305, "step": 598, "weight_max": 3.0886087188264355e-05, "weight_mean": 7.552507213404169e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3987582860863768e-05 }, { "epoch": 0.19406133376602305, "step": 598, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.788476947695017e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9784678443102166e-05 }, { "epoch": 0.19406133376602305, "step": 598, "weight_max": 3.843663216684945e-05, "weight_mean": 9.089024388231337e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.68642436619848e-05 }, { "epoch": 0.19406133376602305, "step": 598, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.194320227834396e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.702074041531887e-05 }, { "epoch": 0.19438585104656822, "step": 599, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843847364099929e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3697530448553152e-05 }, { "epoch": 0.19438585104656822, "step": 599, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1558158702391665e-06, "weight_min": 0.0, "weight_std": 1.458120641473215e-05 }, { "epoch": 0.19438585104656822, "step": 599, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3285132556338795e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.985353446798399e-05 }, { "epoch": 0.19438585104656822, "step": 599, "weight_max": 4.252122016623616e-05, "weight_mean": 7.869821274653077e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.572090150148142e-05 }, { "epoch": 0.19471036832711341, "grad_norm": 0.5814707279205322, "learning_rate": 8.983673469387756e-06, "loss": 0.1815, "step": 600 }, { "epoch": 0.19471036832711341, "eval_accuracy": 0.494, "eval_auc": 0.667647141188331, "eval_loss": 2.305392265319824, "eval_runtime": 155.8331, "eval_samples_per_second": 12.834, "eval_steps_per_second": 1.604, "step": 600 }, { "epoch": 0.19471036832711341, "step": 600, "weight_max": 4.252122016623616e-05, "weight_mean": 8.908200470614247e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6896952729439363e-05 }, { "epoch": 0.19471036832711341, "step": 600, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311815458408091e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9095326933893375e-05 }, { "epoch": 0.19471036832711341, "step": 600, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0551657396717928e-05, "weight_min": 0.0, "weight_std": 1.9543762391549535e-05 }, { "epoch": 0.19471036832711341, "step": 600, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155794497113675e-06, "weight_min": 0.0, "weight_std": 1.4581214600184467e-05 }, { "epoch": 0.1950348856076586, "step": 601, "weight_max": 3.9051607018336654e-05, "weight_mean": 5.352376319933683e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3679906260222197e-05 }, { "epoch": 0.1950348856076586, "step": 601, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805411663255654e-06, "weight_min": 0.0, "weight_std": 1.3589065929409117e-05 }, { "epoch": 0.1950348856076586, "step": 601, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156068709766259e-06, "weight_min": 0.0, "weight_std": 1.4581104551325552e-05 }, { "epoch": 0.1950348856076586, "step": 601, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9870738469762728e-05, "weight_min": 7.721521200210191e-10, "weight_std": 2.1254267267067917e-05 }, { "epoch": 0.1953594028882038, "step": 602, "weight_max": 3.967146767536178e-05, "weight_mean": 9.533529009786434e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6122650777106173e-05 }, { "epoch": 0.1953594028882038, "step": 602, "weight_max": 4.187939703115262e-05, "weight_mean": 2.8053093046764843e-05, "weight_min": 0.0, "weight_std": 1.792430703062564e-05 }, { "epoch": 0.1953594028882038, "step": 602, "weight_max": 3.967146767536178e-05, "weight_mean": 8.274870197055861e-06, "weight_min": 7.298507354036587e-10, "weight_std": 1.5687806808273308e-05 }, { "epoch": 0.1953594028882038, "step": 602, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.84442762171966e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3697296708414797e-05 }, { "epoch": 0.195683920168749, "step": 603, "weight_max": 1.2301966307859402e-05, "weight_mean": 1.5384862308565062e-06, "weight_min": 5.771870759119224e-10, "weight_std": 4.349102709966246e-06 }, { "epoch": 0.195683920168749, "step": 603, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0941377442795783e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.2391513994080015e-05 }, { "epoch": 0.195683920168749, "step": 603, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156101906322874e-06, "weight_min": 0.0, "weight_std": 1.4581089999410324e-05 }, { "epoch": 0.195683920168749, "step": 603, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882338544121012e-06, "weight_min": 0.0, "weight_std": 1.3806468814436812e-05 }, { "epoch": 0.19600843744929417, "step": 604, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.330724424216896e-06, "weight_min": 0.0, "weight_std": 1.7350754205835983e-05 }, { "epoch": 0.19600843744929417, "step": 604, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6638683518976904e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.877470094768796e-05 }, { "epoch": 0.19600843744929417, "step": 604, "weight_max": 2.8994179956498556e-05, "weight_mean": 3.83700216843863e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0181985089730006e-05 }, { "epoch": 0.19600843744929417, "step": 604, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8790560716297477e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.02125811483711e-05 }, { "epoch": 0.19633295472983936, "step": 605, "weight_max": 3.967146767536178e-05, "weight_mean": 8.670674105815124e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.437256287317723e-05 }, { "epoch": 0.19633295472983936, "step": 605, "weight_max": 4.187939703115262e-05, "weight_mean": 8.859698027663399e-06, "weight_min": 0.0, "weight_std": 1.6761317965574563e-05 }, { "epoch": 0.19633295472983936, "step": 605, "weight_max": 4.187939703115262e-05, "weight_mean": 8.793826964392792e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6671048797434196e-05 }, { "epoch": 0.19633295472983936, "step": 605, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9195646018488333e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.058458630926907e-05 }, { "epoch": 0.19665747201038455, "step": 606, "weight_max": 2.821082034643041e-06, "weight_mean": 3.54598199692191e-07, "weight_min": 0.0, "weight_std": 9.966182688003755e-07 }, { "epoch": 0.19665747201038455, "step": 606, "weight_max": 3.722132169059478e-05, "weight_mean": 8.277445886051282e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5482526578125544e-05 }, { "epoch": 0.19665747201038455, "step": 606, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4319489966728725e-05, "weight_min": 0.0, "weight_std": 2.025961111939978e-05 }, { "epoch": 0.19665747201038455, "step": 606, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355523045291193e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.919598253152799e-05 }, { "epoch": 0.19698198929092975, "step": 607, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.9279096856480464e-06, "weight_min": 0.0, "weight_std": 1.3665597180079203e-05 }, { "epoch": 0.19698198929092975, "step": 607, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4765108971914742e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.039573882939294e-05 }, { "epoch": 0.19698198929092975, "step": 607, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844374871026957e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.3697317626792938e-05 }, { "epoch": 0.19698198929092975, "step": 607, "weight_max": 4.3167921830900013e-05, "weight_mean": 6.607600880670361e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.515633994131349e-05 }, { "epoch": 0.19730650657147494, "step": 608, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4048664525034837e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9756005713134073e-05 }, { "epoch": 0.19730650657147494, "step": 608, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5397597962873988e-05, "weight_min": 0.0, "weight_std": 2.1295996702974662e-05 }, { "epoch": 0.19730650657147494, "step": 608, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.301256452279631e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.8329146769247018e-05 }, { "epoch": 0.19730650657147494, "step": 608, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.677508043940179e-05, "weight_min": 0.0, "weight_std": 1.8393438949715346e-05 }, { "epoch": 0.1976310238520201, "step": 609, "weight_max": 4.381950566312298e-05, "weight_mean": 1.887827784230467e-05, "weight_min": 0.0, "weight_std": 2.0504669009824283e-05 }, { "epoch": 0.1976310238520201, "step": 609, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355543054174632e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9195969798602164e-05 }, { "epoch": 0.1976310238520201, "step": 609, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3212415979069192e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8372247723164037e-05 }, { "epoch": 0.1976310238520201, "step": 609, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.0540950825088657e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.5924975741654634e-05 }, { "epoch": 0.1979555411325653, "grad_norm": 0.34276318550109863, "learning_rate": 8.963265306122449e-06, "loss": 0.1036, "step": 610 }, { "epoch": 0.1979555411325653, "step": 610, "weight_max": 3.843663216684945e-05, "weight_mean": 9.124857569986489e-06, "weight_min": 0.0, "weight_std": 1.692649493634235e-05 }, { "epoch": 0.1979555411325653, "step": 610, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.398575922299642e-06, "weight_min": 0.0, "weight_std": 1.5261121006915346e-05 }, { "epoch": 0.1979555411325653, "step": 610, "weight_max": 4.187939703115262e-05, "weight_mean": 1.391723253618693e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9641949620563537e-05 }, { "epoch": 0.1979555411325653, "step": 610, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0827495973207988e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.2283924408839084e-05 }, { "epoch": 0.1982800584131105, "step": 611, "weight_max": 1.0509849657225345e-09, "weight_mean": 7.686274394735904e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.9478375856696317e-10 }, { "epoch": 0.1982800584131105, "step": 611, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.767218216438778e-06, "weight_min": 0.0, "weight_std": 1.810294452297967e-05 }, { "epoch": 0.1982800584131105, "step": 611, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4999022823758423e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0716975996037945e-05 }, { "epoch": 0.1982800584131105, "step": 611, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9206678340851795e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.391596561006736e-05 }, { "epoch": 0.1986045756936557, "step": 612, "weight_max": 3.9360926166409627e-05, "weight_mean": 7.821290637366474e-06, "weight_min": 0.0, "weight_std": 1.5109551895875484e-05 }, { "epoch": 0.1986045756936557, "step": 612, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.409100332239177e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9529792552930303e-05 }, { "epoch": 0.1986045756936557, "step": 612, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843544047616888e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3697654139832594e-05 }, { "epoch": 0.1986045756936557, "step": 612, "weight_max": 4.252122016623616e-05, "weight_mean": 5.317100658430718e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5032734154374339e-05 }, { "epoch": 0.1989290929742009, "step": 613, "weight_max": 4.252122016623616e-05, "weight_mean": 1.027459711622214e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.903883457998745e-05 }, { "epoch": 0.1989290929742009, "step": 613, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8502854800317436e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.796105243556667e-05 }, { "epoch": 0.1989290929742009, "step": 613, "weight_max": 3.312147964606993e-05, "weight_mean": 5.322922788764117e-06, "weight_min": 0.0, "weight_std": 1.1709534192050342e-05 }, { "epoch": 0.1989290929742009, "step": 613, "weight_max": 4.252122016623616e-05, "weight_mean": 9.782575034478214e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.820153192966245e-05 }, { "epoch": 0.19925361025474606, "step": 614, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.903603884391487e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6699510524631478e-05 }, { "epoch": 0.19925361025474606, "step": 614, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.157858246704563e-06, "weight_min": 0.0, "weight_std": 1.569507185195107e-05 }, { "epoch": 0.19925361025474606, "step": 614, "weight_max": 4.252122016623616e-05, "weight_mean": 2.231922189821489e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0121588022448123e-05 }, { "epoch": 0.19925361025474606, "step": 614, "weight_max": 4.252122016623616e-05, "weight_mean": 9.782238521438558e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.820174111344386e-05 }, { "epoch": 0.19957812753529125, "step": 615, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0946965630864725e-05, "weight_min": 0.0, "weight_std": 1.9141592929372564e-05 }, { "epoch": 0.19957812753529125, "step": 615, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.007330507098231e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.551763853058219e-05 }, { "epoch": 0.19957812753529125, "step": 615, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.157101441000123e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4580686183762737e-05 }, { "epoch": 0.19957812753529125, "step": 615, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4778947843296919e-05, "weight_min": 0.0, "weight_std": 2.0480034436332062e-05 }, { "epoch": 0.19990264481583644, "step": 616, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.616653313860297e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3052743270236533e-05 }, { "epoch": 0.19990264481583644, "step": 616, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3049881999904756e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8388720491202548e-05 }, { "epoch": 0.19990264481583644, "step": 616, "weight_max": 4.252122016623616e-05, "weight_mean": 1.92265106306877e-05, "weight_min": 7.614652242082798e-10, "weight_std": 2.0748191673192196e-05 }, { "epoch": 0.19990264481583644, "step": 616, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156666247785324e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4580861716240179e-05 }, { "epoch": 0.20022716209638164, "step": 617, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.953836908796802e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.093494731525425e-05 }, { "epoch": 0.20022716209638164, "step": 617, "weight_max": 3.967146767536178e-05, "weight_mean": 4.961374997947132e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4024995834915899e-05 }, { "epoch": 0.20022716209638164, "step": 617, "weight_max": 3.006794759130571e-05, "weight_mean": 3.759049604923348e-06, "weight_min": 0.0, "weight_std": 1.0630399629008025e-05 }, { "epoch": 0.20022716209638164, "step": 617, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6874264474608935e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8925184122053906e-05 }, { "epoch": 0.20055167937692683, "step": 618, "weight_max": 4.252122016623616e-05, "weight_mean": 1.964315742952749e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1160789401619695e-05 }, { "epoch": 0.20055167937692683, "step": 618, "weight_max": 2.8994179956498556e-05, "weight_mean": 3.6251922210794874e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.0250619197904598e-05 }, { "epoch": 0.20055167937692683, "step": 618, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882458142674295e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3806420611217618e-05 }, { "epoch": 0.20055167937692683, "step": 618, "weight_max": 3.782653220696375e-05, "weight_mean": 9.344169484393205e-06, "weight_min": 0.0, "weight_std": 1.7301796106039546e-05 }, { "epoch": 0.200876196657472, "step": 619, "weight_max": 1.1182503811824063e-08, "weight_mean": 3.3075220340350597e-09, "weight_min": 6.586902134841921e-10, "weight_std": 4.530129604773947e-09 }, { "epoch": 0.200876196657472, "step": 619, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.482934390485752e-06, "weight_min": 0.0, "weight_std": 1.4422936146729626e-05 }, { "epoch": 0.200876196657472, "step": 619, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077089099359e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.4357627151184715e-05 }, { "epoch": 0.200876196657472, "step": 619, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471138011780567e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9390117813600227e-05 }, { "epoch": 0.2012007139380172, "grad_norm": 1.7755098342895508, "learning_rate": 8.942857142857142e-06, "loss": 0.0942, "step": 620 }, { "epoch": 0.2012007139380172, "step": 620, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998512395104626e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4135957826511003e-05 }, { "epoch": 0.2012007139380172, "step": 620, "weight_max": 4.381950566312298e-05, "weight_mean": 1.8547279978520237e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.033217970165424e-05 }, { "epoch": 0.2012007139380172, "step": 620, "weight_max": 3.782653220696375e-05, "weight_mean": 8.904755304683931e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.652909122640267e-05 }, { "epoch": 0.2012007139380172, "step": 620, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3207158190198243e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.849683940235991e-05 }, { "epoch": 0.2015252312185624, "step": 621, "weight_max": 3.632264997577295e-05, "weight_mean": 8.685204193170648e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6093061276478693e-05 }, { "epoch": 0.2015252312185624, "step": 621, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0095265679410659e-05, "weight_min": 0.0, "weight_std": 1.865066951722838e-05 }, { "epoch": 0.2015252312185624, "step": 621, "weight_max": 4.252122016623616e-05, "weight_mean": 1.509590947534889e-05, "weight_min": 0.0, "weight_std": 2.087046595988795e-05 }, { "epoch": 0.2015252312185624, "step": 621, "weight_max": 9.062063877074422e-10, "weight_mean": 6.300291954808301e-10, "weight_min": 0.0, "weight_std": 2.8201241342173944e-10 }, { "epoch": 0.20184974849910758, "step": 622, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.048311428865418e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4967625247663818e-05 }, { "epoch": 0.20184974849910758, "step": 622, "weight_max": 4.187939703115262e-05, "weight_mean": 1.299472205573693e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.877265822258778e-05 }, { "epoch": 0.20184974849910758, "step": 622, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7706261132843792e-05, "weight_min": 9.772551656794803e-10, "weight_std": 2.0259562006685883e-05 }, { "epoch": 0.20184974849910758, "step": 622, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3149499611463398e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9992119632661343e-05 }, { "epoch": 0.20217426577965278, "step": 623, "weight_max": 4.187939703115262e-05, "weight_mean": 9.062445315066725e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7046009816112928e-05 }, { "epoch": 0.20217426577965278, "step": 623, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3484273040376138e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8815679140971042e-05 }, { "epoch": 0.20217426577965278, "step": 623, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3216880688560195e-05, "weight_min": 0.0, "weight_std": 1.8414122678223066e-05 }, { "epoch": 0.20217426577965278, "step": 623, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.472464828111697e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0331826817709953e-05 }, { "epoch": 0.20249878306019795, "step": 624, "weight_max": 1.56184469233267e-05, "weight_mean": 5.0723133426799905e-06, "weight_min": 6.390288853630466e-10, "weight_std": 7.197439572337316e-06 }, { "epoch": 0.20249878306019795, "step": 624, "weight_max": 3.813097282545641e-05, "weight_mean": 4.766958227264695e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3481097994372249e-05 }, { "epoch": 0.20249878306019795, "step": 624, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9005692593054846e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.9770322978729382e-05 }, { "epoch": 0.20249878306019795, "step": 624, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5651407011318952e-05, "weight_min": 0.0, "weight_std": 1.8873166482080705e-05 }, { "epoch": 0.20282330034074314, "step": 625, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.31352042243816e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7280353858950548e-05 }, { "epoch": 0.20282330034074314, "step": 625, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8015241948887706e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9881990738213062e-05 }, { "epoch": 0.20282330034074314, "step": 625, "weight_max": 4.187939703115262e-05, "weight_mean": 6.584822585864458e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4752587048860732e-05 }, { "epoch": 0.20282330034074314, "step": 625, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4422364984056912e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.993468868022319e-05 }, { "epoch": 0.20314781762128833, "step": 626, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3981203665025532e-05, "weight_min": 0.0, "weight_std": 1.93939504242735e-05 }, { "epoch": 0.20314781762128833, "step": 626, "weight_max": 3.5434961318969727e-05, "weight_mean": 9.152508937404491e-06, "weight_min": 0.0, "weight_std": 1.5529834854532965e-05 }, { "epoch": 0.20314781762128833, "step": 626, "weight_max": 2.2490561605081894e-05, "weight_mean": 2.8117956389905885e-06, "weight_min": 0.0, "weight_std": 7.951422048790846e-06 }, { "epoch": 0.20314781762128833, "step": 626, "weight_max": 4.381950566312298e-05, "weight_mean": 2.481051706126891e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0716986909974366e-05 }, { "epoch": 0.20347233490183353, "step": 627, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.868444451247342e-05, "weight_min": 0.0, "weight_std": 2.0270392269594595e-05 }, { "epoch": 0.20347233490183353, "step": 627, "weight_max": 3.782653220696375e-05, "weight_mean": 8.555238309781998e-06, "weight_min": 0.0, "weight_std": 1.595714820723515e-05 }, { "epoch": 0.20347233490183353, "step": 627, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843566784984432e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3697645044885576e-05 }, { "epoch": 0.20347233490183353, "step": 627, "weight_max": 3.632264997577295e-05, "weight_mean": 8.970688213594258e-06, "weight_min": 0.0, "weight_std": 1.6609745216555893e-05 }, { "epoch": 0.20379685218237872, "step": 628, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6939937268034555e-05, "weight_min": 0.0, "weight_std": 1.8669654309633188e-05 }, { "epoch": 0.20379685218237872, "step": 628, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.240816325473133e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.715648249955848e-05 }, { "epoch": 0.20379685218237872, "step": 628, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728729436465073e-06, "weight_min": 0.0, "weight_std": 1.3373532055993564e-05 }, { "epoch": 0.20379685218237872, "step": 628, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5010560673545115e-05, "weight_min": 0.0, "weight_std": 2.078655961668119e-05 }, { "epoch": 0.2041213694629239, "step": 629, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390898751211353e-05, "weight_min": 0.0, "weight_std": 1.9239312678109854e-05 }, { "epoch": 0.2041213694629239, "step": 629, "weight_max": 3.722132169059478e-05, "weight_mean": 1.3375110938795842e-05, "weight_min": 0.0, "weight_std": 1.8514678231440485e-05 }, { "epoch": 0.2041213694629239, "step": 629, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.606133633293211e-06, "weight_min": 0.0, "weight_std": 1.5942303434712812e-05 }, { "epoch": 0.2041213694629239, "step": 629, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212239673506701e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1911834917555097e-05 }, { "epoch": 0.20444588674346909, "grad_norm": 0.44419312477111816, "learning_rate": 8.922448979591837e-06, "loss": 0.3153, "step": 630 }, { "epoch": 0.20444588674346909, "step": 630, "weight_max": 3.4558255720185116e-05, "weight_mean": 7.463000201823888e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3687624232261442e-05 }, { "epoch": 0.20444588674346909, "step": 630, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2123760572867468e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9193801563233137e-05 }, { "epoch": 0.20444588674346909, "step": 630, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9260061890236102e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0830799257964827e-05 }, { "epoch": 0.20444588674346909, "step": 630, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3534905519918539e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9360491933184676e-05 }, { "epoch": 0.20477040402401428, "step": 631, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320361313148169e-06, "weight_min": 0.0, "weight_std": 1.2217954463267233e-05 }, { "epoch": 0.20477040402401428, "step": 631, "weight_max": 7.2748242985198885e-09, "weight_mean": 1.90712245995428e-09, "weight_min": 6.053385570581327e-10, "weight_std": 2.2436930091629392e-09 }, { "epoch": 0.20477040402401428, "step": 631, "weight_max": 3.9051607018336654e-05, "weight_mean": 6.598105755983852e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3965173820906784e-05 }, { "epoch": 0.20477040402401428, "step": 631, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.469735232414678e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7593518350622617e-05 }, { "epoch": 0.20509492130455947, "step": 632, "weight_max": 9.087236740867866e-09, "weight_mean": 1.6980548078748825e-09, "weight_min": 0.0, "weight_std": 3.0013849183774255e-09 }, { "epoch": 0.20509492130455947, "step": 632, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3885426596971229e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9483357391436584e-05 }, { "epoch": 0.20509492130455947, "step": 632, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3663347999681719e-05, "weight_min": 0.0, "weight_std": 1.9172653992427513e-05 }, { "epoch": 0.20509492130455947, "step": 632, "weight_max": 3.998322426923551e-05, "weight_mean": 9.576286174706183e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7752719941199757e-05 }, { "epoch": 0.20541943858510467, "step": 633, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0037533684226219e-05, "weight_min": 0.0, "weight_std": 1.859320946095977e-05 }, { "epoch": 0.20541943858510467, "step": 633, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233556167804636e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.895384230010677e-05 }, { "epoch": 0.20541943858510467, "step": 633, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235882781562395e-06, "weight_min": 0.0, "weight_std": 1.4806216313445475e-05 }, { "epoch": 0.20541943858510467, "step": 633, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1898240700247698e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7020742234308273e-05 }, { "epoch": 0.20574395586564984, "step": 634, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3886139640817419e-05, "weight_min": 8.490471659072796e-10, "weight_std": 1.9482777133816853e-05 }, { "epoch": 0.20574395586564984, "step": 634, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728990006697131e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3373426554608159e-05 }, { "epoch": 0.20574395586564984, "step": 634, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8046262994175777e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9741704818443395e-05 }, { "epoch": 0.20574395586564984, "step": 634, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.2927171812625602e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.929845893755555e-05 }, { "epoch": 0.20606847314619503, "step": 635, "weight_max": 3.0886087188264355e-05, "weight_mean": 7.025311788311228e-06, "weight_min": 0.0, "weight_std": 1.243382939719595e-05 }, { "epoch": 0.20606847314619503, "step": 635, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3357953321246896e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9004895875696093e-05 }, { "epoch": 0.20606847314619503, "step": 635, "weight_max": 3.998322426923551e-05, "weight_mean": 9.427752047486138e-06, "weight_min": 0.0, "weight_std": 1.7497897715657018e-05 }, { "epoch": 0.20606847314619503, "step": 635, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.7077310985769145e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.830141991376877e-05 }, { "epoch": 0.20639299042674022, "step": 636, "weight_max": 4.187939703115262e-05, "weight_mean": 9.027919077198021e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6997222701320425e-05 }, { "epoch": 0.20639299042674022, "step": 636, "weight_max": 5.042821271672437e-07, "weight_mean": 6.735709234817477e-08, "weight_min": 5.679522963042416e-10, "weight_std": 1.768508752775233e-07 }, { "epoch": 0.20639299042674022, "step": 636, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.460374374408275e-06, "weight_min": 0.0, "weight_std": 1.5669269487261772e-05 }, { "epoch": 0.20639299042674022, "step": 636, "weight_max": 3.312147964606993e-05, "weight_mean": 9.88837746263016e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4422173080674838e-05 }, { "epoch": 0.20671750770728542, "step": 637, "weight_max": 3.28377827827353e-05, "weight_mean": 4.10630127589684e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.1609271496126894e-05 }, { "epoch": 0.20671750770728542, "step": 637, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.0676481653936207e-05, "weight_min": 1.0261104188558079e-09, "weight_std": 1.7789090634323657e-05 }, { "epoch": 0.20671750770728542, "step": 637, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1875963537022471e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7442789612687193e-05 }, { "epoch": 0.20671750770728542, "step": 637, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.982667890843004e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1317418941180222e-05 }, { "epoch": 0.20704202498783061, "step": 638, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.770729042822495e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.736024751153309e-05 }, { "epoch": 0.20704202498783061, "step": 638, "weight_max": 4.252122016623616e-05, "weight_mean": 5.31581417817506e-06, "weight_min": 0.0, "weight_std": 1.5033254385343753e-05 }, { "epoch": 0.20704202498783061, "step": 638, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.31128999782959e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.804628482204862e-05 }, { "epoch": 0.20704202498783061, "step": 638, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.232034244632814e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7397600458934903e-05 }, { "epoch": 0.20736654226837578, "step": 639, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076936758996453e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4357688996824436e-05 }, { "epoch": 0.20736654226837578, "step": 639, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4803144949837588e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0443807443371043e-05 }, { "epoch": 0.20736654226837578, "step": 639, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4380108041223139e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.986186543945223e-05 }, { "epoch": 0.20736654226837578, "step": 639, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.696272172732279e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.800053178158123e-05 }, { "epoch": 0.20769105954892098, "grad_norm": 1.186802864074707, "learning_rate": 8.90204081632653e-06, "loss": 0.2273, "step": 640 }, { "epoch": 0.20769105954892098, "step": 640, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3253849829197861e-05, "weight_min": 7.508524357824342e-10, "weight_std": 1.9268953110440634e-05 }, { "epoch": 0.20769105954892098, "step": 640, "weight_max": 2.952862632810138e-05, "weight_mean": 3.6919257127010496e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0439603101985995e-05 }, { "epoch": 0.20769105954892098, "step": 640, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0244824554538354e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.171344203816261e-05 }, { "epoch": 0.20769105954892098, "step": 640, "weight_max": 1.1273959543700585e-09, "weight_mean": 7.63737739717385e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.0385527987887286e-10 }, { "epoch": 0.20801557682946617, "step": 641, "weight_max": 4.252122016623616e-05, "weight_mean": 1.501674523751717e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.082074570353143e-05 }, { "epoch": 0.20801557682946617, "step": 641, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235936896497151e-06, "weight_min": 0.0, "weight_std": 1.480619357607793e-05 }, { "epoch": 0.20801557682946617, "step": 641, "weight_max": 3.782653220696375e-05, "weight_mean": 1.0240557458018884e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.700551547401119e-05 }, { "epoch": 0.20801557682946617, "step": 641, "weight_max": 8.651113603264093e-09, "weight_mean": 2.191787418226454e-09, "weight_min": 5.958802340444436e-10, "weight_std": 2.7530202562076056e-09 }, { "epoch": 0.20834009411001136, "step": 642, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504020580498036e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2736614735331386e-05 }, { "epoch": 0.20834009411001136, "step": 642, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.42827229865361e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7125568774645217e-05 }, { "epoch": 0.20834009411001136, "step": 642, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4682520486530848e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.042256892309524e-05 }, { "epoch": 0.20834009411001136, "step": 642, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2394761951291e-06, "weight_min": 0.0, "weight_std": 1.480476748838555e-05 }, { "epoch": 0.20866461139055656, "step": 643, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3123868484399281e-05, "weight_min": 0.0, "weight_std": 1.8359591194894165e-05 }, { "epoch": 0.20866461139055656, "step": 643, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1503505447763018e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.795356183720287e-05 }, { "epoch": 0.20866461139055656, "step": 643, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235498974798247e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4806370927544776e-05 }, { "epoch": 0.20866461139055656, "step": 643, "weight_max": 3.061215829802677e-05, "weight_mean": 4.23948404204566e-06, "weight_min": 0.0, "weight_std": 1.0718351404648274e-05 }, { "epoch": 0.20898912867110173, "step": 644, "weight_max": 3.752332122530788e-05, "weight_mean": 1.6767691704444587e-05, "weight_min": 0.0, "weight_std": 1.8014143279287964e-05 }, { "epoch": 0.20898912867110173, "step": 644, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4299963519733865e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2527897524705622e-05 }, { "epoch": 0.20898912867110173, "step": 644, "weight_max": 7.829137893544669e-10, "weight_mean": 6.6496819162154e-10, "weight_min": 5.679522963042416e-10, "weight_std": 7.679625685375058e-11 }, { "epoch": 0.20898912867110173, "step": 644, "weight_max": 3.843663216684945e-05, "weight_mean": 1.498010169598274e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7404758182237856e-05 }, { "epoch": 0.20931364595164692, "step": 645, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3848314665665384e-05, "weight_min": 0.0, "weight_std": 1.9285209418740124e-05 }, { "epoch": 0.20931364595164692, "step": 645, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1320356861688197e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.6447831512778066e-05 }, { "epoch": 0.20931364595164692, "step": 645, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0468427464948036e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.190311897720676e-05 }, { "epoch": 0.20931364595164692, "step": 645, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8199349142378196e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9895940567948855e-05 }, { "epoch": 0.20963816323219212, "step": 646, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.439967132289894e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.605006036697887e-05 }, { "epoch": 0.20963816323219212, "step": 646, "weight_max": 4.252122016623616e-05, "weight_mean": 1.546871499158442e-05, "weight_min": 0.0, "weight_std": 2.1358850062824786e-05 }, { "epoch": 0.20963816323219212, "step": 646, "weight_max": 4.252122016623616e-05, "weight_mean": 1.878062539617531e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0447630959097296e-05 }, { "epoch": 0.20963816323219212, "step": 646, "weight_max": 4.187939703115262e-05, "weight_mean": 7.706444193900097e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5443481970578432e-05 }, { "epoch": 0.2099626805127373, "step": 647, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.2256197805982083e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6964848327916116e-05 }, { "epoch": 0.2099626805127373, "step": 647, "weight_max": 4.187939703115262e-05, "weight_mean": 9.81311677605845e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.822323065425735e-05 }, { "epoch": 0.2099626805127373, "step": 647, "weight_max": 2.716206836339552e-05, "weight_mean": 3.3970536605920643e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.602516001905315e-06 }, { "epoch": 0.2099626805127373, "step": 647, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235527623881353e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4806359104113653e-05 }, { "epoch": 0.2102871977932825, "step": 648, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248047844157554e-06, "weight_min": 0.0, "weight_std": 1.2013475497951731e-05 }, { "epoch": 0.2102871977932825, "step": 648, "weight_max": 2.926079287135508e-05, "weight_mean": 3.658685727714328e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0344813745177817e-05 }, { "epoch": 0.2102871977932825, "step": 648, "weight_max": 3.662098970380612e-05, "weight_mean": 9.015357136377133e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.667557626205962e-05 }, { "epoch": 0.2102871977932825, "step": 648, "weight_max": 4.187939703115262e-05, "weight_mean": 1.004006844596006e-05, "weight_min": 0.0, "weight_std": 1.8611955965752713e-05 }, { "epoch": 0.21061171507382767, "step": 649, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4627674318035133e-05, "weight_min": 0.0, "weight_std": 2.0142017092439346e-05 }, { "epoch": 0.21061171507382767, "step": 649, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3276221579872072e-05, "weight_min": 0.0, "weight_std": 1.843227437348105e-05 }, { "epoch": 0.21061171507382767, "step": 649, "weight_max": 1.753466261078529e-09, "weight_mean": 8.605267609596012e-10, "weight_min": 5.958802340444436e-10, "weight_std": 3.896236944189724e-10 }, { "epoch": 0.21061171507382767, "step": 649, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4412229575100355e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.7520507753943093e-05 }, { "epoch": 0.21093623235437287, "grad_norm": 11.251177787780762, "learning_rate": 8.881632653061226e-06, "loss": 0.4706, "step": 650 }, { "epoch": 0.21093623235437287, "step": 650, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.622008292353712e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7876092897495255e-05 }, { "epoch": 0.21093623235437287, "step": 650, "weight_max": 1.753466261078529e-09, "weight_mean": 8.433231335480684e-10, "weight_min": 0.0, "weight_std": 5.511319733919606e-10 }, { "epoch": 0.21093623235437287, "step": 650, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0633289093675558e-05, "weight_min": 0.0, "weight_std": 1.929076825035736e-05 }, { "epoch": 0.21093623235437287, "step": 650, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316104306984926e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5033137060527224e-05 }, { "epoch": 0.21126074963491806, "step": 651, "weight_max": 3.0886087188264355e-05, "weight_mean": 7.653727152501233e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4171379007166252e-05 }, { "epoch": 0.21126074963491806, "step": 651, "weight_max": 1.2064879095774472e-09, "weight_mean": 7.861301609679572e-10, "weight_min": 0.0, "weight_std": 3.724066610644172e-10 }, { "epoch": 0.21126074963491806, "step": 651, "weight_max": 1.4305072681963793e-09, "weight_mean": 5.688994830777006e-10, "weight_min": 0.0, "weight_std": 4.4927672693262366e-10 }, { "epoch": 0.21126074963491806, "step": 651, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9500685084494762e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.088394103338942e-05 }, { "epoch": 0.21158526691546325, "step": 652, "weight_max": 3.0339444492710754e-05, "weight_mean": 6.967135050217621e-06, "weight_min": 0.0, "weight_std": 1.296585105592385e-05 }, { "epoch": 0.21158526691546325, "step": 652, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.1840714250865858e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.641492963244673e-05 }, { "epoch": 0.21158526691546325, "step": 652, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9177390640834346e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0649496946134605e-05 }, { "epoch": 0.21158526691546325, "step": 652, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.801276635495014e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4937359082978219e-05 }, { "epoch": 0.21190978419600845, "step": 653, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.701339538674802e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6406274880864657e-05 }, { "epoch": 0.21190978419600845, "step": 653, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.353708835784346e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4724019820278045e-05 }, { "epoch": 0.21190978419600845, "step": 653, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.158944986760616e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4579942217096686e-05 }, { "epoch": 0.21190978419600845, "step": 653, "weight_max": 4.187939703115262e-05, "weight_mean": 5.245138254394988e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4802486475673504e-05 }, { "epoch": 0.21223430147655362, "step": 654, "weight_max": 8.926079999582726e-07, "weight_mean": 1.1238544317393462e-07, "weight_min": 5.679522963042416e-10, "weight_std": 3.1525817689725955e-07 }, { "epoch": 0.21223430147655362, "step": 654, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8052224883576855e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3589141417469364e-05 }, { "epoch": 0.21223430147655362, "step": 654, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9859122403431684e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1325295165297575e-05 }, { "epoch": 0.21223430147655362, "step": 654, "weight_max": 3.4558255720185116e-05, "weight_mean": 6.628975370404078e-06, "weight_min": 0.0, "weight_std": 1.3005020264245104e-05 }, { "epoch": 0.2125588187570988, "step": 655, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.96006531547755e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.668462391535286e-05 }, { "epoch": 0.2125588187570988, "step": 655, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0845563338079955e-05, "weight_min": 0.0, "weight_std": 1.6357553249690682e-05 }, { "epoch": 0.2125588187570988, "step": 655, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.229502933914773e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.018741906795185e-05 }, { "epoch": 0.2125588187570988, "step": 655, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2710937880910933e-05, "weight_min": 0.0, "weight_std": 1.816419535316527e-05 }, { "epoch": 0.212883336037644, "step": 656, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.461515219300054e-06, "weight_min": 0.0, "weight_std": 1.7535494407638907e-05 }, { "epoch": 0.212883336037644, "step": 656, "weight_max": 3.998322426923551e-05, "weight_mean": 9.576224329066463e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.775275813997723e-05 }, { "epoch": 0.212883336037644, "step": 656, "weight_max": 3.752332122530788e-05, "weight_mean": 9.078115908778273e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6747002518968657e-05 }, { "epoch": 0.212883336037644, "step": 656, "weight_max": 4.187939703115262e-05, "weight_mean": 1.931734186655376e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9979641365353018e-05 }, { "epoch": 0.2132078533181892, "step": 657, "weight_max": 1.2129309652664233e-05, "weight_mean": 1.516854354122188e-06, "weight_min": 6.100955851628953e-10, "weight_std": 4.2880797082034405e-06 }, { "epoch": 0.2132078533181892, "step": 657, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156087809154997e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4581096365873236e-05 }, { "epoch": 0.2132078533181892, "step": 657, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4176399417920038e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.999038977373857e-05 }, { "epoch": 0.2132078533181892, "step": 657, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.77475305160624e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2208000043756329e-05 }, { "epoch": 0.2135323705987344, "step": 658, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396843334892765e-06, "weight_min": 0.0, "weight_std": 1.5261819498846307e-05 }, { "epoch": 0.2135323705987344, "step": 658, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.977029781613965e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.548422733321786e-05 }, { "epoch": 0.2135323705987344, "step": 658, "weight_max": 3.143761568935588e-05, "weight_mean": 7.722699592704885e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4301273040473461e-05 }, { "epoch": 0.2135323705987344, "step": 658, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.85183289431734e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.069538095383905e-05 }, { "epoch": 0.21385688787927956, "step": 659, "weight_max": 4.252122016623616e-05, "weight_mean": 2.7578331355471164e-05, "weight_min": 8.490471659072796e-10, "weight_std": 1.7763974028639495e-05 }, { "epoch": 0.21385688787927956, "step": 659, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153446055483073e-05, "weight_min": 0.0, "weight_std": 1.8802958948072046e-05 }, { "epoch": 0.21385688787927956, "step": 659, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0115255463460926e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.873201472335495e-05 }, { "epoch": 0.21385688787927956, "step": 659, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315621365298284e-06, "weight_min": 0.0, "weight_std": 1.5033331692393404e-05 }, { "epoch": 0.21418140515982476, "grad_norm": 4.832530498504639, "learning_rate": 8.861224489795919e-06, "loss": 0.2905, "step": 660 }, { "epoch": 0.21418140515982476, "step": 660, "weight_max": 3.998322426923551e-05, "weight_mean": 9.803220564208459e-06, "weight_min": 0.0, "weight_std": 1.815492032619659e-05 }, { "epoch": 0.21418140515982476, "step": 660, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.041128578246571e-06, "weight_min": 0.0, "weight_std": 1.6907830286072567e-05 }, { "epoch": 0.21418140515982476, "step": 660, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4048951925360598e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.944946779985912e-05 }, { "epoch": 0.21418140515982476, "step": 660, "weight_max": 1.1273959543700585e-09, "weight_mean": 8.239399162945915e-10, "weight_min": 6.196655966128617e-10, "weight_std": 1.8019952197079192e-10 }, { "epoch": 0.21450592244036995, "step": 661, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9314144083182327e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0874505935353227e-05 }, { "epoch": 0.21450592244036995, "step": 661, "weight_max": 4.252122016623616e-05, "weight_mean": 2.31761769100558e-05, "weight_min": 8.94625595826426e-10, "weight_std": 1.9689476175699383e-05 }, { "epoch": 0.21450592244036995, "step": 661, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7359712728648447e-05, "weight_min": 0.0, "weight_std": 1.996604260057211e-05 }, { "epoch": 0.21450592244036995, "step": 661, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4002957868797239e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.966274794540368e-05 }, { "epoch": 0.21483043972091515, "step": 662, "weight_max": 3.662098970380612e-05, "weight_mean": 1.7575728634255938e-05, "weight_min": 0.0, "weight_std": 1.8808799723046832e-05 }, { "epoch": 0.21483043972091515, "step": 662, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861365712509723e-06, "weight_min": 0.0, "weight_std": 1.0919637134065852e-05 }, { "epoch": 0.21483043972091515, "step": 662, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5356039511971176e-05, "weight_min": 0.0, "weight_std": 2.120498538715765e-05 }, { "epoch": 0.21483043972091515, "step": 662, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0076041689899284e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7869911971502006e-05 }, { "epoch": 0.21515495700146034, "step": 663, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.7252864331094315e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0534838111198042e-05 }, { "epoch": 0.21515495700146034, "step": 663, "weight_max": 3.28377827827353e-05, "weight_mean": 7.863818609621376e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4578284208255354e-05 }, { "epoch": 0.21515495700146034, "step": 663, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.792731023277156e-06, "weight_min": 0.0, "weight_std": 1.6830033928272314e-05 }, { "epoch": 0.21515495700146034, "step": 663, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.313816579175182e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.569320374983363e-05 }, { "epoch": 0.2154794742820055, "step": 664, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396814685809659e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.526183041278273e-05 }, { "epoch": 0.2154794742820055, "step": 664, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232392014586367e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.894552588055376e-05 }, { "epoch": 0.2154794742820055, "step": 664, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.616433216142468e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3052833310212009e-05 }, { "epoch": 0.2154794742820055, "step": 664, "weight_max": 4.187939703115262e-05, "weight_mean": 2.054388460237533e-05, "weight_min": 0.0, "weight_std": 2.1967553038848564e-05 }, { "epoch": 0.2158039915625507, "step": 665, "weight_max": 2.0431491066119634e-05, "weight_mean": 2.8137035315012326e-06, "weight_min": 5.771870759119224e-10, "weight_std": 7.1506210588268004e-06 }, { "epoch": 0.2158039915625507, "step": 665, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0316619409422856e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9128478015773e-05 }, { "epoch": 0.2158039915625507, "step": 665, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.433656487090047e-06, "weight_min": 0.0, "weight_std": 1.7532522178953514e-05 }, { "epoch": 0.2158039915625507, "step": 665, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7686670616967604e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9234530554967932e-05 }, { "epoch": 0.2161285088430959, "step": 666, "weight_max": 3.813097282545641e-05, "weight_mean": 9.196177416015416e-06, "weight_min": 0.0, "weight_std": 1.7042175386450253e-05 }, { "epoch": 0.2161285088430959, "step": 666, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3234044672572054e-05, "weight_min": 0.0, "weight_std": 1.713084930088371e-05 }, { "epoch": 0.2161285088430959, "step": 666, "weight_max": 4.187939703115262e-05, "weight_mean": 9.482747373112943e-06, "weight_min": 0.0, "weight_std": 1.7684280464891344e-05 }, { "epoch": 0.2161285088430959, "step": 666, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0711904906202108e-05, "weight_min": 0.0, "weight_std": 1.983346555789467e-05 }, { "epoch": 0.2164530261236411, "step": 667, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.976225757971406e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.144056634278968e-05 }, { "epoch": 0.2164530261236411, "step": 667, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8231970898341388e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.004523776122369e-05 }, { "epoch": 0.2164530261236411, "step": 667, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.296157259086613e-06, "weight_min": 0.0, "weight_std": 1.7347805624012835e-05 }, { "epoch": 0.2164530261236411, "step": 667, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3448598110699095e-05, "weight_min": 0.0, "weight_std": 1.8783035557135008e-05 }, { "epoch": 0.21677754340418628, "step": 668, "weight_max": 4.9337902652268895e-09, "weight_mean": 1.825727458104609e-09, "weight_min": 5.864964625068581e-10, "weight_std": 1.5506610440141344e-09 }, { "epoch": 0.21677754340418628, "step": 668, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105848347535357e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.160945430456195e-05 }, { "epoch": 0.21677754340418628, "step": 668, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.896633865631884e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4431197996600531e-05 }, { "epoch": 0.21677754340418628, "step": 668, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236869583313819e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4805817045271397e-05 }, { "epoch": 0.21710206068473145, "step": 669, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.825893357628956e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8309718143427745e-05 }, { "epoch": 0.21710206068473145, "step": 669, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3343203793046996e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.965647425095085e-05 }, { "epoch": 0.21710206068473145, "step": 669, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0711593859014101e-05, "weight_min": 0.0, "weight_std": 1.9833656551782042e-05 }, { "epoch": 0.21710206068473145, "step": 669, "weight_max": 3.998322426923551e-05, "weight_mean": 9.726796633913182e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8018325135926716e-05 }, { "epoch": 0.21742657796527665, "grad_norm": 0.48042866587638855, "learning_rate": 8.840816326530614e-06, "loss": 0.1467, "step": 670 }, { "epoch": 0.21742657796527665, "step": 670, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9150811567669734e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0655877960962243e-05 }, { "epoch": 0.21742657796527665, "step": 670, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.431567387859104e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5248091585817747e-05 }, { "epoch": 0.21742657796527665, "step": 670, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233452485408634e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8953905964735895e-05 }, { "epoch": 0.21742657796527665, "step": 670, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861247478198493e-06, "weight_min": 0.0, "weight_std": 1.0919684427790344e-05 }, { "epoch": 0.21775109524582184, "step": 671, "weight_max": 4.187939703115262e-05, "weight_mean": 8.7280141087831e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6583546312176622e-05 }, { "epoch": 0.21775109524582184, "step": 671, "weight_max": 3.632264997577295e-05, "weight_mean": 4.555860869004391e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2835786947107408e-05 }, { "epoch": 0.21775109524582184, "step": 671, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2225348655192647e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8012357031693682e-05 }, { "epoch": 0.21775109524582184, "step": 671, "weight_max": 3.199401908204891e-05, "weight_mean": 4.023768269689754e-06, "weight_min": 0.0, "weight_std": 1.1301886843284592e-05 }, { "epoch": 0.21807561252636704, "step": 672, "weight_max": 4.187939703115262e-05, "weight_mean": 9.41451799008064e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.7569478586665355e-05 }, { "epoch": 0.21807561252636704, "step": 672, "weight_max": 3.998322426923551e-05, "weight_mean": 9.318128832092043e-06, "weight_min": 0.0, "weight_std": 1.7313541320618242e-05 }, { "epoch": 0.21807561252636704, "step": 672, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4312821804196574e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0123123249504715e-05 }, { "epoch": 0.21807561252636704, "step": 672, "weight_max": 1.1273959543700585e-09, "weight_mean": 8.188396072306148e-10, "weight_min": 5.864964625068581e-10, "weight_std": 1.906688279484925e-10 }, { "epoch": 0.21840012980691223, "step": 673, "weight_max": 4.252122016623616e-05, "weight_mean": 9.129422323894687e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7162225049105473e-05 }, { "epoch": 0.21840012980691223, "step": 673, "weight_max": 3.998322426923551e-05, "weight_mean": 1.132139004766941e-05, "weight_min": 0.0, "weight_std": 1.7473565094405785e-05 }, { "epoch": 0.21840012980691223, "step": 673, "weight_max": 2.7679425329552032e-05, "weight_mean": 3.4960921766469255e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.772049452294596e-06 }, { "epoch": 0.21840012980691223, "step": 673, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805268417840125e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.358912413707003e-05 }, { "epoch": 0.2187246470874574, "step": 674, "weight_max": 3.843663216684945e-05, "weight_mean": 5.117653927300125e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3491247045749333e-05 }, { "epoch": 0.2187246470874574, "step": 674, "weight_max": 3.782653220696375e-05, "weight_mean": 9.232056982000358e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7099866454373114e-05 }, { "epoch": 0.2187246470874574, "step": 674, "weight_max": 3.967146767536178e-05, "weight_mean": 8.889176569937263e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6604686607024632e-05 }, { "epoch": 0.2187246470874574, "step": 674, "weight_max": 3.340639523230493e-05, "weight_mean": 1.2437196346581914e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.4604772331949789e-05 }, { "epoch": 0.2190491643680026, "step": 675, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.146237971435767e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.707054252619855e-05 }, { "epoch": 0.2190491643680026, "step": 675, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1731777703971602e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.7376825780957006e-05 }, { "epoch": 0.2190491643680026, "step": 675, "weight_max": 3.0886087188264355e-05, "weight_mean": 7.192455996118952e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3364661754167173e-05 }, { "epoch": 0.2190491643680026, "step": 675, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959685611538589e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4025677955942228e-05 }, { "epoch": 0.21937368164854779, "step": 676, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0159052180824801e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8835615264833905e-05 }, { "epoch": 0.21937368164854779, "step": 676, "weight_max": 1.2183443587332476e-08, "weight_mean": 2.4510056206850095e-09, "weight_min": 5.864964625068581e-10, "weight_std": 3.98102706355985e-09 }, { "epoch": 0.21937368164854779, "step": 676, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728745352622354e-06, "weight_min": 0.0, "weight_std": 1.337352478003595e-05 }, { "epoch": 0.21937368164854779, "step": 676, "weight_max": 4.187939703115262e-05, "weight_mean": 9.887056876323186e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.76954927155748e-05 }, { "epoch": 0.21969819892909298, "step": 677, "weight_max": 4.252122016623616e-05, "weight_mean": 1.527252607047558e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1093457689858042e-05 }, { "epoch": 0.21969819892909298, "step": 677, "weight_max": 2.4894337911973707e-05, "weight_mean": 3.1123036023927853e-06, "weight_min": 0.0, "weight_std": 8.801271178526804e-06 }, { "epoch": 0.21969819892909298, "step": 677, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0159637642791495e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1598278181045316e-05 }, { "epoch": 0.21969819892909298, "step": 677, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2842016076319851e-05, "weight_min": 0.0, "weight_std": 1.877233080449514e-05 }, { "epoch": 0.22002271620963818, "step": 678, "weight_max": 4.187939703115262e-05, "weight_mean": 1.011686799756717e-05, "weight_min": 0.0, "weight_std": 1.8746823116089217e-05 }, { "epoch": 0.22002271620963818, "step": 678, "weight_max": 2.952862632810138e-05, "weight_mean": 6.742400728398934e-06, "weight_min": 0.0, "weight_std": 1.2557335139717907e-05 }, { "epoch": 0.22002271620963818, "step": 678, "weight_max": 3.843663216684945e-05, "weight_mean": 1.13444693852216e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6121815860969946e-05 }, { "epoch": 0.22002271620963818, "step": 678, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076961770100752e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357679901877418e-05 }, { "epoch": 0.22034723349018334, "step": 679, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155990038474556e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4581135474145412e-05 }, { "epoch": 0.22034723349018334, "step": 679, "weight_max": 3.722132169059478e-05, "weight_mean": 7.985565389390104e-06, "weight_min": 0.0, "weight_std": 1.5049788999021985e-05 }, { "epoch": 0.22034723349018334, "step": 679, "weight_max": 4.252122016623616e-05, "weight_mean": 9.52780646912288e-06, "weight_min": 0.0, "weight_std": 1.77964993781643e-05 }, { "epoch": 0.22034723349018334, "step": 679, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504047410591738e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2736602911900263e-05 }, { "epoch": 0.22067175077072854, "grad_norm": 0.3145492970943451, "learning_rate": 8.820408163265307e-06, "loss": 0.2794, "step": 680 }, { "epoch": 0.22067175077072854, "step": 680, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0663890861906111e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8013539374805987e-05 }, { "epoch": 0.22067175077072854, "step": 680, "weight_max": 4.381950566312298e-05, "weight_mean": 1.7006928828777745e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9796087144641206e-05 }, { "epoch": 0.22067175077072854, "step": 680, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7740347175276838e-05, "weight_min": 0.0, "weight_std": 1.9367036657058634e-05 }, { "epoch": 0.22067175077072854, "step": 680, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2324653653195128e-05, "weight_min": 0.0, "weight_std": 1.765656634233892e-05 }, { "epoch": 0.22099626805127373, "step": 681, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2600533409568015e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.92349234566791e-05 }, { "epoch": 0.22099626805127373, "step": 681, "weight_max": 4.825951638309789e-09, "weight_mean": 1.3343746108063215e-09, "weight_min": 6.488223291967188e-10, "weight_std": 1.448555830840803e-09 }, { "epoch": 0.22099626805127373, "step": 681, "weight_max": 4.252122016623616e-05, "weight_mean": 8.127094588417094e-06, "weight_min": 0.0, "weight_std": 1.597087612026371e-05 }, { "epoch": 0.22099626805127373, "step": 681, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998528766009258e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4135952369542792e-05 }, { "epoch": 0.22132078533181893, "step": 682, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315710950526409e-06, "weight_min": 0.0, "weight_std": 1.5033297131594736e-05 }, { "epoch": 0.22132078533181893, "step": 682, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0153299626836088e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.879853334685322e-05 }, { "epoch": 0.22132078533181893, "step": 682, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9326675101183355e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1341173123801127e-05 }, { "epoch": 0.22132078533181893, "step": 682, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4125240340945311e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9559964130166918e-05 }, { "epoch": 0.22164530261236412, "step": 683, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1541758794919588e-05, "weight_min": 0.0, "weight_std": 1.8647948309080675e-05 }, { "epoch": 0.22164530261236412, "step": 683, "weight_max": 3.843663216684945e-05, "weight_mean": 9.384043551108334e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7378102711518295e-05 }, { "epoch": 0.22164530261236412, "step": 683, "weight_max": 3.843663216684945e-05, "weight_mean": 1.7435664631193504e-05, "weight_min": 9.178617865757133e-10, "weight_std": 1.8959144654218107e-05 }, { "epoch": 0.22164530261236412, "step": 683, "weight_max": 4.252122016623616e-05, "weight_mean": 1.516550219093915e-05, "weight_min": 0.0, "weight_std": 2.0993622456444427e-05 }, { "epoch": 0.2219698198929093, "step": 684, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1920882570848335e-05, "weight_min": 0.0, "weight_std": 1.9223507479182445e-05 }, { "epoch": 0.2219698198929093, "step": 684, "weight_max": 3.813097282545641e-05, "weight_mean": 4.766933216160396e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3481107998813968e-05 }, { "epoch": 0.2219698198929093, "step": 684, "weight_max": 4.187939703115262e-05, "weight_mean": 1.039067683450412e-05, "weight_min": 0.0, "weight_std": 1.923944910231512e-05 }, { "epoch": 0.2219698198929093, "step": 684, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.725281415740028e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.800796962925233e-05 }, { "epoch": 0.22229433717345448, "step": 685, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4030548300070222e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9513194274622947e-05 }, { "epoch": 0.22229433717345448, "step": 685, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805293883691775e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3589113223133609e-05 }, { "epoch": 0.22229433717345448, "step": 685, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.465708899369929e-06, "weight_min": 0.0, "weight_std": 1.7754358850652352e-05 }, { "epoch": 0.22229433717345448, "step": 685, "weight_max": 3.632264997577295e-05, "weight_mean": 7.440800800395664e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4215679584594909e-05 }, { "epoch": 0.22261885445399968, "step": 686, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4265411664382555e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9795103071373887e-05 }, { "epoch": 0.22261885445399968, "step": 686, "weight_max": 4.447596802492626e-05, "weight_mean": 1.542240715934895e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1432228095363826e-05 }, { "epoch": 0.22261885445399968, "step": 686, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3002279956708662e-05, "weight_min": 0.0, "weight_std": 1.8127051589544863e-05 }, { "epoch": 0.22261885445399968, "step": 686, "weight_max": 4.381950566312298e-05, "weight_mean": 2.0550951376208104e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.200651761086192e-05 }, { "epoch": 0.22294337173454487, "step": 687, "weight_max": 4.252122016623616e-05, "weight_mean": 1.894037995953113e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.041011794062797e-05 }, { "epoch": 0.22294337173454487, "step": 687, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4307412129710428e-05, "weight_min": 0.0, "weight_std": 1.9780665752477944e-05 }, { "epoch": 0.22294337173454487, "step": 687, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0401183317299001e-05, "weight_min": 0.0, "weight_std": 1.924201933434233e-05 }, { "epoch": 0.22294337173454487, "step": 687, "weight_max": 3.998322426923551e-05, "weight_mean": 1.346382123301737e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.870327105280012e-05 }, { "epoch": 0.22326788901509007, "step": 688, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0278338777425233e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9061342754866928e-05 }, { "epoch": 0.22326788901509007, "step": 688, "weight_max": 2.538973967602942e-05, "weight_mean": 3.1744741590955527e-06, "weight_min": 5.679522963042416e-10, "weight_std": 8.976322533271741e-06 }, { "epoch": 0.22326788901509007, "step": 688, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311220648873132e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.909117418108508e-05 }, { "epoch": 0.22326788901509007, "step": 688, "weight_max": 1.1016275669462061e-09, "weight_mean": 6.985957923255626e-10, "weight_min": 0.0, "weight_std": 3.371919132799661e-10 }, { "epoch": 0.22359240629563523, "step": 689, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0117270903720055e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8746575733530335e-05 }, { "epoch": 0.22359240629563523, "step": 689, "weight_max": 3.340639523230493e-05, "weight_mean": 7.968625141074881e-06, "weight_min": 0.0, "weight_std": 1.4776803254790138e-05 }, { "epoch": 0.22359240629563523, "step": 689, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7499256500741467e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9034354409086518e-05 }, { "epoch": 0.22359240629563523, "step": 689, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396869255491765e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5261808584909886e-05 }, { "epoch": 0.22391692357618043, "grad_norm": 0.4389163553714752, "learning_rate": 8.8e-06, "loss": 0.234, "step": 690 }, { "epoch": 0.22391692357618043, "step": 690, "weight_max": 7.194614903838215e-10, "weight_mean": 6.7000688330765e-10, "weight_min": 5.864964625068581e-10, "weight_std": 4.719450177881335e-11 }, { "epoch": 0.22391692357618043, "step": 690, "weight_max": 3.8743506593164057e-05, "weight_mean": 5.496855919773225e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.355721087747952e-05 }, { "epoch": 0.22391692357618043, "step": 690, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1333908080123365e-05, "weight_min": 0.0, "weight_std": 1.8248909327667207e-05 }, { "epoch": 0.22391692357618043, "step": 690, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0772518989106175e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4357561667566188e-05 }, { "epoch": 0.22424144085672562, "step": 691, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.93586729944218e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6555542970309034e-05 }, { "epoch": 0.22424144085672562, "step": 691, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.055018381332047e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6819625670905225e-05 }, { "epoch": 0.22424144085672562, "step": 691, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.881987933913479e-06, "weight_min": 0.0, "weight_std": 1.3806610695610289e-05 }, { "epoch": 0.22424144085672562, "step": 691, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2114385754102841e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7917158402269706e-05 }, { "epoch": 0.22456595813727082, "step": 692, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7091417248593643e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.936973239935469e-05 }, { "epoch": 0.22456595813727082, "step": 692, "weight_max": 3.813097282545641e-05, "weight_mean": 9.086889804166276e-06, "weight_min": 0.0, "weight_std": 1.6850906831678003e-05 }, { "epoch": 0.22456595813727082, "step": 692, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.2032727429177612e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9975634131697007e-05 }, { "epoch": 0.22456595813727082, "step": 692, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.433023478777613e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7532911442685872e-05 }, { "epoch": 0.224890475417816, "step": 693, "weight_max": 1.5496667282732801e-09, "weight_mean": 9.332371542214446e-10, "weight_min": 5.864964625068581e-10, "weight_std": 3.4584229924305987e-10 }, { "epoch": 0.224890475417816, "step": 693, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.303406502818689e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5655385141144507e-05 }, { "epoch": 0.224890475417816, "step": 693, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.937575330492109e-06, "weight_min": 0.0, "weight_std": 1.675074963713996e-05 }, { "epoch": 0.224890475417816, "step": 693, "weight_max": 4.252122016623616e-05, "weight_mean": 8.520941264578141e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6408075680374168e-05 }, { "epoch": 0.22521499269836118, "step": 694, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9281798813608475e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0703675545519218e-05 }, { "epoch": 0.22521499269836118, "step": 694, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156136012374191e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581077266484499e-05 }, { "epoch": 0.22521499269836118, "step": 694, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4397522793151438e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8920181901194155e-05 }, { "epoch": 0.22521499269836118, "step": 694, "weight_max": 3.998322426923551e-05, "weight_mean": 4.999487373424927e-06, "weight_min": 0.0, "weight_std": 1.4135564924799837e-05 }, { "epoch": 0.22553950997890637, "step": 695, "weight_max": 4.187939703115262e-05, "weight_mean": 5.246305590844713e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.48020226333756e-05 }, { "epoch": 0.22553950997890637, "step": 695, "weight_max": 4.187939703115262e-05, "weight_mean": 8.22503261588281e-06, "weight_min": 0.0, "weight_std": 1.5967620129231364e-05 }, { "epoch": 0.22553950997890637, "step": 695, "weight_max": 4.187939703115262e-05, "weight_mean": 8.472034096485004e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6256752132903785e-05 }, { "epoch": 0.22553950997890637, "step": 695, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.302778946410399e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.8270109649165533e-05 }, { "epoch": 0.22586402725945157, "step": 696, "weight_max": 4.252122016623616e-05, "weight_mean": 5.322716788214166e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.503047769801924e-05 }, { "epoch": 0.22586402725945157, "step": 696, "weight_max": 2.3357602074725037e-08, "weight_mean": 4.085501714712336e-09, "weight_min": 6.100955851628953e-10, "weight_std": 7.942663415860807e-09 }, { "epoch": 0.22586402725945157, "step": 696, "weight_max": 4.187939703115262e-05, "weight_mean": 1.548089858260937e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.1336674763006158e-05 }, { "epoch": 0.22586402725945157, "step": 696, "weight_max": 1.2882609423670033e-09, "weight_mean": 6.472827829284711e-10, "weight_min": 0.0, "weight_std": 4.541825804338373e-10 }, { "epoch": 0.22618854453999676, "step": 697, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.158996827958617e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4579922208213247e-05 }, { "epoch": 0.22618854453999676, "step": 697, "weight_max": 3.1715208024252206e-05, "weight_mean": 4.56361658507376e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.109745880967239e-05 }, { "epoch": 0.22618854453999676, "step": 697, "weight_max": 4.252122016623616e-05, "weight_mean": 5.689838417310966e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4918813576514367e-05 }, { "epoch": 0.22618854453999676, "step": 697, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0146755812456831e-05, "weight_min": 0.0, "weight_std": 1.8155820725951344e-05 }, { "epoch": 0.22651306182054196, "step": 698, "weight_max": 3.813097282545641e-05, "weight_mean": 8.731521120353136e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6256470189546235e-05 }, { "epoch": 0.22651306182054196, "step": 698, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.862068297166843e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0919352462224197e-05 }, { "epoch": 0.22651306182054196, "step": 698, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.98308644030476e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1484167518792674e-05 }, { "epoch": 0.22651306182054196, "step": 698, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.487259037792683e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6190262613235973e-05 }, { "epoch": 0.22683757910108712, "step": 699, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.968027603055816e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5234645616146736e-05 }, { "epoch": 0.22683757910108712, "step": 699, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.742901627556421e-06, "weight_min": 0.0, "weight_std": 1.6333453459083103e-05 }, { "epoch": 0.22683757910108712, "step": 699, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.506313290330581e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7654952898737974e-05 }, { "epoch": 0.22683757910108712, "step": 699, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5548663213849068e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1469275452545844e-05 }, { "epoch": 0.22716209638163232, "grad_norm": 12.52347183227539, "learning_rate": 8.779591836734694e-06, "loss": 0.2564, "step": 700 }, { "epoch": 0.22716209638163232, "eval_accuracy": 0.494, "eval_auc": 0.7670994623225744, "eval_loss": 2.163256883621216, "eval_runtime": 156.0999, "eval_samples_per_second": 12.812, "eval_steps_per_second": 1.602, "step": 700 }, { "epoch": 0.22716209638163232, "step": 700, "weight_max": 4.252122016623616e-05, "weight_mean": 1.993380101339426e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.1392526832642034e-05 }, { "epoch": 0.22716209638163232, "step": 700, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3505405149771832e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8775857824948616e-05 }, { "epoch": 0.22716209638163232, "step": 700, "weight_max": 3.998322426923551e-05, "weight_mean": 2.4898086849134415e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6128929928527214e-05 }, { "epoch": 0.22716209638163232, "step": 700, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316073838912416e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5033148883958347e-05 }, { "epoch": 0.2274866136621775, "step": 701, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3171061153989285e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5032732335384935e-05 }, { "epoch": 0.2274866136621775, "step": 701, "weight_max": 3.967146767536178e-05, "weight_mean": 9.260986189474352e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6341409718734212e-05 }, { "epoch": 0.2274866136621775, "step": 701, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.77125182544114e-06, "weight_min": 0.0, "weight_std": 1.812751725083217e-05 }, { "epoch": 0.2274866136621775, "step": 701, "weight_max": 4.187939703115262e-05, "weight_mean": 1.709733987809159e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8726250345935114e-05 }, { "epoch": 0.2278111309427227, "step": 702, "weight_max": 4.187939703115262e-05, "weight_mean": 5.328095994627802e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4771122550882865e-05 }, { "epoch": 0.2278111309427227, "step": 702, "weight_max": 3.5141503758495674e-05, "weight_mean": 5.931641680945177e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.2562919437186792e-05 }, { "epoch": 0.2278111309427227, "step": 702, "weight_max": 3.782653220696375e-05, "weight_mean": 8.693134077475406e-06, "weight_min": 0.0, "weight_std": 1.6178188161575235e-05 }, { "epoch": 0.2278111309427227, "step": 702, "weight_max": 2.7679425329552032e-05, "weight_mean": 5.877944204257801e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.100028839573497e-05 }, { "epoch": 0.2281356482232679, "step": 703, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.6549147403566167e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.809467539715115e-05 }, { "epoch": 0.2281356482232679, "step": 703, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0114859833265655e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8732258467935026e-05 }, { "epoch": 0.2281356482232679, "step": 703, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3252329154056497e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8445338355377316e-05 }, { "epoch": 0.2281356482232679, "step": 703, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.329695785709191e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.565891579957679e-05 }, { "epoch": 0.22846016550381307, "step": 704, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.90566097444389e-05, "weight_min": 0.0, "weight_std": 2.0916981156915426e-05 }, { "epoch": 0.22846016550381307, "step": 704, "weight_max": 4.252122016623616e-05, "weight_mean": 8.743842954572756e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6683094145264477e-05 }, { "epoch": 0.22846016550381307, "step": 704, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4516862393065821e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0145738744759e-05 }, { "epoch": 0.22846016550381307, "step": 704, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4656371604360174e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0275865608709864e-05 }, { "epoch": 0.22878468278435826, "step": 705, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.884432984108571e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.8323173208045773e-05 }, { "epoch": 0.22878468278435826, "step": 705, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745214811118785e-06, "weight_min": 0.0, "weight_std": 1.8142047338187695e-05 }, { "epoch": 0.22878468278435826, "step": 705, "weight_max": 3.22740561387036e-05, "weight_mean": 8.051355507632252e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.4696096513944212e-05 }, { "epoch": 0.22878468278435826, "step": 705, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0078539162350353e-05, "weight_min": 0.0, "weight_std": 1.867897117335815e-05 }, { "epoch": 0.22910920006490346, "step": 706, "weight_max": 1.582971620450735e-08, "weight_mean": 2.760331963003182e-09, "weight_min": 0.0, "weight_std": 5.345772269294002e-09 }, { "epoch": 0.22910920006490346, "step": 706, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155813141755061e-06, "weight_min": 0.0, "weight_std": 1.4581207324226853e-05 }, { "epoch": 0.22910920006490346, "step": 706, "weight_max": 2.952862632810138e-05, "weight_mean": 3.6917608667863533e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0439670404593926e-05 }, { "epoch": 0.22910920006490346, "step": 706, "weight_max": 3.813097282545641e-05, "weight_mean": 1.3296556062414311e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8481332517694682e-05 }, { "epoch": 0.22943371734544865, "step": 707, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4700760402774904e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0522818886092864e-05 }, { "epoch": 0.22943371734544865, "step": 707, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6180045349756256e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8313317923457362e-05 }, { "epoch": 0.22943371734544865, "step": 707, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235778644419042e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4806258150201757e-05 }, { "epoch": 0.22943371734544865, "step": 707, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235506250755861e-06, "weight_min": 0.0, "weight_std": 1.480636728956597e-05 }, { "epoch": 0.22975823462599385, "step": 708, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0094574210816063e-05, "weight_min": 0.0, "weight_std": 2.15856507566059e-05 }, { "epoch": 0.22975823462599385, "step": 708, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631396435201168e-05, "weight_min": 0.0, "weight_std": 1.9687344320118427e-05 }, { "epoch": 0.22975823462599385, "step": 708, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3893048162572086e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9471348423394375e-05 }, { "epoch": 0.22975823462599385, "step": 708, "weight_max": 4.252122016623616e-05, "weight_mean": 1.462989894207567e-05, "weight_min": 0.0, "weight_std": 2.0450128431548364e-05 }, { "epoch": 0.230082751906539, "step": 709, "weight_max": 3.199401908204891e-05, "weight_mean": 4.000636181444861e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1311034540995024e-05 }, { "epoch": 0.230082751906539, "step": 709, "weight_max": 3.662098970380612e-05, "weight_mean": 4.8713864089222625e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.285497910430422e-05 }, { "epoch": 0.230082751906539, "step": 709, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.0737221600720659e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.6562455130042508e-05 }, { "epoch": 0.230082751906539, "step": 709, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.417659384780563e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.737566344672814e-05 }, { "epoch": 0.2304072691870842, "grad_norm": 0.8096270561218262, "learning_rate": 8.759183673469389e-06, "loss": 0.1727, "step": 710 }, { "epoch": 0.2304072691870842, "step": 710, "weight_max": 1.5955802235012584e-09, "weight_mean": 7.270916091428603e-10, "weight_min": 5.679522963042416e-10, "weight_std": 3.5230549033649083e-10 }, { "epoch": 0.2304072691870842, "step": 710, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.32831610244466e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8456776160746813e-05 }, { "epoch": 0.2304072691870842, "step": 710, "weight_max": 3.199401908204891e-05, "weight_mean": 7.624210866197245e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.413835161656607e-05 }, { "epoch": 0.2304072691870842, "step": 710, "weight_max": 3.116124207735993e-05, "weight_mean": 5.921899173699785e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1669398190861102e-05 }, { "epoch": 0.2307317864676294, "step": 711, "weight_max": 4.252122016623616e-05, "weight_mean": 1.466322828491684e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.031750227615703e-05 }, { "epoch": 0.2307317864676294, "step": 711, "weight_max": 4.252122016623616e-05, "weight_mean": 1.012140819511842e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.876884198281914e-05 }, { "epoch": 0.2307317864676294, "step": 711, "weight_max": 3.3961886636291183e-09, "weight_mean": 1.3876471083307251e-09, "weight_min": 7.508524357824342e-10, "weight_std": 9.262363098727633e-10 }, { "epoch": 0.2307317864676294, "step": 711, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0290153770474717e-05, "weight_min": 0.0, "weight_std": 2.16936859942507e-05 }, { "epoch": 0.2310563037481746, "step": 712, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155556083191186e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.881520620372612e-05 }, { "epoch": 0.2310563037481746, "step": 712, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.804350156628061e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7307529560639523e-05 }, { "epoch": 0.2310563037481746, "step": 712, "weight_max": 3.813097282545641e-05, "weight_mean": 1.337671528744977e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.852281275205314e-05 }, { "epoch": 0.2310563037481746, "step": 712, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7980844859266654e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.985767812584527e-05 }, { "epoch": 0.2313808210287198, "step": 713, "weight_max": 3.967146767536178e-05, "weight_mean": 9.841585779213347e-06, "weight_min": 0.0, "weight_std": 1.8220862330053933e-05 }, { "epoch": 0.2313808210287198, "step": 713, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.32518890092615e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5455878383363597e-05 }, { "epoch": 0.2313808210287198, "step": 713, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.1185669791302644e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.5786928997840732e-05 }, { "epoch": 0.2313808210287198, "step": 713, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0117003512277734e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8746739442576654e-05 }, { "epoch": 0.23170533830926496, "step": 714, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4002685020386707e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9352904928382486e-05 }, { "epoch": 0.23170533830926496, "step": 714, "weight_max": 3.340639523230493e-05, "weight_mean": 6.104406566009857e-06, "weight_min": 0.0, "weight_std": 1.2281062481633853e-05 }, { "epoch": 0.23170533830926496, "step": 714, "weight_max": 3.116124207735993e-05, "weight_mean": 3.8960911297181156e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1016784810635727e-05 }, { "epoch": 0.23170533830926496, "step": 714, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.29593534237938e-06, "weight_min": 0.0, "weight_std": 1.73479420482181e-05 }, { "epoch": 0.23202985558981015, "step": 715, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9049060938414186e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0344943550298922e-05 }, { "epoch": 0.23202985558981015, "step": 715, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5670651919208467e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.164913348678965e-05 }, { "epoch": 0.23202985558981015, "step": 715, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691128197009675e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3266208952700254e-05 }, { "epoch": 0.23202985558981015, "step": 715, "weight_max": 4.187939703115262e-05, "weight_mean": 1.863219222286716e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7748656318872236e-05 }, { "epoch": 0.23235437287035535, "step": 716, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.440282737050438e-06, "weight_min": 0.0, "weight_std": 1.2523774785222486e-05 }, { "epoch": 0.23235437287035535, "step": 716, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.433356555935461e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0082032278878614e-05 }, { "epoch": 0.23235437287035535, "step": 716, "weight_max": 3.255531191825867e-05, "weight_mean": 7.520490271417657e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3934955859440379e-05 }, { "epoch": 0.23235437287035535, "step": 716, "weight_max": 3.782653220696375e-05, "weight_mean": 8.799085662758444e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.635020089452155e-05 }, { "epoch": 0.23267889015090054, "step": 717, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.356957560958108e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2320752830419224e-05 }, { "epoch": 0.23267889015090054, "step": 717, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3862355444871355e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.92021961993305e-05 }, { "epoch": 0.23267889015090054, "step": 717, "weight_max": 1.783980223990511e-05, "weight_mean": 2.230708787465119e-06, "weight_min": 6.100955851628953e-10, "weight_std": 6.307026069407584e-06 }, { "epoch": 0.23267889015090054, "step": 717, "weight_max": 4.187939703115262e-05, "weight_mean": 2.362188752158545e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9803628674708307e-05 }, { "epoch": 0.23300340743144574, "step": 718, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8061440641758963e-05, "weight_min": 7.614652242082798e-10, "weight_std": 1.9663752027554438e-05 }, { "epoch": 0.23300340743144574, "step": 718, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.674545824760571e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.6228405002038926e-05 }, { "epoch": 0.23300340743144574, "step": 718, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.176497431122698e-06, "weight_min": 0.0, "weight_std": 1.5810594049980864e-05 }, { "epoch": 0.23300340743144574, "step": 718, "weight_max": 3.752332122530788e-05, "weight_mean": 8.483535566483624e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.582366348884534e-05 }, { "epoch": 0.2333279247119909, "step": 719, "weight_max": 3.28377827827353e-05, "weight_mean": 7.762749191897456e-06, "weight_min": 0.0, "weight_std": 1.440452160750283e-05 }, { "epoch": 0.2333279247119909, "step": 719, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.300710548297502e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.954186154762283e-05 }, { "epoch": 0.2333279247119909, "step": 719, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.309716915595345e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.725563924992457e-05 }, { "epoch": 0.2333279247119909, "step": 719, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2558799426187761e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8052320228889585e-05 }, { "epoch": 0.2336524419925361, "grad_norm": 7.888647556304932, "learning_rate": 8.738775510204082e-06, "loss": 0.2419, "step": 720 }, { "epoch": 0.2336524419925361, "step": 720, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4057004591450095e-05, "weight_min": 0.0, "weight_std": 1.9500081180012785e-05 }, { "epoch": 0.2336524419925361, "step": 720, "weight_max": 3.782653220696375e-05, "weight_mean": 9.122808478423394e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6902944480534643e-05 }, { "epoch": 0.2336524419925361, "step": 720, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.420674379914999e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.63348886417225e-05 }, { "epoch": 0.2336524419925361, "step": 720, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7866203052108176e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9262024579802528e-05 }, { "epoch": 0.2339769592730813, "step": 721, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.881498615664896e-06, "weight_min": 0.0, "weight_std": 1.672378130024299e-05 }, { "epoch": 0.2339769592730813, "step": 721, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4434900549531449e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.000555286940653e-05 }, { "epoch": 0.2339769592730813, "step": 721, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.121032806229778e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5118422197701875e-05 }, { "epoch": 0.2339769592730813, "step": 721, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.025894774415065e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.411035282217199e-05 }, { "epoch": 0.2343014765536265, "step": 722, "weight_max": 4.187939703115262e-05, "weight_mean": 1.788027657312341e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.963371323654428e-05 }, { "epoch": 0.2343014765536265, "step": 722, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.728567440295592e-05, "weight_min": 0.0, "weight_std": 1.9241291738580912e-05 }, { "epoch": 0.2343014765536265, "step": 722, "weight_max": 3.632264997577295e-05, "weight_mean": 8.333216101163998e-06, "weight_min": 0.0, "weight_std": 1.55116413225187e-05 }, { "epoch": 0.2343014765536265, "step": 722, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156182851351332e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4581058167095762e-05 }, { "epoch": 0.23462599383417168, "step": 723, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212085514154751e-06, "weight_min": 0.0, "weight_std": 1.1911896763194818e-05 }, { "epoch": 0.23462599383417168, "step": 723, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2521809367171954e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7739273971528746e-05 }, { "epoch": 0.23462599383417168, "step": 723, "weight_max": 3.722132169059478e-05, "weight_mean": 8.214904482883867e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5176867236732505e-05 }, { "epoch": 0.23462599383417168, "step": 723, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.041283192345873e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6907737517612986e-05 }, { "epoch": 0.23495051111471685, "step": 724, "weight_max": 1.7053184420134926e-09, "weight_mean": 7.732600115772925e-10, "weight_min": 0.0, "weight_std": 4.670432929287927e-10 }, { "epoch": 0.23495051111471685, "step": 724, "weight_max": 3.143761568935588e-05, "weight_mean": 7.654364708287176e-06, "weight_min": 0.0, "weight_std": 1.3718180525756907e-05 }, { "epoch": 0.23495051111471685, "step": 724, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0195471986662596e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8883583834394813e-05 }, { "epoch": 0.23495051111471685, "step": 724, "weight_max": 3.602553260861896e-05, "weight_mean": 4.94999039801769e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.261839770450024e-05 }, { "epoch": 0.23527502839526204, "step": 725, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3320752259460278e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8828075553756207e-05 }, { "epoch": 0.23527502839526204, "step": 725, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615636498783715e-06, "weight_min": 0.0, "weight_std": 1.3053153452347033e-05 }, { "epoch": 0.23527502839526204, "step": 725, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.424071888322942e-05, "weight_min": 0.0, "weight_std": 1.9726247046492063e-05 }, { "epoch": 0.23527502839526204, "step": 725, "weight_max": 3.22740561387036e-05, "weight_mean": 1.4249222658690996e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.543995131214615e-05 }, { "epoch": 0.23559954567580724, "step": 726, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.1352609362802468e-05, "weight_min": 0.0, "weight_std": 1.640522896195762e-05 }, { "epoch": 0.23559954567580724, "step": 726, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.883193469140679e-06, "weight_min": 0.0, "weight_std": 1.380612411594484e-05 }, { "epoch": 0.23559954567580724, "step": 726, "weight_max": 1.8352018793521552e-09, "weight_mean": 7.659801126713717e-10, "weight_min": 0.0, "weight_std": 5.166325700578511e-10 }, { "epoch": 0.23559954567580724, "step": 726, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2421837709553074e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7961674529942684e-05 }, { "epoch": 0.23592406295635243, "step": 727, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7476499124313705e-05, "weight_min": 0.0, "weight_std": 1.9460569092188962e-05 }, { "epoch": 0.23592406295635243, "step": 727, "weight_max": 4.187939703115262e-05, "weight_mean": 5.238287030806532e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4805245882598683e-05 }, { "epoch": 0.23592406295635243, "step": 727, "weight_max": 3.967146767536178e-05, "weight_mean": 8.291112862934824e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5739329683128744e-05 }, { "epoch": 0.23592406295635243, "step": 727, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843584065383766e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.369763685943326e-05 }, { "epoch": 0.23624858023689763, "step": 728, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4445638953475282e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0225688786013052e-05 }, { "epoch": 0.23624858023689763, "step": 728, "weight_max": 4.252122016623616e-05, "weight_mean": 1.330304985458497e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8739399820333347e-05 }, { "epoch": 0.23624858023689763, "step": 728, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1998504204675555e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4563916920451447e-05 }, { "epoch": 0.23624858023689763, "step": 728, "weight_max": 3.255531191825867e-05, "weight_mean": 4.0796671783027705e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1505926522659138e-05 }, { "epoch": 0.2365730975174428, "step": 729, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.357708793482743e-06, "weight_min": 0.0, "weight_std": 1.5501011148444377e-05 }, { "epoch": 0.2365730975174428, "step": 729, "weight_max": 3.199401908204891e-05, "weight_mean": 7.792266842443496e-06, "weight_min": 0.0, "weight_std": 1.443380824639462e-05 }, { "epoch": 0.2365730975174428, "step": 729, "weight_max": 3.998322426923551e-05, "weight_mean": 9.996410881285556e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8508262655814178e-05 }, { "epoch": 0.2365730975174428, "step": 729, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.157098257768666e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.458068800275214e-05 }, { "epoch": 0.236897614797988, "grad_norm": 7.360884189605713, "learning_rate": 8.718367346938777e-06, "loss": 0.1725, "step": 730 }, { "epoch": 0.236897614797988, "step": 730, "weight_max": 3.782653220696375e-05, "weight_mean": 8.728227840038016e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.6234918803093024e-05 }, { "epoch": 0.236897614797988, "step": 730, "weight_max": 8.508121318584472e-09, "weight_mean": 1.5812716691598894e-09, "weight_min": 0.0, "weight_std": 2.829395828740644e-09 }, { "epoch": 0.236897614797988, "step": 730, "weight_max": 3.998322426923551e-05, "weight_mean": 9.99636176857166e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.850829357863404e-05 }, { "epoch": 0.236897614797988, "step": 730, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316564966051374e-06, "weight_min": 0.0, "weight_std": 1.503295061411336e-05 }, { "epoch": 0.23722213207853318, "step": 731, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6753963791416027e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.887699545477517e-05 }, { "epoch": 0.23722213207853318, "step": 731, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3552203199651558e-05, "weight_min": 0.0, "weight_std": 1.8881595678976737e-05 }, { "epoch": 0.23722213207853318, "step": 731, "weight_max": 3.061215829802677e-05, "weight_mean": 7.81236303737387e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3270027011458296e-05 }, { "epoch": 0.23722213207853318, "step": 731, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5331770555349067e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.7760388800525106e-05 }, { "epoch": 0.23754664935907838, "step": 732, "weight_max": 1.783980223990511e-05, "weight_mean": 2.2313508907245705e-06, "weight_min": 5.864964625068581e-10, "weight_std": 6.30676731816493e-06 }, { "epoch": 0.23754664935907838, "step": 732, "weight_max": 4.252122016623616e-05, "weight_mean": 1.421837805537507e-05, "weight_min": 0.0, "weight_std": 1.9917923054890707e-05 }, { "epoch": 0.23754664935907838, "step": 732, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5078549040481448e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0836811017943546e-05 }, { "epoch": 0.23754664935907838, "step": 732, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4619101420976222e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0229163055773824e-05 }, { "epoch": 0.23787116663962357, "step": 733, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9751749277929775e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.124596903740894e-05 }, { "epoch": 0.23787116663962357, "step": 733, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691061803896446e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3266235328046605e-05 }, { "epoch": 0.23787116663962357, "step": 733, "weight_max": 3.2623432844047784e-08, "weight_mean": 5.129962676875266e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.1150033785156666e-08 }, { "epoch": 0.23787116663962357, "step": 733, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9985951591224875e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.413592508470174e-05 }, { "epoch": 0.23819568392016874, "step": 734, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0128780559170991e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8359092791797593e-05 }, { "epoch": 0.23819568392016874, "step": 734, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.857870302046649e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.002737528528087e-05 }, { "epoch": 0.23819568392016874, "step": 734, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8837123207049444e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.057452547887806e-05 }, { "epoch": 0.23819568392016874, "step": 734, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3583674192195758e-05, "weight_min": 1.0509849657225345e-09, "weight_std": 1.9919096303055994e-05 }, { "epoch": 0.23852020120071393, "step": 735, "weight_max": 3.397988257347606e-05, "weight_mean": 1.11333738459507e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.561189492349513e-05 }, { "epoch": 0.23852020120071393, "step": 735, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.8516355087049305e-05, "weight_min": 0.0, "weight_std": 1.7757029127096757e-05 }, { "epoch": 0.23852020120071393, "step": 735, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4097349523799494e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.011404103541281e-05 }, { "epoch": 0.23852020120071393, "step": 735, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861310688080266e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0919658052443992e-05 }, { "epoch": 0.23884471848125913, "step": 736, "weight_max": 4.252122016623616e-05, "weight_mean": 9.202096407534555e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.725695801724214e-05 }, { "epoch": 0.23884471848125913, "step": 736, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2798451280104928e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7916192518896423e-05 }, { "epoch": 0.23884471848125913, "step": 736, "weight_max": 3.998322426923551e-05, "weight_mean": 6.5579015426919796e-06, "weight_min": 0.0, "weight_std": 1.4193843526300043e-05 }, { "epoch": 0.23884471848125913, "step": 736, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.219629307859577e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.576976319483947e-05 }, { "epoch": 0.23916923576180432, "step": 737, "weight_max": 3.397988257347606e-05, "weight_mean": 4.549602181214141e-06, "weight_min": 7.937498991417158e-10, "weight_std": 1.192124454973964e-05 }, { "epoch": 0.23916923576180432, "step": 737, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5197052562143654e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0977928215870634e-05 }, { "epoch": 0.23916923576180432, "step": 737, "weight_max": 3.8743506593164057e-05, "weight_mean": 6.652512183791259e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3920000128564425e-05 }, { "epoch": 0.23916923576180432, "step": 737, "weight_max": 4.252122016623616e-05, "weight_mean": 1.055079155776184e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9535258616087958e-05 }, { "epoch": 0.23949375304234952, "step": 738, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.56279875227483e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1579562599072233e-05 }, { "epoch": 0.23949375304234952, "step": 738, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232170097879134e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8945662304759026e-05 }, { "epoch": 0.23949375304234952, "step": 738, "weight_max": 3.998322426923551e-05, "weight_mean": 5.008015250496101e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.413213976775296e-05 }, { "epoch": 0.23949375304234952, "step": 738, "weight_max": 4.187939703115262e-05, "weight_mean": 9.850448805082124e-06, "weight_min": 0.0, "weight_std": 1.828645144996699e-05 }, { "epoch": 0.23981827032289468, "step": 739, "weight_max": 3.602553260861896e-05, "weight_mean": 9.073570254258811e-06, "weight_min": 7.508524357824342e-10, "weight_std": 1.595790672581643e-05 }, { "epoch": 0.23981827032289468, "step": 739, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3666382074006833e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.919466558319982e-05 }, { "epoch": 0.23981827032289468, "step": 739, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0116862540598959e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8746826754068024e-05 }, { "epoch": 0.23981827032289468, "step": 739, "weight_max": 2.8055460177256464e-08, "weight_mean": 6.700910937240678e-09, "weight_min": 0.0, "weight_std": 1.013194861343436e-08 }, { "epoch": 0.24014278760343988, "grad_norm": 3.975144624710083, "learning_rate": 8.69795918367347e-06, "loss": 0.3085, "step": 740 }, { "epoch": 0.24014278760343988, "step": 740, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.847353920922615e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.3696117093786597e-05 }, { "epoch": 0.24014278760343988, "step": 740, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653498763218522e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3159387890482321e-05 }, { "epoch": 0.24014278760343988, "step": 740, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0707095498219132e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8357026419835165e-05 }, { "epoch": 0.24014278760343988, "step": 740, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959395937476074e-06, "weight_min": 0.0, "weight_std": 1.4025793461769354e-05 }, { "epoch": 0.24046730488398507, "step": 741, "weight_max": 3.340639523230493e-05, "weight_mean": 4.176671154709766e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1810591786343139e-05 }, { "epoch": 0.24046730488398507, "step": 741, "weight_max": 9.652274535199012e-10, "weight_mean": 7.38493211027702e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.652477182867429e-10 }, { "epoch": 0.24046730488398507, "step": 741, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.0128665962838568e-05, "weight_min": 0.0, "weight_std": 1.7268010196858086e-05 }, { "epoch": 0.24046730488398507, "step": 741, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8236836694995873e-05, "weight_min": 0.0, "weight_std": 2.065527041850146e-05 }, { "epoch": 0.24079182216453027, "step": 742, "weight_max": 3.22740561387036e-05, "weight_mean": 4.0347235881199595e-06, "weight_min": 0.0, "weight_std": 1.141041411756305e-05 }, { "epoch": 0.24079182216453027, "step": 742, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3662443961948156e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9597253412939608e-05 }, { "epoch": 0.24079182216453027, "step": 742, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.18312673800392e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5177592104009818e-05 }, { "epoch": 0.24079182216453027, "step": 742, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882476787315682e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3806413335260004e-05 }, { "epoch": 0.24111633944507546, "step": 743, "weight_max": 3.998322426923551e-05, "weight_mean": 9.804791261558421e-06, "weight_min": 0.0, "weight_std": 1.8153952623833902e-05 }, { "epoch": 0.24111633944507546, "step": 743, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235637217992917e-06, "weight_min": 0.0, "weight_std": 1.4806315448367968e-05 }, { "epoch": 0.24111633944507546, "step": 743, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.484219501435291e-05, "weight_min": 0.0, "weight_std": 2.0500410755630583e-05 }, { "epoch": 0.24111633944507546, "step": 743, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.846958861686289e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.070564758265391e-05 }, { "epoch": 0.24144085672562063, "step": 744, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4645689589087851e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.025819594564382e-05 }, { "epoch": 0.24144085672562063, "step": 744, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4301505871117115e-05, "weight_min": 0.0, "weight_std": 1.9905954104615375e-05 }, { "epoch": 0.24144085672562063, "step": 744, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.396801033290103e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.74727520061424e-05 }, { "epoch": 0.24144085672562063, "step": 744, "weight_max": 5.564223556575598e-06, "weight_mean": 6.967966328375041e-07, "weight_min": 6.586902134841921e-10, "weight_std": 1.966737954717246e-06 }, { "epoch": 0.24176537400616582, "step": 745, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.533821867080405e-06, "weight_min": 0.0, "weight_std": 1.765506385709159e-05 }, { "epoch": 0.24176537400616582, "step": 745, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4899663256073836e-05, "weight_min": 0.0, "weight_std": 2.0658462744904682e-05 }, { "epoch": 0.24176537400616582, "step": 745, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.695878932485357e-06, "weight_min": 0.0, "weight_std": 1.671281825110782e-05 }, { "epoch": 0.24176537400616582, "step": 745, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2065750524925534e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8526920030126348e-05 }, { "epoch": 0.24208989128671102, "step": 746, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4442692190641537e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.9967845219071023e-05 }, { "epoch": 0.24208989128671102, "step": 746, "weight_max": 7.614652242082798e-10, "weight_mean": 6.743590130753319e-10, "weight_min": 5.864964625068581e-10, "weight_std": 6.454219797502603e-11 }, { "epoch": 0.24208989128671102, "step": 746, "weight_max": 3.998322426923551e-05, "weight_mean": 2.2836131392978132e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9159795556333847e-05 }, { "epoch": 0.24208989128671102, "step": 746, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23544395036879e-06, "weight_min": 0.0, "weight_std": 1.4806392755417619e-05 }, { "epoch": 0.2424144085672562, "step": 747, "weight_max": 4.187939703115262e-05, "weight_mean": 9.519046216155402e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7741449482855387e-05 }, { "epoch": 0.2424144085672562, "step": 747, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8104079572367482e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0216877601342276e-05 }, { "epoch": 0.2424144085672562, "step": 747, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0195117283728905e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8883802113123238e-05 }, { "epoch": 0.2424144085672562, "step": 747, "weight_max": 4.187939703115262e-05, "weight_mean": 9.628207408240996e-06, "weight_min": 0.0, "weight_std": 1.7917231161845848e-05 }, { "epoch": 0.2427389258478014, "step": 748, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1559482017182745e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581152754544746e-05 }, { "epoch": 0.2427389258478014, "step": 748, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.27834573504515e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7982885765377432e-05 }, { "epoch": 0.2427389258478014, "step": 748, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.016323767544236e-05, "weight_min": 0.0, "weight_std": 1.8864589947042987e-05 }, { "epoch": 0.2427389258478014, "step": 748, "weight_max": 4.252122016623616e-05, "weight_mean": 2.063198553514667e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.209890772064682e-05 }, { "epoch": 0.24306344312834657, "step": 749, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998592430638382e-06, "weight_min": 0.0, "weight_std": 1.4135925994196441e-05 }, { "epoch": 0.24306344312834657, "step": 749, "weight_max": 3.061215829802677e-05, "weight_mean": 3.8271628000075e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0822772310348228e-05 }, { "epoch": 0.24306344312834657, "step": 749, "weight_max": 4.187939703115262e-05, "weight_mean": 9.165821211354341e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7196800399688073e-05 }, { "epoch": 0.24306344312834657, "step": 749, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3629022760142107e-05, "weight_min": 0.0, "weight_std": 1.9008917661267333e-05 }, { "epoch": 0.24338796040889177, "grad_norm": 45.465232849121094, "learning_rate": 8.679591836734695e-06, "loss": 0.3403, "step": 750 }, { "epoch": 0.24338796040889177, "step": 750, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8292405002284795e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9786775737884454e-05 }, { "epoch": 0.24338796040889177, "step": 750, "weight_max": 3.998322426923551e-05, "weight_mean": 9.138768291450106e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.7019123333739117e-05 }, { "epoch": 0.24338796040889177, "step": 750, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5869254639255814e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1913720047450624e-05 }, { "epoch": 0.24338796040889177, "step": 750, "weight_max": 1.504423474685268e-09, "weight_mean": 9.083673813137239e-10, "weight_min": 6.053385570581327e-10, "weight_std": 3.428664852034302e-10 }, { "epoch": 0.24371247768943696, "step": 751, "weight_max": 3.967146767536178e-05, "weight_mean": 9.880779543891549e-06, "weight_min": 0.0, "weight_std": 1.829156644816976e-05 }, { "epoch": 0.24371247768943696, "step": 751, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.736855847411789e-05, "weight_min": 0.0, "weight_std": 1.9082912331214175e-05 }, { "epoch": 0.24371247768943696, "step": 751, "weight_max": 1.56184469233267e-05, "weight_mean": 1.9545998384273844e-06, "weight_min": 6.196655966128617e-10, "weight_std": 5.521029379451647e-06 }, { "epoch": 0.24371247768943696, "step": 751, "weight_max": 1.3161142176087992e-09, "weight_mean": 6.253279005719037e-10, "weight_min": 0.0, "weight_std": 4.59753124459894e-10 }, { "epoch": 0.24403699496998216, "step": 752, "weight_max": 4.252122016623616e-05, "weight_mean": 1.787400287867058e-05, "weight_min": 8.490471659072796e-10, "weight_std": 1.9451203115750104e-05 }, { "epoch": 0.24403699496998216, "step": 752, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.8812161013484e-06, "weight_min": 0.0, "weight_std": 1.8304848708794452e-05 }, { "epoch": 0.24403699496998216, "step": 752, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.055772352439817e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7169772036140785e-05 }, { "epoch": 0.24403699496998216, "step": 752, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155800863576587e-06, "weight_min": 0.0, "weight_std": 1.4581211871700361e-05 }, { "epoch": 0.24436151225052735, "step": 753, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1831592928501777e-05, "weight_min": 8.267048712262692e-10, "weight_std": 1.6975853213807568e-05 }, { "epoch": 0.24436151225052735, "step": 753, "weight_max": 4.381950566312298e-05, "weight_mean": 1.713791789370589e-05, "weight_min": 0.0, "weight_std": 2.1010333512094803e-05 }, { "epoch": 0.24436151225052735, "step": 753, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.396706572151743e-06, "weight_min": 0.0, "weight_std": 1.2422778127074707e-05 }, { "epoch": 0.24436151225052735, "step": 753, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5547451766906306e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.145420148735866e-05 }, { "epoch": 0.24468602953107252, "step": 754, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.0385179595905356e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1795476641273126e-05 }, { "epoch": 0.24468602953107252, "step": 754, "weight_max": 2.8994179956498556e-05, "weight_mean": 3.783487954933662e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.0196215043833945e-05 }, { "epoch": 0.24468602953107252, "step": 754, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4174357602314558e-05, "weight_min": 0.0, "weight_std": 1.9678444004966877e-05 }, { "epoch": 0.24468602953107252, "step": 754, "weight_max": 3.752332122530788e-05, "weight_mean": 8.216144124162383e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5414711015182547e-05 }, { "epoch": 0.2450105468116177, "step": 755, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6154528566403314e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.6560468793613836e-05 }, { "epoch": 0.2450105468116177, "step": 755, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3553497157990932e-05, "weight_min": 0.0, "weight_std": 1.9930586859118193e-05 }, { "epoch": 0.2450105468116177, "step": 755, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1571107279451098e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7302529158769175e-05 }, { "epoch": 0.2450105468116177, "step": 755, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5234925740514882e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.1049245333415456e-05 }, { "epoch": 0.2453350640921629, "step": 756, "weight_max": 4.252122016623616e-05, "weight_mean": 6.070941708458122e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4878970432619099e-05 }, { "epoch": 0.2453350640921629, "step": 756, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5011664800113067e-05, "weight_min": 0.0, "weight_std": 2.0789862901438028e-05 }, { "epoch": 0.2453350640921629, "step": 756, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8071748854708858e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.0643356037908234e-05 }, { "epoch": 0.2453350640921629, "step": 756, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.7940910765901208e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9343638996360824e-05 }, { "epoch": 0.2456595813727081, "step": 757, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.940411311807111e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0873560060863383e-05 }, { "epoch": 0.2456595813727081, "step": 757, "weight_max": 1.0509849657225345e-09, "weight_mean": 6.498894200568373e-10, "weight_min": 0.0, "weight_std": 3.0606928103082964e-10 }, { "epoch": 0.2456595813727081, "step": 757, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076986781205051e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357669897435699e-05 }, { "epoch": 0.2456595813727081, "step": 757, "weight_max": 3.4268457966391e-05, "weight_mean": 4.4527746467792895e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2056568266416434e-05 }, { "epoch": 0.2459840986532533, "step": 758, "weight_max": 1.3161142176087992e-09, "weight_mean": 7.451807504388341e-10, "weight_min": 0.0, "weight_std": 4.023524291518754e-10 }, { "epoch": 0.2459840986532533, "step": 758, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7837668565334752e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0141933418926783e-05 }, { "epoch": 0.2459840986532533, "step": 758, "weight_max": 3.998322426923551e-05, "weight_mean": 5.03282808494987e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.412240271747578e-05 }, { "epoch": 0.2459840986532533, "step": 758, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3849794413545169e-05, "weight_min": 0.0, "weight_std": 1.9300170606584288e-05 }, { "epoch": 0.24630861593379846, "step": 759, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.97451388684567e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.854969741543755e-05 }, { "epoch": 0.24630861593379846, "step": 759, "weight_max": 4.187939703115262e-05, "weight_mean": 9.850696187641006e-06, "weight_min": 0.0, "weight_std": 1.828629865485709e-05 }, { "epoch": 0.24630861593379846, "step": 759, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5387422536150552e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.123643753293436e-05 }, { "epoch": 0.24630861593379846, "step": 759, "weight_max": 3.255531191825867e-05, "weight_mean": 4.07144352720934e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.150922071246896e-05 }, { "epoch": 0.24663313321434366, "grad_norm": 1.600200891494751, "learning_rate": 8.659183673469388e-06, "loss": 0.2883, "step": 760 }, { "epoch": 0.24663313321434366, "step": 760, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5048020941321738e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.083033905364573e-05 }, { "epoch": 0.24663313321434366, "step": 760, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5470692233066075e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1369143723859452e-05 }, { "epoch": 0.24663313321434366, "step": 760, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.959201108547859e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.1346095309127122e-05 }, { "epoch": 0.24663313321434366, "step": 760, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5316918506869115e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.1176285372348502e-05 }, { "epoch": 0.24695765049488885, "step": 761, "weight_max": 4.0610404539620504e-05, "weight_mean": 6.166020284581464e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4247560102376156e-05 }, { "epoch": 0.24695765049488885, "step": 761, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.695726314210333e-05, "weight_min": 0.0, "weight_std": 1.814014103729278e-05 }, { "epoch": 0.24695765049488885, "step": 761, "weight_max": 4.187939703115262e-05, "weight_mean": 1.570580388943199e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1673793526133522e-05 }, { "epoch": 0.24695765049488885, "step": 761, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3564478649641387e-05, "weight_min": 0.0, "weight_std": 1.9229388271924108e-05 }, { "epoch": 0.24728216777543405, "step": 762, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.4285258657764643e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9344774045748636e-05 }, { "epoch": 0.24728216777543405, "step": 762, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7630931324674748e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9481074559735134e-05 }, { "epoch": 0.24728216777543405, "step": 762, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4245316378946882e-05, "weight_min": 0.0, "weight_std": 1.9707809769897722e-05 }, { "epoch": 0.24728216777543405, "step": 762, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076874913356733e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357714462676086e-05 }, { "epoch": 0.24760668505597924, "step": 763, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396731467044447e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.52618667925708e-05 }, { "epoch": 0.24760668505597924, "step": 763, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805267053598072e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.358912413707003e-05 }, { "epoch": 0.24760668505597924, "step": 763, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5280872907605954e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.969632628781255e-06 }, { "epoch": 0.24760668505597924, "step": 763, "weight_max": 4.4065102677848245e-09, "weight_mean": 1.2920787773040843e-09, "weight_min": 5.864964625068581e-10, "weight_std": 1.3425087708185401e-09 }, { "epoch": 0.2479312023365244, "step": 764, "weight_max": 3.9051607018336654e-05, "weight_mean": 6.008110176480841e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3718412446905859e-05 }, { "epoch": 0.2479312023365244, "step": 764, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8504737454350106e-05, "weight_min": 0.0, "weight_std": 1.8386139345238917e-05 }, { "epoch": 0.2479312023365244, "step": 764, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0493553418200463e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.775744567567017e-05 }, { "epoch": 0.2479312023365244, "step": 764, "weight_max": 2.5017732241394697e-07, "weight_mean": 4.918656770769303e-08, "weight_min": 5.679522963042416e-10, "weight_std": 9.394690181352416e-08 }, { "epoch": 0.2482557196170696, "step": 765, "weight_max": 3.572963396436535e-05, "weight_mean": 7.486653885280248e-06, "weight_min": 0.0, "weight_std": 1.420227090420667e-05 }, { "epoch": 0.2482557196170696, "step": 765, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.7266207578359172e-05, "weight_min": 0.0, "weight_std": 1.8559019736130722e-05 }, { "epoch": 0.2482557196170696, "step": 765, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4761375496163964e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.037440390267875e-05 }, { "epoch": 0.2482557196170696, "step": 765, "weight_max": 1.344265809777312e-09, "weight_mean": 9.370901832284062e-10, "weight_min": 6.686325937366178e-10, "weight_std": 2.8201385671167145e-10 }, { "epoch": 0.2485802368976148, "step": 766, "weight_max": 3.602553260861896e-05, "weight_mean": 7.615580216224771e-06, "weight_min": 0.0, "weight_std": 1.4410266885533929e-05 }, { "epoch": 0.2485802368976148, "step": 766, "weight_max": 3.061215829802677e-05, "weight_mean": 7.6197875387151726e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4107225979387295e-05 }, { "epoch": 0.2485802368976148, "step": 766, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1257398000452667e-05, "weight_min": 0.0, "weight_std": 1.606533260201104e-05 }, { "epoch": 0.2485802368976148, "step": 766, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8056725821224973e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9329041606397368e-05 }, { "epoch": 0.24890475417816, "step": 767, "weight_max": 9.772551656794803e-10, "weight_mean": 7.455251971322241e-10, "weight_min": 5.864964625068581e-10, "weight_std": 1.4979115170454804e-10 }, { "epoch": 0.24890475417816, "step": 767, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.686402336228639e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.793443334463518e-05 }, { "epoch": 0.24890475417816, "step": 767, "weight_max": 3.752332122530788e-05, "weight_mean": 1.1443172297731508e-05, "weight_min": 0.0, "weight_std": 1.634067484701518e-05 }, { "epoch": 0.24890475417816, "step": 767, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4457694962620735e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0303181372582912e-05 }, { "epoch": 0.2492292714587052, "step": 768, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2485039405873977e-05, "weight_min": 0.0, "weight_std": 1.7358672266709618e-05 }, { "epoch": 0.2492292714587052, "step": 768, "weight_max": 3.340639523230493e-05, "weight_mean": 4.176538823230658e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1810645446530543e-05 }, { "epoch": 0.2492292714587052, "step": 768, "weight_max": 4.252122016623616e-05, "weight_mean": 1.25940614452702e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.797668846847955e-05 }, { "epoch": 0.2492292714587052, "step": 768, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.387946551607456e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0131668861722574e-05 }, { "epoch": 0.24955378873925035, "step": 769, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4036659194971435e-05, "weight_min": 0.0, "weight_std": 1.9555069229681976e-05 }, { "epoch": 0.24955378873925035, "step": 769, "weight_max": 4.187939703115262e-05, "weight_mean": 1.062893534253817e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8919894500868395e-05 }, { "epoch": 0.24955378873925035, "step": 769, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1752636964956764e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.7491831385996193e-05 }, { "epoch": 0.24955378873925035, "step": 769, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140890268899966e-06, "weight_min": 0.0, "weight_std": 1.1709927093761507e-05 }, { "epoch": 0.24987830601979555, "grad_norm": 1.3369460105895996, "learning_rate": 8.638775510204083e-06, "loss": 0.274, "step": 770 }, { "epoch": 0.24987830601979555, "step": 770, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5905650798231363e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0495850549195893e-05 }, { "epoch": 0.24987830601979555, "step": 770, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5160572729655541e-05, "weight_min": 0.0, "weight_std": 2.0966039301129058e-05 }, { "epoch": 0.24987830601979555, "step": 770, "weight_max": 3.998322426923551e-05, "weight_mean": 2.161603515560273e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8109001757693477e-05 }, { "epoch": 0.24987830601979555, "step": 770, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.899682481773198e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.842854544520378e-05 }, { "epoch": 0.25020282330034077, "step": 771, "weight_max": 4.252122016623616e-05, "weight_mean": 9.282486644224264e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7421631127945147e-05 }, { "epoch": 0.25020282330034077, "step": 771, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315919679560466e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.503321163909277e-05 }, { "epoch": 0.25020282330034077, "step": 771, "weight_max": 3.4268457966391e-05, "weight_mean": 8.845283446135e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.562553552503232e-05 }, { "epoch": 0.25020282330034077, "step": 771, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615710167854559e-06, "weight_min": 0.0, "weight_std": 1.3053124348516576e-05 }, { "epoch": 0.25052734058088594, "step": 772, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8381040717940778e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9939056073781103e-05 }, { "epoch": 0.25052734058088594, "step": 772, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5309655282180756e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1134475900908e-05 }, { "epoch": 0.25052734058088594, "step": 772, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2637778127100319e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.887174221337773e-05 }, { "epoch": 0.25052734058088594, "step": 772, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998632903152611e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4135909623291809e-05 }, { "epoch": 0.2508518578614311, "step": 773, "weight_max": 2.6141986381844617e-05, "weight_mean": 3.2685281894373475e-06, "weight_min": 6.196655966128617e-10, "weight_std": 9.242273335985374e-06 }, { "epoch": 0.2508518578614311, "step": 773, "weight_max": 3.998322426923551e-05, "weight_mean": 8.824951692076866e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6530166249140166e-05 }, { "epoch": 0.2508518578614311, "step": 773, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248294317221735e-06, "weight_min": 0.0, "weight_std": 1.2013376363029238e-05 }, { "epoch": 0.2508518578614311, "step": 773, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235827302385587e-06, "weight_min": 0.0, "weight_std": 1.480623905081302e-05 }, { "epoch": 0.2511763751419763, "step": 774, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3580714039562736e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9048075046157464e-05 }, { "epoch": 0.2511763751419763, "step": 774, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2204285667394288e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.764028456818778e-05 }, { "epoch": 0.2511763751419763, "step": 774, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6313206288032234e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7905573258758523e-05 }, { "epoch": 0.2511763751419763, "step": 774, "weight_max": 2.690521432668902e-05, "weight_mean": 3.364609938216745e-06, "weight_min": 0.0, "weight_std": 9.511840289633255e-06 }, { "epoch": 0.2515008924225215, "step": 775, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5440160495927557e-05, "weight_min": 7.298507354036587e-10, "weight_std": 2.136479815817438e-05 }, { "epoch": 0.2515008924225215, "step": 775, "weight_max": 4.187939703115262e-05, "weight_mean": 1.222422179125715e-05, "weight_min": 0.0, "weight_std": 1.7422584278392605e-05 }, { "epoch": 0.2515008924225215, "step": 775, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0388208465883508e-05, "weight_min": 0.0, "weight_std": 1.9062837964156643e-05 }, { "epoch": 0.2515008924225215, "step": 775, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0159257726627402e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8835487935575657e-05 }, { "epoch": 0.2518254097030667, "step": 776, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0244099030387588e-05, "weight_min": 0.0, "weight_std": 1.902850272017531e-05 }, { "epoch": 0.2518254097030667, "step": 776, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0551751984166913e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9543702364899218e-05 }, { "epoch": 0.2518254097030667, "step": 776, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2046599294990301e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.710425567580387e-05 }, { "epoch": 0.2518254097030667, "step": 776, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.802111890166998e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.814882671169471e-05 }, { "epoch": 0.2521499269836119, "step": 777, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076951765659032e-06, "weight_min": 0.0, "weight_std": 1.4357683539856225e-05 }, { "epoch": 0.2521499269836119, "step": 777, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3681185919267591e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8931088561657816e-05 }, { "epoch": 0.2521499269836119, "step": 777, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076890829514014e-06, "weight_min": 0.0, "weight_std": 1.4357708096213173e-05 }, { "epoch": 0.2521499269836119, "step": 777, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550649676588364e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9535344108589925e-05 }, { "epoch": 0.25247444426415705, "step": 778, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311241567251272e-05, "weight_min": 0.0, "weight_std": 1.9091161448159255e-05 }, { "epoch": 0.25247444426415705, "step": 778, "weight_max": 3.998322426923551e-05, "weight_mean": 2.4293593014590442e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0123414287809283e-05 }, { "epoch": 0.25247444426415705, "step": 778, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311765436199494e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9095357856713235e-05 }, { "epoch": 0.25247444426415705, "step": 778, "weight_max": 3.998322426923551e-05, "weight_mean": 9.210355528921355e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7134756490122527e-05 }, { "epoch": 0.25279896154470227, "step": 779, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.828371750889346e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9740095012821257e-05 }, { "epoch": 0.25279896154470227, "step": 779, "weight_max": 2.5890016331686638e-05, "weight_mean": 4.796301254827995e-06, "weight_min": 5.771870759119224e-10, "weight_std": 9.576038792147301e-06 }, { "epoch": 0.25279896154470227, "step": 779, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5321533030364662e-05, "weight_min": 7.091466858177853e-10, "weight_std": 2.1202706193435006e-05 }, { "epoch": 0.25279896154470227, "step": 779, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9069968402618542e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.070246046059765e-05 }, { "epoch": 0.25312347882524744, "grad_norm": 11.973822593688965, "learning_rate": 8.618367346938776e-06, "loss": 0.1866, "step": 780 }, { "epoch": 0.25312347882524744, "step": 780, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.3815631720935926e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6338013665517792e-05 }, { "epoch": 0.25312347882524744, "step": 780, "weight_max": 3.662098970380612e-05, "weight_mean": 7.302563972189091e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.4088606803852599e-05 }, { "epoch": 0.25312347882524744, "step": 780, "weight_max": 3.998322426923551e-05, "weight_mean": 1.468494247092167e-05, "weight_min": 0.0, "weight_std": 2.0272189431125298e-05 }, { "epoch": 0.25312347882524744, "step": 780, "weight_max": 7.4070909406032115e-09, "weight_mean": 1.7076376979119345e-09, "weight_min": 5.864964625068581e-10, "weight_std": 2.328072845614315e-09 }, { "epoch": 0.25344799610579266, "step": 781, "weight_max": 3.998322426923551e-05, "weight_mean": 4.999485554435523e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4135564924799837e-05 }, { "epoch": 0.25344799610579266, "step": 781, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3177355867810547e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8631764760357328e-05 }, { "epoch": 0.25344799610579266, "step": 781, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3282590771268588e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.902732765302062e-05 }, { "epoch": 0.25344799610579266, "step": 781, "weight_max": 3.752332122530788e-05, "weight_mean": 9.306240826845169e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7230693629244342e-05 }, { "epoch": 0.25377251338633783, "step": 782, "weight_max": 3.8743506593164057e-05, "weight_mean": 5.38233871338889e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.356378743366804e-05 }, { "epoch": 0.25377251338633783, "step": 782, "weight_max": 3.998322426923551e-05, "weight_mean": 9.464905815548263e-06, "weight_min": 0.0, "weight_std": 1.7560465494170785e-05 }, { "epoch": 0.25377251338633783, "step": 782, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0043952897831332e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.863887155195698e-05 }, { "epoch": 0.25377251338633783, "step": 782, "weight_max": 1.4281727999332361e-05, "weight_mean": 1.7897599491334404e-06, "weight_min": 5.958802340444436e-10, "weight_std": 5.047529157309327e-06 }, { "epoch": 0.254097030666883, "step": 783, "weight_max": 2.7549629066925263e-07, "weight_mean": 3.517954283438485e-08, "weight_min": 6.053385570581327e-10, "weight_std": 9.710311132948846e-08 }, { "epoch": 0.254097030666883, "step": 783, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.43548196926713e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.252568563359091e-05 }, { "epoch": 0.254097030666883, "step": 783, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998627900931751e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4135911442281213e-05 }, { "epoch": 0.254097030666883, "step": 783, "weight_max": 2.133443376806099e-05, "weight_mean": 2.667370154085802e-06, "weight_min": 5.771870759119224e-10, "weight_std": 7.5426332841743715e-06 }, { "epoch": 0.2544215479474282, "step": 784, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2330150639172643e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7894812117447145e-05 }, { "epoch": 0.2544215479474282, "step": 784, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.846475222730078e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.825734943849966e-05 }, { "epoch": 0.2544215479474282, "step": 784, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.46176987781655e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.752518437569961e-05 }, { "epoch": 0.2544215479474282, "step": 784, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6164882254088297e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9205799617338926e-05 }, { "epoch": 0.2547460652279734, "step": 785, "weight_max": 3.813097282545641e-05, "weight_mean": 9.382923963130452e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.7373145965393633e-05 }, { "epoch": 0.2547460652279734, "step": 785, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.73420355876442e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8063379684463143e-05 }, { "epoch": 0.2547460652279734, "step": 785, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2564125427161343e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8194081349065527e-05 }, { "epoch": 0.2547460652279734, "step": 785, "weight_max": 4.381950566312298e-05, "weight_mean": 1.8149254174204543e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0759052858920768e-05 }, { "epoch": 0.2550705825085186, "step": 786, "weight_max": 3.116124207735993e-05, "weight_mean": 3.896232556144241e-06, "weight_min": 0.0, "weight_std": 1.1016727512469515e-05 }, { "epoch": 0.2550705825085186, "step": 786, "weight_max": 3.632264997577295e-05, "weight_mean": 1.6955040337052196e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.827177038649097e-05 }, { "epoch": 0.2550705825085186, "step": 786, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467638973437715e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2631752724701073e-05 }, { "epoch": 0.2550705825085186, "step": 786, "weight_max": 4.252122016623616e-05, "weight_mean": 5.319608135323506e-06, "weight_min": 0.0, "weight_std": 1.503172461525537e-05 }, { "epoch": 0.2553950997890638, "step": 787, "weight_max": 4.187939703115262e-05, "weight_mean": 5.64345600650995e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4685588212159928e-05 }, { "epoch": 0.2553950997890638, "step": 787, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4971898053772748e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0736311853397638e-05 }, { "epoch": 0.2553950997890638, "step": 787, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6020643670344725e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9915709344786592e-05 }, { "epoch": 0.2553950997890638, "step": 787, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5003964108473156e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0742552806041203e-05 }, { "epoch": 0.25571961706960894, "step": 788, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.93776086741127e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.675063685979694e-05 }, { "epoch": 0.25571961706960894, "step": 788, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.916822480387054e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.660707675910089e-05 }, { "epoch": 0.25571961706960894, "step": 788, "weight_max": 3.1316551574178675e-09, "weight_mean": 1.5120629193177137e-09, "weight_min": 6.053385570581327e-10, "weight_std": 1.0404014316733878e-09 }, { "epoch": 0.25571961706960894, "step": 788, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9156068447045982e-05, "weight_min": 0.0, "weight_std": 2.0690829842351377e-05 }, { "epoch": 0.25604413435015416, "step": 789, "weight_max": 4.187939703115262e-05, "weight_mean": 6.524340733449208e-06, "weight_min": 0.0, "weight_std": 1.47341024785419e-05 }, { "epoch": 0.25604413435015416, "step": 789, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.483275915845297e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2274826076463796e-05 }, { "epoch": 0.25604413435015416, "step": 789, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.789809175534174e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9275596059742384e-05 }, { "epoch": 0.25604413435015416, "step": 789, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965211817558156e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.121269178838702e-05 }, { "epoch": 0.25636865163069933, "grad_norm": 0.2943289279937744, "learning_rate": 8.59795918367347e-06, "loss": 0.128, "step": 790 }, { "epoch": 0.25636865163069933, "step": 790, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.079004949948285e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.435685499018291e-05 }, { "epoch": 0.25636865163069933, "step": 790, "weight_max": 4.252122016623616e-05, "weight_mean": 1.885557867353782e-05, "weight_min": 7.829137893544669e-10, "weight_std": 2.0389939891174436e-05 }, { "epoch": 0.25636865163069933, "step": 790, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.316012094321195e-05, "weight_min": 0.0, "weight_std": 1.840457298385445e-05 }, { "epoch": 0.25636865163069933, "step": 790, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4492343325400725e-05, "weight_min": 0.0, "weight_std": 2.0007448256365024e-05 }, { "epoch": 0.25669316891124455, "step": 791, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930603270418942e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1114511835330632e-05 }, { "epoch": 0.25669316891124455, "step": 791, "weight_max": 4.187939703115262e-05, "weight_mean": 2.740315358096268e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.74990163941402e-05 }, { "epoch": 0.25669316891124455, "step": 791, "weight_max": 3.572963396436535e-05, "weight_mean": 8.362429980479646e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.552963840367738e-05 }, { "epoch": 0.25669316891124455, "step": 791, "weight_max": 4.252122016623616e-05, "weight_mean": 9.968404810933862e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8510721929487772e-05 }, { "epoch": 0.2570176861917897, "step": 792, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4463757906924002e-05, "weight_min": 0.0, "weight_std": 2.0009110812679864e-05 }, { "epoch": 0.2570176861917897, "step": 792, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2187709216959774e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8160506442654878e-05 }, { "epoch": 0.2570176861917897, "step": 792, "weight_max": 3.782653220696375e-05, "weight_mean": 1.5014233213150874e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.7600863429834135e-05 }, { "epoch": 0.2570176861917897, "step": 792, "weight_max": 3.255531191825867e-05, "weight_mean": 8.139842975651845e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5069554137880914e-05 }, { "epoch": 0.2573422034723349, "step": 793, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3541971384256613e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8824361177394167e-05 }, { "epoch": 0.2573422034723349, "step": 793, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4479999663308263e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.0058882000739686e-05 }, { "epoch": 0.2573422034723349, "step": 793, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.023211189021822e-05, "weight_min": 0.0, "weight_std": 1.89457005035365e-05 }, { "epoch": 0.2573422034723349, "step": 793, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631796612869948e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9687096937559545e-05 }, { "epoch": 0.2576667207528801, "step": 794, "weight_max": 3.36925259034615e-05, "weight_mean": 7.936912879813462e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4731556802871637e-05 }, { "epoch": 0.2576667207528801, "step": 794, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.84327016340103e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8236762116430327e-05 }, { "epoch": 0.2576667207528801, "step": 794, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0138142897631042e-05, "weight_min": 0.0, "weight_std": 1.8615917724673636e-05 }, { "epoch": 0.2576667207528801, "step": 794, "weight_max": 3.998322426923551e-05, "weight_mean": 8.893634912965354e-06, "weight_min": 0.0, "weight_std": 1.6634334315313026e-05 }, { "epoch": 0.2579912380334253, "step": 795, "weight_max": 4.187939703115262e-05, "weight_mean": 2.1403342543635517e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.885547135316301e-05 }, { "epoch": 0.2579912380334253, "step": 795, "weight_max": 3.116124207735993e-05, "weight_mean": 3.896223006449873e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1016731150448322e-05 }, { "epoch": 0.2579912380334253, "step": 795, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.181748282571789e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7125958038377576e-05 }, { "epoch": 0.2579912380334253, "step": 795, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430260105436901e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2527792023320217e-05 }, { "epoch": 0.2583157553139705, "step": 796, "weight_max": 3.998322426923551e-05, "weight_mean": 9.95725167740602e-06, "weight_min": 0.0, "weight_std": 1.8436410755384713e-05 }, { "epoch": 0.2583157553139705, "step": 796, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.59437022882048e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.620834336790722e-05 }, { "epoch": 0.2583157553139705, "step": 796, "weight_max": 3.813097282545641e-05, "weight_mean": 8.45801241666777e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5827676179469563e-05 }, { "epoch": 0.2583157553139705, "step": 796, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4711465155414771e-05, "weight_min": 0.0, "weight_std": 2.0424475223990157e-05 }, { "epoch": 0.25864027259451566, "step": 797, "weight_max": 3.662098970380612e-05, "weight_mean": 1.176174555439502e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6418489394709468e-05 }, { "epoch": 0.25864027259451566, "step": 797, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.477378620824311e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0437189959920943e-05 }, { "epoch": 0.25864027259451566, "step": 797, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9891209376510233e-05, "weight_min": 0.0, "weight_std": 2.148531530110631e-05 }, { "epoch": 0.25864027259451566, "step": 797, "weight_max": 2.6141986381844617e-05, "weight_mean": 3.268394038968836e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.242326996172778e-06 }, { "epoch": 0.25896478987506083, "step": 798, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0117075362359174e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.874669578683097e-05 }, { "epoch": 0.25896478987506083, "step": 798, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.554911068524234e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1473173546837643e-05 }, { "epoch": 0.25896478987506083, "step": 798, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3679134756093845e-05, "weight_min": 2.517239972021912e-09, "weight_std": 1.9350498405401595e-05 }, { "epoch": 0.25896478987506083, "step": 798, "weight_max": 3.1316551574178675e-09, "weight_mean": 9.634156805660155e-10, "weight_min": 5.864964625068581e-10, "weight_std": 8.805956519530866e-10 }, { "epoch": 0.25928930715560605, "step": 799, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844284376304131e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3697354006581008e-05 }, { "epoch": 0.25928930715560605, "step": 799, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470355846337043e-05, "weight_min": 0.0, "weight_std": 1.938608147611376e-05 }, { "epoch": 0.25928930715560605, "step": 799, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4699000530526973e-05, "weight_min": 0.0, "weight_std": 2.034792851191014e-05 }, { "epoch": 0.25928930715560605, "step": 799, "weight_max": 4.381950566312298e-05, "weight_mean": 2.8647033104789443e-05, "weight_min": 0.0, "weight_std": 1.854942638601642e-05 }, { "epoch": 0.2596138244361512, "grad_norm": 0.36117103695869446, "learning_rate": 8.577551020408163e-06, "loss": 0.2084, "step": 800 }, { "epoch": 0.2596138244361512, "eval_accuracy": 0.494, "eval_auc": 0.9732806524139476, "eval_loss": 1.491502285003662, "eval_runtime": 155.3763, "eval_samples_per_second": 12.872, "eval_steps_per_second": 1.609, "step": 800 }, { "epoch": 0.2596138244361512, "step": 800, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0120365004695486e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.876948408607859e-05 }, { "epoch": 0.2596138244361512, "step": 800, "weight_max": 4.187939703115262e-05, "weight_mean": 9.206544746120926e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7244699847651646e-05 }, { "epoch": 0.2596138244361512, "step": 800, "weight_max": 4.252122016623616e-05, "weight_mean": 5.975261956336908e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.488176167185884e-05 }, { "epoch": 0.2596138244361512, "step": 800, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315763701219112e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5033274394227192e-05 }, { "epoch": 0.25993834171669644, "step": 801, "weight_max": 4.187939703115262e-05, "weight_mean": 9.888098247756716e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8350223399465904e-05 }, { "epoch": 0.25993834171669644, "step": 801, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.390672150591854e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7384811144438572e-05 }, { "epoch": 0.25993834171669644, "step": 801, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882095254288288e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3806567949359305e-05 }, { "epoch": 0.25993834171669644, "step": 801, "weight_max": 3.0339444492710754e-05, "weight_mean": 7.22069034964079e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3391034372034483e-05 }, { "epoch": 0.2602628589972416, "step": 802, "weight_max": 8.209445695683826e-06, "weight_mean": 1.0271897963320953e-06, "weight_min": 6.196655966128617e-10, "weight_std": 2.9020695819781395e-06 }, { "epoch": 0.2602628589972416, "step": 802, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3893010873289313e-05, "weight_min": 0.0, "weight_std": 1.937834531418048e-05 }, { "epoch": 0.2602628589972416, "step": 802, "weight_max": 2.6649584469851106e-05, "weight_mean": 8.52132507134229e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2000981769233476e-05 }, { "epoch": 0.2602628589972416, "step": 802, "weight_max": 2.7939937353949063e-05, "weight_mean": 3.4932677408505697e-06, "weight_min": 6.100955851628953e-10, "weight_std": 9.877946467895526e-06 }, { "epoch": 0.2605873762777868, "step": 803, "weight_max": 3.998322426923551e-05, "weight_mean": 1.9113836970063858e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.047430280072149e-05 }, { "epoch": 0.2605873762777868, "step": 803, "weight_max": 3.602553260861896e-05, "weight_mean": 6.8935555646021385e-06, "weight_min": 0.0, "weight_std": 1.35383224915131e-05 }, { "epoch": 0.2605873762777868, "step": 803, "weight_max": 4.187939703115262e-05, "weight_mean": 8.440400051767938e-06, "weight_min": 0.0, "weight_std": 1.621874253032729e-05 }, { "epoch": 0.2605873762777868, "step": 803, "weight_max": 4.187939703115262e-05, "weight_mean": 9.270406735595316e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7353366274619475e-05 }, { "epoch": 0.260911893558332, "step": 804, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.155883162748069e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.695295941317454e-05 }, { "epoch": 0.260911893558332, "step": 804, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.862831575446762e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.836873525462579e-05 }, { "epoch": 0.260911893558332, "step": 804, "weight_max": 4.3167921830900013e-05, "weight_mean": 6.766782007616712e-06, "weight_min": 0.0, "weight_std": 1.5199455447145738e-05 }, { "epoch": 0.260911893558332, "step": 804, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5163646569126286e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0913097614538856e-05 }, { "epoch": 0.26123641083887716, "step": 805, "weight_max": 1.076157385426768e-09, "weight_mean": 6.987400658076126e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.0806938116901819e-10 }, { "epoch": 0.26123641083887716, "step": 805, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0614829079713672e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.868095569079742e-05 }, { "epoch": 0.26123641083887716, "step": 805, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.96499808630324e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.1212778190383688e-05 }, { "epoch": 0.26123641083887716, "step": 805, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3511580366175622e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9880770196323283e-05 }, { "epoch": 0.2615609281194224, "step": 806, "weight_max": 7.829137893544669e-10, "weight_mean": 4.855345570042857e-10, "weight_min": 0.0, "weight_std": 3.0664576433636626e-10 }, { "epoch": 0.2615609281194224, "step": 806, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.88144165603444e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.830470864661038e-05 }, { "epoch": 0.2615609281194224, "step": 806, "weight_max": 3.967146767536178e-05, "weight_mean": 9.537224286759738e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7676671632216312e-05 }, { "epoch": 0.2615609281194224, "step": 806, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4128294424153864e-05, "weight_min": 0.0, "weight_std": 1.9736460671992972e-05 }, { "epoch": 0.26188544539996755, "step": 807, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4934546925360337e-05, "weight_min": 0.0, "weight_std": 2.0692086764029227e-05 }, { "epoch": 0.26188544539996755, "step": 807, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.157764007890364e-06, "weight_min": 0.0, "weight_std": 1.4580419701815117e-05 }, { "epoch": 0.26188544539996755, "step": 807, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236201104708016e-06, "weight_min": 0.0, "weight_std": 1.4806087165197823e-05 }, { "epoch": 0.26188544539996755, "step": 807, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2357540880620945e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806268154643476e-05 }, { "epoch": 0.2622099626805127, "step": 808, "weight_max": 1.2475846233428456e-05, "weight_mean": 1.5607315617671702e-06, "weight_min": 0.0, "weight_std": 4.410372639540583e-06 }, { "epoch": 0.2622099626805127, "step": 808, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.003590296022594e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8582923075882718e-05 }, { "epoch": 0.2622099626805127, "step": 808, "weight_max": 3.22740561387036e-05, "weight_mean": 8.027407602639869e-06, "weight_min": 0.0, "weight_std": 1.4709949937241618e-05 }, { "epoch": 0.2622099626805127, "step": 808, "weight_max": 3.967146767536178e-05, "weight_mean": 9.6875619419734e-06, "weight_min": 0.0, "weight_std": 1.7943884813576005e-05 }, { "epoch": 0.26253447996105794, "step": 809, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235953722149134e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806187209615018e-05 }, { "epoch": 0.26253447996105794, "step": 809, "weight_max": 4.252122016623616e-05, "weight_mean": 9.856050382950343e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8323502445127815e-05 }, { "epoch": 0.26253447996105794, "step": 809, "weight_max": 1.1931196031156333e-09, "weight_mean": 6.485069148354228e-10, "weight_min": 0.0, "weight_std": 3.253282920834266e-10 }, { "epoch": 0.26253447996105794, "step": 809, "weight_max": 4.252122016623616e-05, "weight_mean": 9.455894542043097e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7686859791865572e-05 }, { "epoch": 0.2628589972416031, "grad_norm": 24.898073196411133, "learning_rate": 8.557142857142858e-06, "loss": 0.3277, "step": 810 }, { "epoch": 0.2628589972416031, "step": 810, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.603573764441535e-05, "weight_min": 0.0, "weight_std": 2.2115504179964773e-05 }, { "epoch": 0.2628589972416031, "step": 810, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.401652288128389e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.525988227513153e-05 }, { "epoch": 0.2628589972416031, "step": 810, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653354608308291e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3159445188648533e-05 }, { "epoch": 0.2628589972416031, "step": 810, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578472271532519e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2947131835971959e-05 }, { "epoch": 0.26318351452214833, "step": 811, "weight_max": 3.752332122530788e-05, "weight_mean": 4.69108454126399e-06, "weight_min": 0.0, "weight_std": 1.3266226233099587e-05 }, { "epoch": 0.26318351452214833, "step": 811, "weight_max": 3.998322426923551e-05, "weight_mean": 8.869862540450413e-06, "weight_min": 0.0, "weight_std": 1.6575457266299054e-05 }, { "epoch": 0.26318351452214833, "step": 811, "weight_max": 4.252122016623616e-05, "weight_mean": 2.330696224817075e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9802966562565416e-05 }, { "epoch": 0.26318351452214833, "step": 811, "weight_max": 3.843663216684945e-05, "weight_mean": 9.272566785512026e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7180294889840297e-05 }, { "epoch": 0.2635080318026935, "step": 812, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.904723472369369e-06, "weight_min": 0.0, "weight_std": 1.6698828403605148e-05 }, { "epoch": 0.2635080318026935, "step": 812, "weight_max": 1.2334475663067224e-09, "weight_mean": 7.951194702648934e-10, "weight_min": 5.864964625068581e-10, "weight_std": 2.0312176940429083e-10 }, { "epoch": 0.2635080318026935, "step": 812, "weight_max": 3.1316551574178675e-09, "weight_mean": 1.0306662190373572e-09, "weight_min": 5.679522963042416e-10, "weight_std": 8.679079122053679e-10 }, { "epoch": 0.2635080318026935, "step": 812, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.881558071356267e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.830463770602364e-05 }, { "epoch": 0.26383254908323867, "step": 813, "weight_max": 1.753466261078529e-09, "weight_mean": 6.989239187404905e-10, "weight_min": 0.0, "weight_std": 4.847164891685907e-10 }, { "epoch": 0.26383254908323867, "step": 813, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4930486941011623e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.034640419878997e-05 }, { "epoch": 0.26383254908323867, "step": 813, "weight_max": 3.998322426923551e-05, "weight_mean": 9.974086424335837e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8426078895572573e-05 }, { "epoch": 0.26383254908323867, "step": 813, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0986038432747591e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7017271602526307e-05 }, { "epoch": 0.2641570663637839, "step": 814, "weight_max": 4.252122016623616e-05, "weight_mean": 1.004396199277835e-05, "weight_min": 0.0, "weight_std": 1.863886609498877e-05 }, { "epoch": 0.2641570663637839, "step": 814, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5354455172200687e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1218686015345156e-05 }, { "epoch": 0.2641570663637839, "step": 814, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.6171288532787e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7843072782852687e-05 }, { "epoch": 0.2641570663637839, "step": 814, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.464136044494808e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5694031390012242e-05 }, { "epoch": 0.26448158364432905, "step": 815, "weight_max": 3.36925259034615e-05, "weight_mean": 7.936796464491636e-06, "weight_min": 0.0, "weight_std": 1.4731628652953077e-05 }, { "epoch": 0.26448158364432905, "step": 815, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2360865083755925e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4806133549427614e-05 }, { "epoch": 0.26448158364432905, "step": 815, "weight_max": 4.252122016623616e-05, "weight_mean": 1.000808879325632e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.857324969023466e-05 }, { "epoch": 0.26448158364432905, "step": 815, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.654601853981148e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.79069465957582e-05 }, { "epoch": 0.2648061009248743, "step": 816, "weight_max": 2.4894337911973707e-05, "weight_mean": 7.434696271957364e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0674978511815425e-05 }, { "epoch": 0.2648061009248743, "step": 816, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0473015208845027e-05, "weight_min": 0.0, "weight_std": 1.9402510588406585e-05 }, { "epoch": 0.2648061009248743, "step": 816, "weight_max": 4.252122016623616e-05, "weight_mean": 2.55460363405291e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.1182646378292702e-05 }, { "epoch": 0.2648061009248743, "step": 816, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3168896556599066e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5032819646876305e-05 }, { "epoch": 0.26513061820541944, "step": 817, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.8062255549011752e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.935818545462098e-05 }, { "epoch": 0.26513061820541944, "step": 817, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4028217265149578e-05, "weight_min": 0.0, "weight_std": 1.9492688807076775e-05 }, { "epoch": 0.26513061820541944, "step": 817, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3333597053133417e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.915410757646896e-05 }, { "epoch": 0.26513061820541944, "step": 817, "weight_max": 4.187939703115262e-05, "weight_mean": 9.375578883918934e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7516133084427565e-05 }, { "epoch": 0.2654551354859646, "step": 818, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235875960352132e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.480621904192958e-05 }, { "epoch": 0.2654551354859646, "step": 818, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4656531675427686e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0438823412405327e-05 }, { "epoch": 0.2654551354859646, "step": 818, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920813353237463e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3915906492911745e-05 }, { "epoch": 0.2654551354859646, "step": 818, "weight_max": 3.9360926166409627e-05, "weight_mean": 7.702661605435424e-06, "weight_min": 0.0, "weight_std": 1.4976190868765116e-05 }, { "epoch": 0.26577965276650983, "step": 819, "weight_max": 8.94625595826426e-10, "weight_mean": 6.246053674274776e-10, "weight_min": 0.0, "weight_std": 2.745995375530441e-10 }, { "epoch": 0.26577965276650983, "step": 819, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9014172721654177e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.05873711820459e-05 }, { "epoch": 0.26577965276650983, "step": 819, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8172833733842708e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.976216481125448e-05 }, { "epoch": 0.26577965276650983, "step": 819, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2991657058591954e-05, "weight_min": 0.0, "weight_std": 1.812843584048096e-05 }, { "epoch": 0.266104170047055, "grad_norm": 9.991578102111816, "learning_rate": 8.536734693877553e-06, "loss": 0.2373, "step": 820 }, { "epoch": 0.266104170047055, "step": 820, "weight_max": 3.998322426923551e-05, "weight_mean": 9.50181856751442e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7623931853449903e-05 }, { "epoch": 0.266104170047055, "step": 820, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5549103409284726e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.146891347365454e-05 }, { "epoch": 0.266104170047055, "step": 820, "weight_max": 3.602553260861896e-05, "weight_mean": 8.96977144293487e-06, "weight_min": 0.0, "weight_std": 1.660804991843179e-05 }, { "epoch": 0.266104170047055, "step": 820, "weight_max": 3.967146767536178e-05, "weight_mean": 9.006344043882564e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6684782167430967e-05 }, { "epoch": 0.2664286873276002, "step": 821, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.078733011032455e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4356964129547123e-05 }, { "epoch": 0.2664286873276002, "step": 821, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8467533664079383e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9926068489439785e-05 }, { "epoch": 0.2664286873276002, "step": 821, "weight_max": 3.006794759130571e-05, "weight_mean": 5.810102720715804e-06, "weight_min": 0.0, "weight_std": 1.1359594282112084e-05 }, { "epoch": 0.2664286873276002, "step": 821, "weight_max": 3.671447146302853e-09, "weight_mean": 1.0636282965492683e-09, "weight_min": 0.0, "weight_std": 1.1058927107399086e-09 }, { "epoch": 0.2667532046081454, "step": 822, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2266827070561703e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8191360140917823e-05 }, { "epoch": 0.2667532046081454, "step": 822, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.120304639509413e-06, "weight_min": 0.0, "weight_std": 1.7076950825867243e-05 }, { "epoch": 0.2667532046081454, "step": 822, "weight_max": 1.504423474685268e-09, "weight_mean": 7.628114806479402e-10, "weight_min": 5.679522963042416e-10, "weight_std": 3.143069693400946e-10 }, { "epoch": 0.2667532046081454, "step": 822, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.3569552872213535e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2320754649408627e-05 }, { "epoch": 0.26707772188869056, "step": 823, "weight_max": 3.662098970380612e-05, "weight_mean": 9.11868210096145e-06, "weight_min": 7.091466858177853e-10, "weight_std": 1.6882911950233392e-05 }, { "epoch": 0.26707772188869056, "step": 823, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0556606614263728e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.203560507041402e-05 }, { "epoch": 0.26707772188869056, "step": 823, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.3669064830755815e-05, "weight_min": 8.156453290553145e-10, "weight_std": 1.7546570234117098e-05 }, { "epoch": 0.26707772188869056, "step": 823, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.2022545888612512e-05, "weight_min": 0.0, "weight_std": 1.669539233262185e-05 }, { "epoch": 0.2674022391692358, "step": 824, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.434899437910644e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2525922102213372e-05 }, { "epoch": 0.2674022391692358, "step": 824, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.217218575940933e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7182135707116686e-05 }, { "epoch": 0.2674022391692358, "step": 824, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.3933678171015345e-06, "weight_min": 0.0, "weight_std": 1.242412326973863e-05 }, { "epoch": 0.2674022391692358, "step": 824, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2353097999002784e-06, "weight_min": 0.0, "weight_std": 1.4806447325099725e-05 }, { "epoch": 0.26772675644978094, "step": 825, "weight_max": 4.252122016623616e-05, "weight_mean": 1.789672751328908e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9942304788855836e-05 }, { "epoch": 0.26772675644978094, "step": 825, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3976066838949919e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.951582817127928e-05 }, { "epoch": 0.26772675644978094, "step": 825, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156100087333471e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4581090908905026e-05 }, { "epoch": 0.26772675644978094, "step": 825, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3674884030479006e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.894311571959406e-05 }, { "epoch": 0.26805127373032617, "step": 826, "weight_max": 4.187939703115262e-05, "weight_mean": 1.384578354191035e-05, "weight_min": 0.0, "weight_std": 1.9270337361376733e-05 }, { "epoch": 0.26805127373032617, "step": 826, "weight_max": 3.602553260861896e-05, "weight_mean": 1.0802126780617982e-05, "weight_min": 0.0, "weight_std": 1.628233803785406e-05 }, { "epoch": 0.26805127373032617, "step": 826, "weight_max": 4.252122016623616e-05, "weight_mean": 2.051405499514658e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.197419416916091e-05 }, { "epoch": 0.26805127373032617, "step": 826, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1319412806187756e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.641293965803925e-05 }, { "epoch": 0.26837579101087133, "step": 827, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471143468748778e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9390115994610824e-05 }, { "epoch": 0.26837579101087133, "step": 827, "weight_max": 3.602553260861896e-05, "weight_mean": 8.933136086852755e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.65402889251709e-05 }, { "epoch": 0.26837579101087133, "step": 827, "weight_max": 2.877845961535286e-09, "weight_mean": 9.425511482419324e-10, "weight_min": 0.0, "weight_std": 8.504211224114044e-10 }, { "epoch": 0.26837579101087133, "step": 827, "weight_max": 1.504423474685268e-09, "weight_mean": 7.892744235959981e-10, "weight_min": 6.053385570581327e-10, "weight_std": 3.097956335906815e-10 }, { "epoch": 0.2687003082914165, "step": 828, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0623683920130134e-05, "weight_min": 0.0, "weight_std": 1.620707735128235e-05 }, { "epoch": 0.2687003082914165, "step": 828, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.80526237981394e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8169463146477938e-05 }, { "epoch": 0.2687003082914165, "step": 828, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23604921909282e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806148101342842e-05 }, { "epoch": 0.2687003082914165, "step": 828, "weight_max": 1.7955494513444137e-06, "weight_mean": 2.2521284392951202e-07, "weight_min": 0.0, "weight_std": 6.345123324535962e-07 }, { "epoch": 0.2690248255719617, "step": 829, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8702976376516744e-05, "weight_min": 0.0, "weight_std": 2.022276930802036e-05 }, { "epoch": 0.2690248255719617, "step": 829, "weight_max": 3.340639523230493e-05, "weight_mean": 4.176580659986939e-06, "weight_min": 0.0, "weight_std": 1.1810629075625911e-05 }, { "epoch": 0.2690248255719617, "step": 829, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316308033798123e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.5033055206004065e-05 }, { "epoch": 0.2690248255719617, "step": 829, "weight_max": 3.843663216684945e-05, "weight_mean": 9.08781021280447e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.593518209119793e-05 }, { "epoch": 0.2693493428525069, "grad_norm": 3.1093804836273193, "learning_rate": 8.516326530612246e-06, "loss": 0.246, "step": 830 }, { "epoch": 0.2693493428525069, "step": 830, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.0108985406986903e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.649415207793936e-05 }, { "epoch": 0.2693493428525069, "step": 830, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.008464076032396e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6685462469467893e-05 }, { "epoch": 0.2693493428525069, "step": 830, "weight_max": 2.952862632810138e-05, "weight_mean": 3.7202714793238556e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0428456334921066e-05 }, { "epoch": 0.2693493428525069, "step": 830, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920780611428199e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3915920135332271e-05 }, { "epoch": 0.2696738601330521, "step": 831, "weight_max": 1.2475846233428456e-05, "weight_mean": 1.5613146615578444e-06, "weight_min": 0.0, "weight_std": 4.4101379899075255e-06 }, { "epoch": 0.2696738601330521, "step": 831, "weight_max": 3.36925259034615e-05, "weight_mean": 8.388298738282174e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5529994925600477e-05 }, { "epoch": 0.2696738601330521, "step": 831, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4242824363464024e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0048104488523677e-05 }, { "epoch": 0.2696738601330521, "step": 831, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4575035493180621e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0445122572709806e-05 }, { "epoch": 0.2699983774135973, "step": 832, "weight_max": 8.267048712262692e-10, "weight_mean": 4.871776315695797e-10, "weight_min": 0.0, "weight_std": 3.1016661461436e-10 }, { "epoch": 0.2699983774135973, "step": 832, "weight_max": 5.663789437448941e-09, "weight_mean": 1.2923395686925687e-09, "weight_min": 0.0, "weight_std": 1.7987832334753762e-09 }, { "epoch": 0.2699983774135973, "step": 832, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6847101505845785e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9165760022588074e-05 }, { "epoch": 0.2699983774135973, "step": 832, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.341374607174657e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7150205167126842e-05 }, { "epoch": 0.27032289469414245, "step": 833, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9064702428295277e-05, "weight_min": 0.0, "weight_std": 2.053614480246324e-05 }, { "epoch": 0.27032289469414245, "step": 833, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.410275207395898e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.525643983768532e-05 }, { "epoch": 0.27032289469414245, "step": 833, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5161298506427556e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0969619072275236e-05 }, { "epoch": 0.27032289469414245, "step": 833, "weight_max": 3.4849264920921996e-05, "weight_mean": 7.001235644565895e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2793022506230045e-05 }, { "epoch": 0.27064741197468767, "step": 834, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9192575564375147e-05, "weight_min": 1.2334475663067224e-09, "weight_std": 2.0714684069389477e-05 }, { "epoch": 0.27064741197468767, "step": 834, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1153820196341258e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9075401723966934e-05 }, { "epoch": 0.27064741197468767, "step": 834, "weight_max": 3.572963396436535e-05, "weight_mean": 8.191713277483359e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.524877006886527e-05 }, { "epoch": 0.27064741197468767, "step": 834, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.793796764104627e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9318769773235545e-05 }, { "epoch": 0.27097192925523284, "step": 835, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998701570002595e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4135882338450756e-05 }, { "epoch": 0.27097192925523284, "step": 835, "weight_max": 3.632264997577295e-05, "weight_mean": 8.970378075900953e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6609936210443266e-05 }, { "epoch": 0.27097192925523284, "step": 835, "weight_max": 2.952862632810138e-05, "weight_mean": 3.691659685500781e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0439711331855506e-05 }, { "epoch": 0.27097192925523284, "step": 835, "weight_max": 1.2064879095774472e-09, "weight_mean": 5.77589753802954e-10, "weight_min": 0.0, "weight_std": 4.1019418417498343e-10 }, { "epoch": 0.27129644653577806, "step": 836, "weight_max": 3.843663216684945e-05, "weight_mean": 4.806238393939566e-06, "weight_min": 0.0, "weight_std": 1.3588731235358864e-05 }, { "epoch": 0.27129644653577806, "step": 836, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074947567773052e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8654094674275257e-05 }, { "epoch": 0.27129644653577806, "step": 836, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3311575457919389e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.862423960119486e-05 }, { "epoch": 0.27129644653577806, "step": 836, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311639016435947e-05, "weight_min": 0.0, "weight_std": 1.9095436073257588e-05 }, { "epoch": 0.2716209638163232, "step": 837, "weight_max": 3.722132169059478e-05, "weight_mean": 9.269924703403376e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7159343769890256e-05 }, { "epoch": 0.2716209638163232, "step": 837, "weight_max": 2.820166446326766e-05, "weight_mean": 3.526085947669344e-06, "weight_min": 5.771870759119224e-10, "weight_std": 9.970439350581728e-06 }, { "epoch": 0.2716209638163232, "step": 837, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4396374353964347e-05, "weight_min": 0.0, "weight_std": 1.9955970856244676e-05 }, { "epoch": 0.2716209638163232, "step": 837, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.331908586318605e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7403823221684434e-05 }, { "epoch": 0.2719454810968684, "step": 838, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.215472007577773e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.523784885648638e-05 }, { "epoch": 0.2719454810968684, "step": 838, "weight_max": 1.2744457711377777e-09, "weight_mean": 7.95641941220282e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.4902951922811667e-10 }, { "epoch": 0.2719454810968684, "step": 838, "weight_max": 2.4786623864514468e-09, "weight_mean": 1.1590068904610007e-09, "weight_min": 5.958802340444436e-10, "weight_std": 7.863504292160428e-10 }, { "epoch": 0.2719454810968684, "step": 838, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.076783499040175e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.696095750958193e-05 }, { "epoch": 0.2722699983774136, "step": 839, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3235452570370398e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8523380276747048e-05 }, { "epoch": 0.2722699983774136, "step": 839, "weight_max": 3.6920544516760856e-05, "weight_mean": 5.778877039119834e-06, "weight_min": 0.0, "weight_std": 1.2997541489312425e-05 }, { "epoch": 0.2722699983774136, "step": 839, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153796210943256e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8802742488333024e-05 }, { "epoch": 0.2722699983774136, "step": 839, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.4439919798169285e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0379617126309313e-05 }, { "epoch": 0.2725945156579588, "grad_norm": 5.011831760406494, "learning_rate": 8.49591836734694e-06, "loss": 0.1949, "step": 840 }, { "epoch": 0.2725945156579588, "step": 840, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0623576347134076e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.205431337642949e-05 }, { "epoch": 0.2725945156579588, "step": 840, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235490789345931e-06, "weight_min": 0.0, "weight_std": 1.4806374565523583e-05 }, { "epoch": 0.2725945156579588, "step": 840, "weight_max": 2.440382377244532e-05, "weight_mean": 3.0535011319443583e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.626836461189669e-06 }, { "epoch": 0.2725945156579588, "step": 840, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2304357369430363e-05, "weight_min": 0.0, "weight_std": 1.824986065912526e-05 }, { "epoch": 0.272919032938504, "step": 841, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4455601558438502e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.996333958231844e-05 }, { "epoch": 0.272919032938504, "step": 841, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691429239755962e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3266087989904918e-05 }, { "epoch": 0.272919032938504, "step": 841, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8043185264104977e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0634055545087904e-05 }, { "epoch": 0.272919032938504, "step": 841, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4974410078139044e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0751729607582092e-05 }, { "epoch": 0.27324355021904917, "step": 842, "weight_max": 4.252122016623616e-05, "weight_mean": 9.600412340660114e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.790864916983992e-05 }, { "epoch": 0.27324355021904917, "step": 842, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.511967618891504e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.769419395714067e-05 }, { "epoch": 0.27324355021904917, "step": 842, "weight_max": 3.572963396436535e-05, "weight_mean": 8.571859325456899e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.5888506823102944e-05 }, { "epoch": 0.27324355021904917, "step": 842, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3792091522191186e-05, "weight_min": 0.0, "weight_std": 1.9119730495731346e-05 }, { "epoch": 0.27356806749959434, "step": 843, "weight_max": 2.3835209361777743e-09, "weight_mean": 7.061444762257452e-10, "weight_min": 0.0, "weight_std": 8.274863572133029e-10 }, { "epoch": 0.27356806749959434, "step": 843, "weight_max": 3.340639523230493e-05, "weight_mean": 1.1627091225818731e-05, "weight_min": 0.0, "weight_std": 1.6090103599708527e-05 }, { "epoch": 0.27356806749959434, "step": 843, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396488631959073e-06, "weight_min": 0.0, "weight_std": 1.5261963199009188e-05 }, { "epoch": 0.27356806749959434, "step": 843, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691029062087182e-06, "weight_min": 0.0, "weight_std": 1.3266250789456535e-05 }, { "epoch": 0.27389258478013956, "step": 844, "weight_max": 4.252122016623616e-05, "weight_mean": 1.388206692354288e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.975002305698581e-05 }, { "epoch": 0.27389258478013956, "step": 844, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930431830667658e-06, "weight_min": 0.0, "weight_std": 1.1114580956927966e-05 }, { "epoch": 0.27389258478013956, "step": 844, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.7812824808061123e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9274790247436613e-05 }, { "epoch": 0.27389258478013956, "step": 844, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232983186142519e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.894516208267305e-05 }, { "epoch": 0.2742171020606847, "step": 845, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.356961198936915e-06, "weight_min": 0.0, "weight_std": 1.2320751920924522e-05 }, { "epoch": 0.2742171020606847, "step": 845, "weight_max": 3.998322426923551e-05, "weight_mean": 9.21077389648417e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7134500012616627e-05 }, { "epoch": 0.2742171020606847, "step": 845, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392543117632158e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.924733442137949e-05 }, { "epoch": 0.2742171020606847, "step": 845, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8867052858695388e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.057407436950598e-05 }, { "epoch": 0.27454161934122995, "step": 846, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.789510841073934e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.825110120989848e-05 }, { "epoch": 0.27454161934122995, "step": 846, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0856434528250247e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9004282876267098e-05 }, { "epoch": 0.27454161934122995, "step": 846, "weight_max": 3.22740561387036e-05, "weight_mean": 4.037262442579959e-06, "weight_min": 8.716874999592505e-10, "weight_std": 1.140938820753945e-05 }, { "epoch": 0.27454161934122995, "step": 846, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653227733797394e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3159497029846534e-05 }, { "epoch": 0.2748661366217751, "step": 847, "weight_max": 3.855956492770929e-06, "weight_mean": 4.830329203286965e-07, "weight_min": 0.0, "weight_std": 1.3628679198518512e-06 }, { "epoch": 0.2748661366217751, "step": 847, "weight_max": 3.632264997577295e-05, "weight_mean": 9.651479558669962e-06, "weight_min": 0.0, "weight_std": 1.6298321497743018e-05 }, { "epoch": 0.2748661366217751, "step": 847, "weight_max": 1.6265615521149357e-09, "weight_mean": 7.160426140906395e-10, "weight_min": 5.679522963042416e-10, "weight_std": 3.6835429151338417e-10 }, { "epoch": 0.2748661366217751, "step": 847, "weight_max": 4.187939703115262e-05, "weight_mean": 5.5653117669862695e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4701572581543587e-05 }, { "epoch": 0.2751906539023203, "step": 848, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.391043593292125e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.555088056193199e-05 }, { "epoch": 0.2751906539023203, "step": 848, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.920252523443196e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.8373011698713526e-05 }, { "epoch": 0.2751906539023203, "step": 848, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2563157724798657e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7659103832556866e-05 }, { "epoch": 0.2751906539023203, "step": 848, "weight_max": 2.926079287135508e-05, "weight_mean": 3.6589856335922377e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0344691872887779e-05 }, { "epoch": 0.2755151711828655, "step": 849, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8057349886221346e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3588934962172061e-05 }, { "epoch": 0.2755151711828655, "step": 849, "weight_max": 4.187939703115262e-05, "weight_mean": 9.664884601079393e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7977130482904613e-05 }, { "epoch": 0.2755151711828655, "step": 849, "weight_max": 2.2490561605081894e-05, "weight_mean": 2.811774265865097e-06, "weight_min": 0.0, "weight_std": 7.951431143737864e-06 }, { "epoch": 0.2755151711828655, "step": 849, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.72863199422136e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8005901438300498e-05 }, { "epoch": 0.27583968846341067, "grad_norm": 11.839210510253906, "learning_rate": 8.475510204081633e-06, "loss": 0.1993, "step": 850 }, { "epoch": 0.27583968846341067, "step": 850, "weight_max": 3.722132169059478e-05, "weight_mean": 9.832428077061195e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.558503572596237e-05 }, { "epoch": 0.27583968846341067, "step": 850, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4833034583716653e-05, "weight_min": 7.508524357824342e-10, "weight_std": 2.040118670265656e-05 }, { "epoch": 0.27583968846341067, "step": 850, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5031951988930814e-05, "weight_min": 0.0, "weight_std": 2.0901537936879322e-05 }, { "epoch": 0.27583968846341067, "step": 850, "weight_max": 3.5434961318969727e-05, "weight_mean": 5.62187415198423e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2499519471020903e-05 }, { "epoch": 0.2761642057439559, "step": 851, "weight_max": 4.252122016623616e-05, "weight_mean": 1.521054218756035e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.729693394736387e-05 }, { "epoch": 0.2761642057439559, "step": 851, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631339137034956e-05, "weight_min": 0.0, "weight_std": 1.9687378880917095e-05 }, { "epoch": 0.2761642057439559, "step": 851, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.395724317873828e-06, "weight_min": 0.0, "weight_std": 1.2423171938280575e-05 }, { "epoch": 0.2761642057439559, "step": 851, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4163348168949597e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9614897610154003e-05 }, { "epoch": 0.27648872302450106, "step": 852, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.500936832599109e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.078118995996192e-05 }, { "epoch": 0.27648872302450106, "step": 852, "weight_max": 3.998322426923551e-05, "weight_mean": 1.354278083454119e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8978738808073103e-05 }, { "epoch": 0.27648872302450106, "step": 852, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2737629731418565e-05, "weight_min": 0.0, "weight_std": 1.800617428671103e-05 }, { "epoch": 0.27648872302450106, "step": 852, "weight_max": 3.602553260861896e-05, "weight_mean": 8.644101399113424e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.60227846208727e-05 }, { "epoch": 0.2768132403050462, "step": 853, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4810735592618585e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0488900190684944e-05 }, { "epoch": 0.2768132403050462, "step": 853, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.53701585077215e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1307616407284513e-05 }, { "epoch": 0.2768132403050462, "step": 853, "weight_max": 3.199401908204891e-05, "weight_mean": 3.999934506282443e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1311318303341977e-05 }, { "epoch": 0.2768132403050462, "step": 853, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6074031009338796e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9195986169506796e-05 }, { "epoch": 0.27713775758559145, "step": 854, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4282497179228812e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.039073660853319e-05 }, { "epoch": 0.27713775758559145, "step": 854, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2926017006975599e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8333468688069843e-05 }, { "epoch": 0.27713775758559145, "step": 854, "weight_max": 3.5434961318969727e-05, "weight_mean": 6.1689934227615595e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2787800187652465e-05 }, { "epoch": 0.27713775758559145, "step": 854, "weight_max": 1.4281727999332361e-05, "weight_mean": 1.7860249954537721e-06, "weight_min": 5.958802340444436e-10, "weight_std": 5.049026640335796e-06 }, { "epoch": 0.2774622748661366, "step": 855, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2455260730348527e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.766821696946863e-05 }, { "epoch": 0.2774622748661366, "step": 855, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4043171177036129e-05, "weight_min": 0.0, "weight_std": 1.9584371329983696e-05 }, { "epoch": 0.2774622748661366, "step": 855, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006147931562737e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.857444476627279e-05 }, { "epoch": 0.2774622748661366, "step": 855, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.431463366927346e-06, "weight_min": 0.0, "weight_std": 1.2527305443654768e-05 }, { "epoch": 0.27778679214668184, "step": 856, "weight_max": 6.79051709084888e-06, "weight_mean": 8.493002496834379e-07, "weight_min": 0.0, "weight_std": 2.400614221187425e-06 }, { "epoch": 0.27778679214668184, "step": 856, "weight_max": 3.624825550829769e-09, "weight_mean": 1.1388054943495263e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.0685535789534129e-09 }, { "epoch": 0.27778679214668184, "step": 856, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843705937673803e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3697587746719364e-05 }, { "epoch": 0.27778679214668184, "step": 856, "weight_max": 3.782653220696375e-05, "weight_mean": 1.1813163837359753e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.656697531871032e-05 }, { "epoch": 0.278111309427227, "step": 857, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393182280182373e-06, "weight_min": 0.0, "weight_std": 1.2424198757798877e-05 }, { "epoch": 0.278111309427227, "step": 857, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.805716601666063e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.888972292363178e-05 }, { "epoch": 0.278111309427227, "step": 857, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5272908058250323e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1093141185701825e-05 }, { "epoch": 0.278111309427227, "step": 857, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140777491556946e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1709971659001894e-05 }, { "epoch": 0.27843582670777217, "step": 858, "weight_max": 3.998322426923551e-05, "weight_mean": 7.2291049946215935e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4632825696025975e-05 }, { "epoch": 0.27843582670777217, "step": 858, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0149018382653594e-05, "weight_min": 0.0, "weight_std": 2.1700743673136458e-05 }, { "epoch": 0.27843582670777217, "step": 858, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.055721082200762e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5355361028923653e-05 }, { "epoch": 0.27843582670777217, "step": 858, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.846204319619574e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.36965800265898e-05 }, { "epoch": 0.2787603439883174, "step": 859, "weight_max": 2.8994179956498556e-05, "weight_mean": 3.6248789001547266e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0250745617668144e-05 }, { "epoch": 0.2787603439883174, "step": 859, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392654075985774e-05, "weight_min": 0.0, "weight_std": 1.924726711877156e-05 }, { "epoch": 0.2787603439883174, "step": 859, "weight_max": 4.252122016623616e-05, "weight_mean": 9.83203608484473e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8254366295877844e-05 }, { "epoch": 0.2787603439883174, "step": 859, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0633224519551732e-05, "weight_min": 0.0, "weight_std": 1.9699768017744645e-05 }, { "epoch": 0.27908486126886256, "grad_norm": 0.1580561250448227, "learning_rate": 8.455102040816328e-06, "loss": 0.2396, "step": 860 }, { "epoch": 0.27908486126886256, "step": 860, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0769112931448035e-06, "weight_min": 0.0, "weight_std": 1.4357699001266155e-05 }, { "epoch": 0.27908486126886256, "step": 860, "weight_max": 3.843663216684945e-05, "weight_mean": 8.330197488248814e-06, "weight_min": 0.0, "weight_std": 1.56642054207623e-05 }, { "epoch": 0.27908486126886256, "step": 860, "weight_max": 3.4268457966391e-05, "weight_mean": 8.495649126416538e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5730338418507017e-05 }, { "epoch": 0.27908486126886256, "step": 860, "weight_max": 3.312147964606993e-05, "weight_mean": 4.1418575165153015e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1709536011039745e-05 }, { "epoch": 0.2794093785494078, "step": 861, "weight_max": 3.9360926166409627e-05, "weight_mean": 6.083971129555721e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.3834502169629559e-05 }, { "epoch": 0.2794093785494078, "step": 861, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3159005801717285e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5033218915050384e-05 }, { "epoch": 0.2794093785494078, "step": 861, "weight_max": 2.5776537313504377e-06, "weight_mean": 3.2761064971964515e-07, "weight_min": 0.0, "weight_std": 9.092537993637961e-07 }, { "epoch": 0.2794093785494078, "step": 861, "weight_max": 2.9865134365536505e-07, "weight_mean": 3.800570880230225e-08, "weight_min": 5.958802340444436e-10, "weight_std": 1.0531731930996102e-07 }, { "epoch": 0.27973389582995295, "step": 862, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.467142010369571e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.03589879674837e-05 }, { "epoch": 0.27973389582995295, "step": 862, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.784498817869462e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.029256575042382e-05 }, { "epoch": 0.27973389582995295, "step": 862, "weight_max": 3.397988257347606e-05, "weight_mean": 7.3300548137922306e-06, "weight_min": 0.0, "weight_std": 1.3796180610370357e-05 }, { "epoch": 0.27973389582995295, "step": 862, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070139311806997e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1509748219395988e-05 }, { "epoch": 0.2800584131104981, "step": 863, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155965936864959e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4581145478587132e-05 }, { "epoch": 0.2800584131104981, "step": 863, "weight_max": 3.782653220696375e-05, "weight_mean": 9.048630090546794e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.677625004958827e-05 }, { "epoch": 0.2800584131104981, "step": 863, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4325822121463716e-05, "weight_min": 0.0, "weight_std": 1.9884459106833674e-05 }, { "epoch": 0.2800584131104981, "step": 863, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315854650689289e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5033238014439121e-05 }, { "epoch": 0.28038293039104334, "step": 864, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.70469193108147e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8613649444887415e-05 }, { "epoch": 0.28038293039104334, "step": 864, "weight_max": 3.722132169059478e-05, "weight_mean": 8.796656402410008e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.631568193261046e-05 }, { "epoch": 0.28038293039104334, "step": 864, "weight_max": 3.843663216684945e-05, "weight_mean": 6.498381026176503e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3747741832048632e-05 }, { "epoch": 0.28038293039104334, "step": 864, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5950732631608844e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.2006495782989077e-05 }, { "epoch": 0.2807074476715885, "step": 865, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.929010613821447e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0764069631695747e-05 }, { "epoch": 0.2807074476715885, "step": 865, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.570732638356276e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1684720195480622e-05 }, { "epoch": 0.2807074476715885, "step": 865, "weight_max": 2.820166446326766e-05, "weight_mean": 3.52576830664475e-06, "weight_min": 5.864964625068581e-10, "weight_std": 9.970567589334678e-06 }, { "epoch": 0.2807074476715885, "step": 865, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8630282283993438e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7545691662235186e-05 }, { "epoch": 0.28103196495213373, "step": 866, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8378465028945357e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9934439478674904e-05 }, { "epoch": 0.28103196495213373, "step": 866, "weight_max": 2.926079287135508e-05, "weight_mean": 3.658515424831421e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.034488195728045e-05 }, { "epoch": 0.28103196495213373, "step": 866, "weight_max": 3.255531191825867e-05, "weight_mean": 4.788156275026267e-06, "weight_min": 9.062063877074422e-10, "weight_std": 1.1397925845813006e-05 }, { "epoch": 0.28103196495213373, "step": 866, "weight_max": 3.397988257347606e-05, "weight_mean": 4.250287929608021e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2012571460218169e-05 }, { "epoch": 0.2813564822326789, "step": 867, "weight_max": 3.998322426923551e-05, "weight_mean": 1.249163324246183e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.793720730347559e-05 }, { "epoch": 0.2813564822326789, "step": 867, "weight_max": 3.967146767536178e-05, "weight_mean": 9.17106808628887e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.705513750493992e-05 }, { "epoch": 0.2813564822326789, "step": 867, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7407974155503325e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9921775674447417e-05 }, { "epoch": 0.2813564822326789, "step": 867, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3236132872407325e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8356415239395574e-05 }, { "epoch": 0.28168099951322406, "step": 868, "weight_max": 3.662098970380612e-05, "weight_mean": 8.925595466280356e-06, "weight_min": 0.0, "weight_std": 1.6296546164085157e-05 }, { "epoch": 0.28168099951322406, "step": 868, "weight_max": 3.998322426923551e-05, "weight_mean": 9.322240657638758e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7311014744336717e-05 }, { "epoch": 0.28168099951322406, "step": 868, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0354548976465594e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9082846847595647e-05 }, { "epoch": 0.28168099951322406, "step": 868, "weight_max": 9.532743206364103e-06, "weight_mean": 1.1924286127396044e-06, "weight_min": 5.864964625068581e-10, "weight_std": 3.3699961932143196e-06 }, { "epoch": 0.2820055167937693, "step": 869, "weight_max": 3.782653220696375e-05, "weight_mean": 1.10974133349373e-05, "weight_min": 0.0, "weight_std": 1.6985235561151057e-05 }, { "epoch": 0.2820055167937693, "step": 869, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4173277122608852e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9673991118906997e-05 }, { "epoch": 0.2820055167937693, "step": 869, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615978923538933e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3053016118647065e-05 }, { "epoch": 0.2820055167937693, "step": 869, "weight_max": 2.563926318543963e-05, "weight_mean": 4.921330400975421e-06, "weight_min": 6.100955851628953e-10, "weight_std": 9.651428626966663e-06 }, { "epoch": 0.28233003407431445, "grad_norm": 0.39309433102607727, "learning_rate": 8.43469387755102e-06, "loss": 0.0396, "step": 870 }, { "epoch": 0.28233003407431445, "step": 870, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156256975169526e-06, "weight_min": 0.0, "weight_std": 1.4581028153770603e-05 }, { "epoch": 0.28233003407431445, "step": 870, "weight_max": 4.187939703115262e-05, "weight_mean": 2.8070535336155444e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8125199858332053e-05 }, { "epoch": 0.28233003407431445, "step": 870, "weight_max": 3.782653220696375e-05, "weight_mean": 7.4640811362769455e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4209074834070634e-05 }, { "epoch": 0.28233003407431445, "step": 870, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.526743290014565e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7915837815962732e-05 }, { "epoch": 0.2826545513548597, "step": 871, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396978849603329e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.52617649291642e-05 }, { "epoch": 0.2826545513548597, "step": 871, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235921435087221e-06, "weight_min": 0.0, "weight_std": 1.4806199942540843e-05 }, { "epoch": 0.2826545513548597, "step": 871, "weight_max": 3.602553260861896e-05, "weight_mean": 6.892910278111231e-06, "weight_min": 0.0, "weight_std": 1.3538698112824932e-05 }, { "epoch": 0.2826545513548597, "step": 871, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232261956844013e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8945605916087516e-05 }, { "epoch": 0.28297906863540484, "step": 872, "weight_max": 2.5141427613561973e-05, "weight_mean": 3.143471303701517e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.888516276783776e-06 }, { "epoch": 0.28297906863540484, "step": 872, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.743473699723836e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6333102394128218e-05 }, { "epoch": 0.28297906863540484, "step": 872, "weight_max": 4.252122016623616e-05, "weight_mean": 9.491550372331403e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7741498595569283e-05 }, { "epoch": 0.28297906863540484, "step": 872, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8834360162145458e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.026272522925865e-05 }, { "epoch": 0.28330358591595, "step": 873, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074792044179048e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.865418926172424e-05 }, { "epoch": 0.28330358591595, "step": 873, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2056021660100669e-05, "weight_min": 0.0, "weight_std": 1.7665681298240088e-05 }, { "epoch": 0.28330358591595, "step": 873, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.950071964529343e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.0900513845845126e-05 }, { "epoch": 0.28330358591595, "step": 873, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.416747394017875e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5055353287607431e-05 }, { "epoch": 0.28362810319649523, "step": 874, "weight_max": 4.252122016623616e-05, "weight_mean": 8.841283488436602e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6810037777759135e-05 }, { "epoch": 0.28362810319649523, "step": 874, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.07684262629482e-06, "weight_min": 0.0, "weight_std": 1.435772719560191e-05 }, { "epoch": 0.28362810319649523, "step": 874, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4651361198048107e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.024890636675991e-05 }, { "epoch": 0.28362810319649523, "step": 874, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2355981097207405e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806330909777898e-05 }, { "epoch": 0.2839526204770404, "step": 875, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.685618371993769e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8437016478856094e-05 }, { "epoch": 0.2839526204770404, "step": 875, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4887018551235087e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.058575591945555e-05 }, { "epoch": 0.2839526204770404, "step": 875, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9678900571307167e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1374695279519074e-05 }, { "epoch": 0.2839526204770404, "step": 875, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.388311809743755e-06, "weight_min": 0.0, "weight_std": 1.740593506838195e-05 }, { "epoch": 0.2842771377575856, "step": 876, "weight_max": 3.199401908204891e-05, "weight_mean": 4.000161879957886e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1311226444377098e-05 }, { "epoch": 0.2842771377575856, "step": 876, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.80009774339851e-05, "weight_min": 0.0, "weight_std": 1.9761799194384366e-05 }, { "epoch": 0.2842771377575856, "step": 876, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.901869477995206e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.6490292182425037e-05 }, { "epoch": 0.2842771377575856, "step": 876, "weight_max": 1.3161142176087992e-09, "weight_mean": 5.335508701520553e-10, "weight_min": 0.0, "weight_std": 4.957397270466402e-10 }, { "epoch": 0.2846016550381308, "step": 877, "weight_max": 3.998322426923551e-05, "weight_mean": 9.87992825685069e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8294349501957186e-05 }, { "epoch": 0.2846016550381308, "step": 877, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8012215150520205e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.982583307835739e-05 }, { "epoch": 0.2846016550381308, "step": 877, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.870826781669166e-05, "weight_min": 0.0, "weight_std": 2.005987880693283e-05 }, { "epoch": 0.2846016550381308, "step": 877, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077011792309349e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357658074004576e-05 }, { "epoch": 0.28492617231867595, "step": 878, "weight_max": 3.662098970380612e-05, "weight_mean": 8.50790547701763e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.581294600327965e-05 }, { "epoch": 0.28492617231867595, "step": 878, "weight_max": 3.967146767536178e-05, "weight_mean": 7.163519967434695e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.451194748369744e-05 }, { "epoch": 0.28492617231867595, "step": 878, "weight_max": 3.722132169059478e-05, "weight_mean": 8.412157512793783e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5690729924244806e-05 }, { "epoch": 0.28492617231867595, "step": 878, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0771318456099834e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4357610780280083e-05 }, { "epoch": 0.2852506895992212, "step": 879, "weight_max": 3.397988257347606e-05, "weight_mean": 4.263900336809456e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.20071499623009e-05 }, { "epoch": 0.2852506895992212, "step": 879, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.3206869122514036e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2217822586535476e-05 }, { "epoch": 0.2852506895992212, "step": 879, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920756509818602e-06, "weight_min": 0.0, "weight_std": 1.3915930139773991e-05 }, { "epoch": 0.2852506895992212, "step": 879, "weight_max": 1.6815759451560552e-08, "weight_mean": 3.2022478002602384e-09, "weight_min": 6.29309993005478e-10, "weight_std": 5.541915815143739e-09 }, { "epoch": 0.28557520687976634, "grad_norm": 2.0422401428222656, "learning_rate": 8.414285714285714e-06, "loss": 0.1492, "step": 880 }, { "epoch": 0.28557520687976634, "step": 880, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0124698494328186e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1577992811216973e-05 }, { "epoch": 0.28557520687976634, "step": 880, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156093266123207e-06, "weight_min": 0.0, "weight_std": 1.4581094546883833e-05 }, { "epoch": 0.28557520687976634, "step": 880, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8836799426935613e-05, "weight_min": 0.0, "weight_std": 2.0440042135305703e-05 }, { "epoch": 0.28557520687976634, "step": 880, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653650194086367e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3159326954337303e-05 }, { "epoch": 0.28589972416031156, "step": 881, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.186680482438533e-06, "weight_min": 0.0, "weight_std": 1.4568990081897937e-05 }, { "epoch": 0.28589972416031156, "step": 881, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315679118211847e-06, "weight_min": 0.0, "weight_std": 1.503330895502586e-05 }, { "epoch": 0.28589972416031156, "step": 881, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.461147783440538e-06, "weight_min": 0.0, "weight_std": 1.753572178131435e-05 }, { "epoch": 0.28589972416031156, "step": 881, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390677743998822e-05, "weight_min": 0.0, "weight_std": 1.923944910231512e-05 }, { "epoch": 0.28622424144085673, "step": 882, "weight_max": 3.662098970380612e-05, "weight_mean": 4.579141659633024e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2946861716045532e-05 }, { "epoch": 0.28622424144085673, "step": 882, "weight_max": 4.3167921830900013e-05, "weight_mean": 6.78667493048124e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.520634123153286e-05 }, { "epoch": 0.28622424144085673, "step": 882, "weight_max": 3.5434961318969727e-05, "weight_mean": 9.363822755403817e-06, "weight_min": 0.0, "weight_std": 1.4438473954214714e-05 }, { "epoch": 0.28622424144085673, "step": 882, "weight_max": 4.187939703115262e-05, "weight_mean": 9.199897249345668e-06, "weight_min": 0.0, "weight_std": 1.7248741642106324e-05 }, { "epoch": 0.2865487587214019, "step": 883, "weight_max": 3.967146767536178e-05, "weight_mean": 5.6481339925085194e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.3881803170079365e-05 }, { "epoch": 0.2865487587214019, "step": 883, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3157131434127223e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.768468428053893e-05 }, { "epoch": 0.2865487587214019, "step": 883, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070256181876175e-06, "weight_min": 0.0, "weight_std": 1.1509700016176794e-05 }, { "epoch": 0.2865487587214019, "step": 883, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23564904142404e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4806309991399758e-05 }, { "epoch": 0.2868732760019471, "step": 884, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1482077499968e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8384107534075156e-05 }, { "epoch": 0.2868732760019471, "step": 884, "weight_max": 3.0339444492710754e-05, "weight_mean": 5.249046807875857e-06, "weight_min": 0.0, "weight_std": 1.0922166438831482e-05 }, { "epoch": 0.2868732760019471, "step": 884, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.5777510270709172e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7248336007469334e-05 }, { "epoch": 0.2868732760019471, "step": 884, "weight_max": 2.926079287135508e-05, "weight_mean": 6.647927875746973e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2390744814183563e-05 }, { "epoch": 0.2871977932824923, "step": 885, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.209783997386694e-05, "weight_min": 7.508524357824342e-10, "weight_std": 1.8836584786185995e-05 }, { "epoch": 0.2871977932824923, "step": 885, "weight_max": 4.3167921830900013e-05, "weight_mean": 6.7052014856017195e-06, "weight_min": 0.0, "weight_std": 1.5181800335994922e-05 }, { "epoch": 0.2871977932824923, "step": 885, "weight_max": 4.187939703115262e-05, "weight_mean": 9.555402357364073e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7799322449718602e-05 }, { "epoch": 0.2871977932824923, "step": 885, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.364252047002083e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4508683307212777e-05 }, { "epoch": 0.2875223105630375, "step": 886, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4070079234661534e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0129167751292698e-05 }, { "epoch": 0.2875223105630375, "step": 886, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.356761110102525e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.232083286595298e-05 }, { "epoch": 0.2875223105630375, "step": 886, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2867976693087257e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.7728802049532533e-05 }, { "epoch": 0.2875223105630375, "step": 886, "weight_max": 3.998322426923551e-05, "weight_mean": 8.893762242223602e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.663425427977927e-05 }, { "epoch": 0.2878468278435827, "step": 887, "weight_max": 4.252122016623616e-05, "weight_mean": 8.275394975498784e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6128087736433372e-05 }, { "epoch": 0.2878468278435827, "step": 887, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.081722626229748e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5023120795376599e-05 }, { "epoch": 0.2878468278435827, "step": 887, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541244834399549e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2841626812587492e-05 }, { "epoch": 0.2878468278435827, "step": 887, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7697069779387675e-05, "weight_min": 0.0, "weight_std": 1.9446069927653298e-05 }, { "epoch": 0.28817134512412784, "step": 888, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3515800421591848e-05, "weight_min": 8.267048712262692e-10, "weight_std": 1.997822073462885e-05 }, { "epoch": 0.28817134512412784, "step": 888, "weight_max": 3.662098970380612e-05, "weight_mean": 1.2532688742794562e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.746638008626178e-05 }, { "epoch": 0.28817134512412784, "step": 888, "weight_max": 3.602553260861896e-05, "weight_mean": 4.508142865233822e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2734951269521844e-05 }, { "epoch": 0.28817134512412784, "step": 888, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4015075976203661e-05, "weight_min": 0.0, "weight_std": 1.9427050574449822e-05 }, { "epoch": 0.28849586240467306, "step": 889, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3850021787220612e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9287312170490623e-05 }, { "epoch": 0.28849586240467306, "step": 889, "weight_max": 2.3357602074725037e-08, "weight_mean": 3.640220791467641e-09, "weight_min": 5.864964625068581e-10, "weight_std": 7.970899496001493e-09 }, { "epoch": 0.28849586240467306, "step": 889, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3866150766261853e-05, "weight_min": 0.0, "weight_std": 1.922300725709647e-05 }, { "epoch": 0.28849586240467306, "step": 889, "weight_max": 3.143761568935588e-05, "weight_mean": 3.9597639442945365e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1103013093816116e-05 }, { "epoch": 0.28882037968521823, "grad_norm": 0.6989885568618774, "learning_rate": 8.393877551020409e-06, "loss": 0.1148, "step": 890 }, { "epoch": 0.28882037968521823, "step": 890, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4694014680571854e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0324583601905033e-05 }, { "epoch": 0.28882037968521823, "step": 890, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4325577467388939e-05, "weight_min": 0.0, "weight_std": 1.9847371731884778e-05 }, { "epoch": 0.28882037968521823, "step": 890, "weight_max": 2.674530374946471e-09, "weight_mean": 9.846130577528811e-10, "weight_min": 5.864964625068581e-10, "weight_std": 7.047468164600446e-10 }, { "epoch": 0.28882037968521823, "step": 890, "weight_max": 3.782653220696375e-05, "weight_mean": 6.933140412002103e-06, "weight_min": 0.0, "weight_std": 1.3924172890256159e-05 }, { "epoch": 0.28914489696576345, "step": 891, "weight_max": 3.662098970380612e-05, "weight_mean": 1.6047903045546263e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.750010414980352e-05 }, { "epoch": 0.28914489696576345, "step": 891, "weight_max": 3.9360926166409627e-05, "weight_mean": 2.1600933905574493e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8576569345896132e-05 }, { "epoch": 0.28914489696576345, "step": 891, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2364393923198804e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4805991668254137e-05 }, { "epoch": 0.28914489696576345, "step": 891, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.855988663621247e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3692665561393369e-05 }, { "epoch": 0.2894694142463086, "step": 892, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105367224838119e-06, "weight_min": 0.0, "weight_std": 1.160964893642813e-05 }, { "epoch": 0.2894694142463086, "step": 892, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.737903277098667e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4627723430749029e-05 }, { "epoch": 0.2894694142463086, "step": 892, "weight_max": 2.464847420924343e-05, "weight_mean": 3.0819974199403077e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.714172508916818e-06 }, { "epoch": 0.2894694142463086, "step": 892, "weight_max": 3.22740561387036e-05, "weight_mean": 8.552722647436894e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.4271008694777265e-05 }, { "epoch": 0.2897939315268538, "step": 893, "weight_max": 4.252122016623616e-05, "weight_mean": 1.454061839467613e-05, "weight_min": 0.0, "weight_std": 2.0274566850275733e-05 }, { "epoch": 0.2897939315268538, "step": 893, "weight_max": 4.381950566312298e-05, "weight_mean": 1.7104328435380012e-05, "weight_min": 0.0, "weight_std": 1.9959597921115346e-05 }, { "epoch": 0.2897939315268538, "step": 893, "weight_max": 3.9360926166409627e-05, "weight_mean": 2.2061978597776033e-05, "weight_min": 1.0261104188558079e-09, "weight_std": 1.8378572349320166e-05 }, { "epoch": 0.2897939315268538, "step": 893, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9298200641060248e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.1082592866150662e-05 }, { "epoch": 0.290118448807399, "step": 894, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1569777497206815e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4580736205971334e-05 }, { "epoch": 0.290118448807399, "step": 894, "weight_max": 3.813097282545641e-05, "weight_mean": 4.78459742225823e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3474056686391123e-05 }, { "epoch": 0.290118448807399, "step": 894, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4147777619655244e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9688701286213472e-05 }, { "epoch": 0.290118448807399, "step": 894, "weight_max": 1.841821415382583e-07, "weight_mean": 2.3590891018443472e-08, "weight_min": 5.958802340444436e-10, "weight_std": 6.488868820042626e-08 }, { "epoch": 0.2904429660879442, "step": 895, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.436959781945916e-06, "weight_min": 0.0, "weight_std": 1.5246027032844722e-05 }, { "epoch": 0.2904429660879442, "step": 895, "weight_max": 3.967146767536178e-05, "weight_mean": 8.855161468090955e-06, "weight_min": 0.0, "weight_std": 1.6550948203075677e-05 }, { "epoch": 0.2904429660879442, "step": 895, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233397915726528e-05, "weight_min": 0.0, "weight_std": 1.895393870654516e-05 }, { "epoch": 0.2904429660879442, "step": 895, "weight_max": 3.340639523230493e-05, "weight_mean": 4.185545094514964e-06, "weight_min": 8.267048712262692e-10, "weight_std": 1.1807022929133382e-05 }, { "epoch": 0.2907674833684894, "step": 896, "weight_max": 8.603301404619401e-10, "weight_mean": 5.329085506211584e-10, "weight_min": 0.0, "weight_std": 3.436544382395823e-10 }, { "epoch": 0.2907674833684894, "step": 896, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3290024071466178e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8650795027497225e-05 }, { "epoch": 0.2907674833684894, "step": 896, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0768671826517675e-06, "weight_min": 0.0, "weight_std": 1.4357718100654893e-05 }, { "epoch": 0.2907674833684894, "step": 896, "weight_max": 3.782653220696375e-05, "weight_mean": 7.481853117496939e-06, "weight_min": 0.0, "weight_std": 1.448199873266276e-05 }, { "epoch": 0.29109200064903457, "step": 897, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9596142162045e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4025706150277983e-05 }, { "epoch": 0.29109200064903457, "step": 897, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3320701327756979e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.867177343228832e-05 }, { "epoch": 0.29109200064903457, "step": 897, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6908230463741347e-05, "weight_min": 0.0, "weight_std": 2.0103321730857715e-05 }, { "epoch": 0.29109200064903457, "step": 897, "weight_max": 3.3961886636291183e-09, "weight_mean": 1.3140708521319766e-09, "weight_min": 6.29309993005478e-10, "weight_std": 8.971393627987823e-10 }, { "epoch": 0.29141651792957973, "step": 898, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315802809491288e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.503325802332256e-05 }, { "epoch": 0.29141651792957973, "step": 898, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4191865375323687e-05, "weight_min": 0.0, "weight_std": 1.961495217983611e-05 }, { "epoch": 0.29141651792957973, "step": 898, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3066872270428576e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8349428501096554e-05 }, { "epoch": 0.29141651792957973, "step": 898, "weight_max": 4.252122016623616e-05, "weight_mean": 9.855967618932482e-06, "weight_min": 0.0, "weight_std": 1.8323553376831114e-05 }, { "epoch": 0.29174103521012495, "step": 899, "weight_max": 8.490471659072796e-10, "weight_mean": 5.931886648546936e-10, "weight_min": 0.0, "weight_std": 2.558387390827477e-10 }, { "epoch": 0.29174103521012495, "step": 899, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745115676196292e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.814210736483801e-05 }, { "epoch": 0.29174103521012495, "step": 899, "weight_max": 4.252122016623616e-05, "weight_mean": 3.019423093064688e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8689761418499984e-05 }, { "epoch": 0.29174103521012495, "step": 899, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4855889276077505e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.057743040495552e-05 }, { "epoch": 0.2920655524906701, "grad_norm": 5.131159782409668, "learning_rate": 8.373469387755104e-06, "loss": 0.195, "step": 900 }, { "epoch": 0.2920655524906701, "eval_accuracy": 0.9415, "eval_auc": 0.9869876262181755, "eval_loss": 0.21867325901985168, "eval_runtime": 155.7813, "eval_samples_per_second": 12.839, "eval_steps_per_second": 1.605, "step": 900 }, { "epoch": 0.2920655524906701, "step": 900, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232386557618156e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8945529518532567e-05 }, { "epoch": 0.2920655524906701, "step": 900, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.277116078010295e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7218075299751945e-05 }, { "epoch": 0.2920655524906701, "step": 900, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235569005890284e-06, "weight_min": 0.0, "weight_std": 1.480634182371432e-05 }, { "epoch": 0.2920655524906701, "step": 900, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9595437303651124e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4025736163603142e-05 }, { "epoch": 0.29239006977121534, "step": 901, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2986626643396448e-05, "weight_min": 7.829137893544669e-10, "weight_std": 1.9019773390027694e-05 }, { "epoch": 0.29239006977121534, "step": 901, "weight_max": 3.782653220696375e-05, "weight_mean": 8.623987014289014e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6066311218310148e-05 }, { "epoch": 0.29239006977121534, "step": 901, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7345835658488795e-05, "weight_min": 0.0, "weight_std": 1.8034113963949494e-05 }, { "epoch": 0.29239006977121534, "step": 901, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4891199498379137e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.061146187770646e-05 }, { "epoch": 0.2927145870517605, "step": 902, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.3968001338944305e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5261837688740343e-05 }, { "epoch": 0.2927145870517605, "step": 902, "weight_max": 3.967146767536178e-05, "weight_mean": 9.426430551684462e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.7482936527812853e-05 }, { "epoch": 0.2927145870517605, "step": 902, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3817751096212305e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9273194993729703e-05 }, { "epoch": 0.2927145870517605, "step": 902, "weight_max": 3.998322426923551e-05, "weight_mean": 9.842051440500654e-06, "weight_min": 7.614652242082798e-10, "weight_std": 1.822396370698698e-05 }, { "epoch": 0.2930391043323057, "step": 903, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.469119615649106e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0307281374698505e-05 }, { "epoch": 0.2930391043323057, "step": 903, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.055193024512846e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9543591406545602e-05 }, { "epoch": 0.2930391043323057, "step": 903, "weight_max": 2.3677199351368472e-05, "weight_mean": 2.9611126137751853e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.370563591597602e-06 }, { "epoch": 0.2930391043323057, "step": 903, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9597356337471865e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.402565794705879e-05 }, { "epoch": 0.2933636216128509, "step": 904, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.922378922055941e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8389759134151973e-05 }, { "epoch": 0.2933636216128509, "step": 904, "weight_max": 4.447596802492626e-05, "weight_mean": 1.3720951756113209e-05, "weight_min": 0.0, "weight_std": 1.91328399523627e-05 }, { "epoch": 0.2933636216128509, "step": 904, "weight_max": 3.312147964606993e-05, "weight_mean": 1.162867556558922e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6031006452976726e-05 }, { "epoch": 0.2933636216128509, "step": 904, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076802153780591e-06, "weight_min": 0.0, "weight_std": 1.4357743566506542e-05 }, { "epoch": 0.29368813889339607, "step": 905, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.380724279442802e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8781922335620038e-05 }, { "epoch": 0.29368813889339607, "step": 905, "weight_max": 4.187939703115262e-05, "weight_mean": 1.515344138169894e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0924388081766665e-05 }, { "epoch": 0.29368813889339607, "step": 905, "weight_max": 3.76558251247161e-09, "weight_mean": 1.3202476889517811e-09, "weight_min": 0.0, "weight_std": 1.164984442247885e-09 }, { "epoch": 0.29368813889339607, "step": 905, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315962425811449e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5033193449198734e-05 }, { "epoch": 0.2940126561739413, "step": 906, "weight_max": 3.843663216684945e-05, "weight_mean": 9.47617991187144e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4428086615225766e-05 }, { "epoch": 0.2940126561739413, "step": 906, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006096090364736e-05, "weight_min": 0.0, "weight_std": 1.857447568909265e-05 }, { "epoch": 0.2940126561739413, "step": 906, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006326192524284e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8574333807919174e-05 }, { "epoch": 0.2940126561739413, "step": 906, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.030496585182846e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.5567267837468535e-05 }, { "epoch": 0.29433717345448646, "step": 907, "weight_max": 4.252122016623616e-05, "weight_mean": 9.9432882052497e-06, "weight_min": 0.0, "weight_std": 1.8028773411060683e-05 }, { "epoch": 0.29433717345448646, "step": 907, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156035967956996e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581116374756675e-05 }, { "epoch": 0.29433717345448646, "step": 907, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.767389201442711e-06, "weight_min": 7.508524357824342e-10, "weight_std": 1.8102839021594264e-05 }, { "epoch": 0.29433717345448646, "step": 907, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156394763616845e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4580972674593795e-05 }, { "epoch": 0.2946616907350316, "step": 908, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767096925206715e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.348104251519544e-05 }, { "epoch": 0.2946616907350316, "step": 908, "weight_max": 1.3183556802687235e-05, "weight_mean": 1.6485639662278118e-06, "weight_min": 0.0, "weight_std": 4.660841113945935e-06 }, { "epoch": 0.2946616907350316, "step": 908, "weight_max": 3.752332122530788e-05, "weight_mean": 9.268645044357982e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7162394215120003e-05 }, { "epoch": 0.2946616907350316, "step": 908, "weight_max": 3.998322426923551e-05, "weight_mean": 8.893840458767954e-06, "weight_min": 0.0, "weight_std": 1.6634206986054778e-05 }, { "epoch": 0.29498620801557685, "step": 909, "weight_max": 5.490859766865697e-09, "weight_mean": 1.915694936016621e-09, "weight_min": 0.0, "weight_std": 1.902466850722817e-09 }, { "epoch": 0.29498620801557685, "step": 909, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8998578525497578e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0348119505797513e-05 }, { "epoch": 0.29498620801557685, "step": 909, "weight_max": 2.926079287135508e-05, "weight_mean": 3.718283778653131e-06, "weight_min": 0.0, "weight_std": 1.0322082744096406e-05 }, { "epoch": 0.29498620801557685, "step": 909, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3851588846591767e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.934421743499115e-05 }, { "epoch": 0.295310725296122, "grad_norm": 18.45730972290039, "learning_rate": 8.353061224489797e-06, "loss": 0.1365, "step": 910 }, { "epoch": 0.295310725296122, "step": 910, "weight_max": 3.813097282545641e-05, "weight_mean": 9.495593985775486e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7580376152181998e-05 }, { "epoch": 0.295310725296122, "step": 910, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9804630180005915e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.4017402463650797e-05 }, { "epoch": 0.295310725296122, "step": 910, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.231053819443332e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4552319044014439e-05 }, { "epoch": 0.295310725296122, "step": 910, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0038983418780845e-05, "weight_min": 0.0, "weight_std": 1.8592314518173225e-05 }, { "epoch": 0.29563524257666723, "step": 911, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.3632703485200182e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6962481822702102e-05 }, { "epoch": 0.29563524257666723, "step": 911, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8581244148663245e-05, "weight_min": 7.091466858177853e-10, "weight_std": 2.0045932615175843e-05 }, { "epoch": 0.29563524257666723, "step": 911, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.259778951469343e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5320785678341053e-05 }, { "epoch": 0.29563524257666723, "step": 911, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284069746063324e-06, "weight_min": 0.0, "weight_std": 1.2115522622480057e-05 }, { "epoch": 0.2959597598572124, "step": 912, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5159261238295585e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0939869500580244e-05 }, { "epoch": 0.2959597598572124, "step": 912, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.711189906927757e-05, "weight_min": 0.0, "weight_std": 1.974767292267643e-05 }, { "epoch": 0.2959597598572124, "step": 912, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316210263117682e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.503309431427624e-05 }, { "epoch": 0.2959597598572124, "step": 912, "weight_max": 3.782653220696375e-05, "weight_mean": 9.085177225642838e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6839554518810473e-05 }, { "epoch": 0.29628427713775757, "step": 913, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315755061019445e-06, "weight_min": 0.0, "weight_std": 1.5033278032205999e-05 }, { "epoch": 0.29628427713775757, "step": 913, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236163815425243e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4806101717113052e-05 }, { "epoch": 0.29628427713775757, "step": 913, "weight_max": 4.152462906859e-08, "weight_mean": 6.523485751586122e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.4292045769082051e-08 }, { "epoch": 0.29628427713775757, "step": 913, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155901817488484e-06, "weight_min": 0.0, "weight_std": 1.4581171853933483e-05 }, { "epoch": 0.2966087944183028, "step": 914, "weight_max": 4.252122016623616e-05, "weight_mean": 1.418714236933738e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9748942577280104e-05 }, { "epoch": 0.2966087944183028, "step": 914, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3394590496318415e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.815363793866709e-05 }, { "epoch": 0.2966087944183028, "step": 914, "weight_max": 3.752332122530788e-05, "weight_mean": 4.690993591793813e-06, "weight_min": 0.0, "weight_std": 1.3266264431877062e-05 }, { "epoch": 0.2966087944183028, "step": 914, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691025424108375e-06, "weight_min": 0.0, "weight_std": 1.3266250789456535e-05 }, { "epoch": 0.29693331169884796, "step": 915, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0049197953776456e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.867391256382689e-05 }, { "epoch": 0.29693331169884796, "step": 915, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315778707881691e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.503326802776428e-05 }, { "epoch": 0.29693331169884796, "step": 915, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2605421943590045e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.800447716959752e-05 }, { "epoch": 0.29693331169884796, "step": 915, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7983589714276604e-05, "weight_min": 0.0, "weight_std": 1.941340815392323e-05 }, { "epoch": 0.2972578289793932, "step": 916, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.57052145875059e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.6563632016186602e-05 }, { "epoch": 0.2972578289793932, "step": 916, "weight_max": 4.252122016623616e-05, "weight_mean": 3.4557579056127e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.4644818293163553e-05 }, { "epoch": 0.2972578289793932, "step": 916, "weight_max": 3.998322426923551e-05, "weight_mean": 8.730869012651965e-06, "weight_min": 0.0, "weight_std": 1.637393143028021e-05 }, { "epoch": 0.2972578289793932, "step": 916, "weight_max": 3.36925259034615e-05, "weight_mean": 1.1147010809509084e-05, "weight_min": 0.0, "weight_std": 1.5555029676761478e-05 }, { "epoch": 0.29758234625993835, "step": 917, "weight_max": 3.813097282545641e-05, "weight_mean": 9.086689715331886e-06, "weight_min": 0.0, "weight_std": 1.6851030522957444e-05 }, { "epoch": 0.29758234625993835, "step": 917, "weight_max": 3.967146767536178e-05, "weight_mean": 9.499856787442695e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7611611838219687e-05 }, { "epoch": 0.29758234625993835, "step": 917, "weight_max": 4.187939703115262e-05, "weight_mean": 9.971070539904758e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.750884621287696e-05 }, { "epoch": 0.29758234625993835, "step": 917, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393372819322394e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2424120541254524e-05 }, { "epoch": 0.2979068635404835, "step": 918, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5626006643287838e-05, "weight_min": 0.0, "weight_std": 2.156495611416176e-05 }, { "epoch": 0.2979068635404835, "step": 918, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.884168321150355e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8323336917092092e-05 }, { "epoch": 0.2979068635404835, "step": 918, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.408160632214276e-06, "weight_min": 0.0, "weight_std": 1.5257274753821548e-05 }, { "epoch": 0.2979068635404835, "step": 918, "weight_max": 4.252122016623616e-05, "weight_mean": 2.2446256480179727e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.906713987409603e-05 }, { "epoch": 0.29823138082102874, "step": 919, "weight_max": 1.178765705844853e-05, "weight_mean": 1.4742358871444594e-06, "weight_min": 0.0, "weight_std": 4.1672515180835035e-06 }, { "epoch": 0.29823138082102874, "step": 919, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882055691268761e-06, "weight_min": 0.0, "weight_std": 1.3806583410769235e-05 }, { "epoch": 0.29823138082102874, "step": 919, "weight_max": 1.3870513626557113e-09, "weight_mean": 8.251606065101669e-10, "weight_min": 5.958802340444436e-10, "weight_std": 3.1142668999173395e-10 }, { "epoch": 0.29823138082102874, "step": 919, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1197583262401167e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8385790099273436e-05 }, { "epoch": 0.2985558981015739, "grad_norm": 11.86892318725586, "learning_rate": 8.33265306122449e-06, "loss": 0.1935, "step": 920 }, { "epoch": 0.2985558981015739, "step": 920, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070551767654251e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1509581781865563e-05 }, { "epoch": 0.2985558981015739, "step": 920, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3246793969301507e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8419785192236304e-05 }, { "epoch": 0.2985558981015739, "step": 920, "weight_max": 3.602553260861896e-05, "weight_mean": 8.715236617717892e-06, "weight_min": 0.0, "weight_std": 1.6148342183441855e-05 }, { "epoch": 0.2985558981015739, "step": 920, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4502494195767213e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.005375426961109e-05 }, { "epoch": 0.2988804153821191, "step": 921, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316422630130546e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033007912279572e-05 }, { "epoch": 0.2988804153821191, "step": 921, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.157082796358736e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4580694369215053e-05 }, { "epoch": 0.2988804153821191, "step": 921, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315981070452835e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033187992230523e-05 }, { "epoch": 0.2988804153821191, "step": 921, "weight_max": 2.952862632810138e-05, "weight_mean": 5.5718883231747895e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1017656106560025e-05 }, { "epoch": 0.2992049326626643, "step": 922, "weight_max": 3.843663216684945e-05, "weight_mean": 4.80539392810897e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.358907320536673e-05 }, { "epoch": 0.2992049326626643, "step": 922, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1394442481105216e-05, "weight_min": 0.0, "weight_std": 1.7793901861296035e-05 }, { "epoch": 0.2992049326626643, "step": 922, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4972662029322237e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.074168332910631e-05 }, { "epoch": 0.2992049326626643, "step": 922, "weight_max": 3.967146767536178e-05, "weight_mean": 9.24317464523483e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7174159438582137e-05 }, { "epoch": 0.29952944994320946, "step": 923, "weight_max": 3.123021770079504e-07, "weight_mean": 3.9762682035870967e-08, "weight_min": 5.679522963042416e-10, "weight_std": 1.1012280509703487e-07 }, { "epoch": 0.29952944994320946, "step": 923, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3436672816169448e-05, "weight_min": 0.0, "weight_std": 1.8703038222156465e-05 }, { "epoch": 0.29952944994320946, "step": 923, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5946479834383354e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.201001552748494e-05 }, { "epoch": 0.29952944994320946, "step": 923, "weight_max": 2.716206836339552e-05, "weight_mean": 3.396946794964606e-06, "weight_min": 0.0, "weight_std": 9.602558748156298e-06 }, { "epoch": 0.2998539672237547, "step": 924, "weight_max": 3.843663216684945e-05, "weight_mean": 6.258278517634608e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.3622668120660819e-05 }, { "epoch": 0.2998539672237547, "step": 924, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2910990335512906e-05, "weight_min": 0.0, "weight_std": 1.7923246559803374e-05 }, { "epoch": 0.2998539672237547, "step": 924, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4048197044758126e-05, "weight_min": 0.0, "weight_std": 1.9450089894235134e-05 }, { "epoch": 0.2998539672237547, "step": 924, "weight_max": 4.252122016623616e-05, "weight_mean": 6.856304935354274e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4964116417104378e-05 }, { "epoch": 0.30017848450429985, "step": 925, "weight_max": 4.252122016623616e-05, "weight_mean": 7.055699825286865e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4756004929949995e-05 }, { "epoch": 0.30017848450429985, "step": 925, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9928216715925373e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.1362711777328514e-05 }, { "epoch": 0.30017848450429985, "step": 925, "weight_max": 5.898533217418844e-09, "weight_mean": 1.5545458253996003e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.7884467240492086e-09 }, { "epoch": 0.30017848450429985, "step": 925, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4964345609769225e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0691828467533924e-05 }, { "epoch": 0.30050300178484507, "step": 926, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1560682550189085e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4581102732336149e-05 }, { "epoch": 0.30050300178484507, "step": 926, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.519683519290993e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.885607525764499e-05 }, { "epoch": 0.30050300178484507, "step": 926, "weight_max": 3.722132169059478e-05, "weight_mean": 9.49682726059109e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5733357940916903e-05 }, { "epoch": 0.30050300178484507, "step": 926, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2370341210043989e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7846221453510225e-05 }, { "epoch": 0.30082751906539024, "step": 927, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8052147576527204e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.358914505544817e-05 }, { "epoch": 0.30082751906539024, "step": 927, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155909093446098e-06, "weight_min": 0.0, "weight_std": 1.4581168215954676e-05 }, { "epoch": 0.30082751906539024, "step": 927, "weight_max": 3.998322426923551e-05, "weight_mean": 9.501605745754205e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.762406282068696e-05 }, { "epoch": 0.30082751906539024, "step": 927, "weight_max": 3.8743506593164057e-05, "weight_mean": 6.820196176704485e-06, "weight_min": 0.0, "weight_std": 1.4035823369340505e-05 }, { "epoch": 0.3011520363459354, "step": 928, "weight_max": 3.998322426923551e-05, "weight_mean": 7.809805538272485e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5196517779259011e-05 }, { "epoch": 0.3011520363459354, "step": 928, "weight_max": 3.752332122530788e-05, "weight_mean": 8.517505193594843e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5877751138759777e-05 }, { "epoch": 0.3011520363459354, "step": 928, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3439809663395863e-05, "weight_min": 0.0, "weight_std": 1.8857572285924107e-05 }, { "epoch": 0.3011520363459354, "step": 928, "weight_max": 2.319888153579086e-05, "weight_mean": 2.9005284432059852e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.20177319837967e-06 }, { "epoch": 0.3014765536264806, "step": 929, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5075411283760332e-05, "weight_min": 0.0, "weight_std": 2.081572347378824e-05 }, { "epoch": 0.3014765536264806, "step": 929, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631780241965316e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9687107851495966e-05 }, { "epoch": 0.3014765536264806, "step": 929, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.169216375274118e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8618486137711443e-05 }, { "epoch": 0.3014765536264806, "step": 929, "weight_max": 1.4305072681963793e-09, "weight_mean": 6.491738258063151e-10, "weight_min": 0.0, "weight_std": 3.875455234503278e-10 }, { "epoch": 0.3018010709070258, "grad_norm": 6.0504889488220215, "learning_rate": 8.312244897959184e-06, "loss": 0.2491, "step": 930 }, { "epoch": 0.3018010709070258, "step": 930, "weight_max": 3.722132169059478e-05, "weight_mean": 6.148236479930347e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3234213838586584e-05 }, { "epoch": 0.3018010709070258, "step": 930, "weight_max": 3.006794759130571e-05, "weight_mean": 3.759008677661768e-06, "weight_min": 0.0, "weight_std": 1.0630416909407359e-05 }, { "epoch": 0.3018010709070258, "step": 930, "weight_max": 3.572963396436535e-05, "weight_mean": 4.469182840693975e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2631131539819762e-05 }, { "epoch": 0.3018010709070258, "step": 930, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5117507246031892e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0889596271445043e-05 }, { "epoch": 0.302125588187571, "step": 931, "weight_max": 4.252122016623616e-05, "weight_mean": 1.407558738719672e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.960657573363278e-05 }, { "epoch": 0.302125588187571, "step": 931, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.267719749710523e-06, "weight_min": 0.0, "weight_std": 1.622986928850878e-05 }, { "epoch": 0.302125588187571, "step": 931, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315811449690955e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5033255294838455e-05 }, { "epoch": 0.302125588187571, "step": 931, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4023943549545947e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.6709194824215956e-05 }, { "epoch": 0.3024501054681162, "step": 932, "weight_max": 3.998322426923551e-05, "weight_mean": 8.266214535979088e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.574538146087434e-05 }, { "epoch": 0.3024501054681162, "step": 932, "weight_max": 3.752332122530788e-05, "weight_mean": 4.6912309699109755e-06, "weight_min": 0.0, "weight_std": 1.3266168025438674e-05 }, { "epoch": 0.3024501054681162, "step": 932, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235535354586318e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806356375629548e-05 }, { "epoch": 0.3024501054681162, "step": 932, "weight_max": 4.252122016623616e-05, "weight_mean": 2.458659582771361e-05, "weight_min": 0.0, "weight_std": 2.0619698261725716e-05 }, { "epoch": 0.30277462274866135, "step": 933, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.041472367243841e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.690762110229116e-05 }, { "epoch": 0.30277462274866135, "step": 933, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.725338046933757e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0534818102314603e-05 }, { "epoch": 0.30277462274866135, "step": 933, "weight_max": 4.252122016623616e-05, "weight_mean": 2.6741023248177953e-05, "weight_min": 0.0, "weight_std": 1.7870484953164123e-05 }, { "epoch": 0.30277462274866135, "step": 933, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9117163901682943e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.105612111336086e-05 }, { "epoch": 0.30309914002920657, "step": 934, "weight_max": 4.252122016623616e-05, "weight_mean": 9.006887012219522e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7033589756465517e-05 }, { "epoch": 0.30309914002920657, "step": 934, "weight_max": 3.22740561387036e-05, "weight_mean": 4.034734956803732e-06, "weight_min": 0.0, "weight_std": 1.141040866059484e-05 }, { "epoch": 0.30309914002920657, "step": 934, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930837010557298e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1114417247881647e-05 }, { "epoch": 0.30309914002920657, "step": 934, "weight_max": 4.0610404539620504e-05, "weight_mean": 6.202659733389737e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4255505448090844e-05 }, { "epoch": 0.30342365730975174, "step": 935, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3688358194485772e-05, "weight_min": 0.0, "weight_std": 1.901618088595569e-05 }, { "epoch": 0.30342365730975174, "step": 935, "weight_max": 3.22740561387036e-05, "weight_mean": 7.793660188326612e-06, "weight_min": 0.0, "weight_std": 1.4440849554375745e-05 }, { "epoch": 0.30342365730975174, "step": 935, "weight_max": 1.0385104998178463e-09, "weight_mean": 7.815709190950315e-10, "weight_min": 5.958802340444436e-10, "weight_std": 1.5580053913666347e-10 }, { "epoch": 0.30342365730975174, "step": 935, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.011994512140518e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4774678675166797e-05 }, { "epoch": 0.30374817459029696, "step": 936, "weight_max": 3.967146767536178e-05, "weight_mean": 2.3573498765472323e-05, "weight_min": 0.0, "weight_std": 1.9554805476218462e-05 }, { "epoch": 0.30374817459029696, "step": 936, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3615153875434771e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8929476937046275e-05 }, { "epoch": 0.30374817459029696, "step": 936, "weight_max": 3.572963396436535e-05, "weight_mean": 8.024902854231186e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4984033441578504e-05 }, { "epoch": 0.30374817459029696, "step": 936, "weight_max": 2.919402941614635e-09, "weight_mean": 9.602714179379745e-10, "weight_min": 0.0, "weight_std": 8.612230373294949e-10 }, { "epoch": 0.3040726918708421, "step": 937, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.396570021635853e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7472892068326473e-05 }, { "epoch": 0.3040726918708421, "step": 937, "weight_max": 3.602553260861896e-05, "weight_mean": 8.860133675625548e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.640685513848439e-05 }, { "epoch": 0.3040726918708421, "step": 937, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4168861525831744e-05, "weight_min": 0.0, "weight_std": 1.9189676095265895e-05 }, { "epoch": 0.3040726918708421, "step": 937, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4853866559860762e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0566783859976567e-05 }, { "epoch": 0.3043972091513873, "step": 938, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.466147841711063e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0305207726778463e-05 }, { "epoch": 0.3043972091513873, "step": 938, "weight_max": 5.722029072785517e-09, "weight_mean": 1.3694474443326499e-09, "weight_min": 6.488223291967188e-10, "weight_std": 1.761227941265986e-09 }, { "epoch": 0.3043972091513873, "step": 938, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.958894224837422e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.8442173313815147e-05 }, { "epoch": 0.3043972091513873, "step": 938, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.2710959708783776e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7558568288222887e-05 }, { "epoch": 0.3047217264319325, "step": 939, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.0095484261401e-06, "weight_min": 0.0, "weight_std": 1.4113925317360554e-05 }, { "epoch": 0.3047217264319325, "step": 939, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653551968658576e-06, "weight_min": 0.0, "weight_std": 1.315936697210418e-05 }, { "epoch": 0.3047217264319325, "step": 939, "weight_max": 1.0224462165808745e-08, "weight_mean": 2.054417747032744e-09, "weight_min": 6.100955851628953e-10, "weight_std": 3.330579145810475e-09 }, { "epoch": 0.3047217264319325, "step": 939, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0203144231345505e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.1656931494362652e-05 }, { "epoch": 0.3050462437124777, "grad_norm": 3.5329017639160156, "learning_rate": 8.291836734693877e-06, "loss": 0.0979, "step": 940 }, { "epoch": 0.3050462437124777, "step": 940, "weight_max": 3.36925259034615e-05, "weight_mean": 7.639908290002495e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4243316400097683e-05 }, { "epoch": 0.3050462437124777, "step": 940, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3074490198050626e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.877626527857501e-05 }, { "epoch": 0.3050462437124777, "step": 940, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.357198577054078e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2320655514486134e-05 }, { "epoch": 0.3050462437124777, "step": 940, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156010047357995e-06, "weight_min": 0.0, "weight_std": 1.4581128198187798e-05 }, { "epoch": 0.3053707609930229, "step": 941, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921727850160096e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3915537238062825e-05 }, { "epoch": 0.3053707609930229, "step": 941, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9012417396879755e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.04212137759896e-05 }, { "epoch": 0.3053707609930229, "step": 941, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4842716154817026e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.049175600404851e-05 }, { "epoch": 0.3053707609930229, "step": 941, "weight_max": 3.662098970380612e-05, "weight_mean": 9.081413736566901e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.681478534010239e-05 }, { "epoch": 0.30569527827356807, "step": 942, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3157646107138135e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5033274394227192e-05 }, { "epoch": 0.30569527827356807, "step": 942, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4266347534430679e-05, "weight_min": 0.0, "weight_std": 1.9873532437486574e-05 }, { "epoch": 0.30569527827356807, "step": 942, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9853720004903153e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.129918175342027e-05 }, { "epoch": 0.30569527827356807, "step": 942, "weight_max": 2.872879122151062e-05, "weight_mean": 3.591777840483701e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.015688667393988e-05 }, { "epoch": 0.30601979555411324, "step": 943, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316167062119348e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5033112504170276e-05 }, { "epoch": 0.30601979555411324, "step": 943, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3724336895393208e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9638582671177574e-05 }, { "epoch": 0.30601979555411324, "step": 943, "weight_max": 2.714597657771378e-09, "weight_mean": 1.0555988305682718e-09, "weight_min": 0.0, "weight_std": 7.967289605836925e-10 }, { "epoch": 0.30601979555411324, "step": 943, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.288491128245369e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7296501027885824e-05 }, { "epoch": 0.30634431283465846, "step": 944, "weight_max": 2.2169586166853605e-09, "weight_mean": 1.0026649510663788e-09, "weight_min": 6.100955851628953e-10, "weight_std": 5.574833927823875e-10 }, { "epoch": 0.30634431283465846, "step": 944, "weight_max": 2.7956259529560157e-09, "weight_mean": 9.80993952737208e-10, "weight_min": 5.771870759119224e-10, "weight_std": 7.423683334728537e-10 }, { "epoch": 0.30634431283465846, "step": 944, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.248490752914222e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.765809065545909e-05 }, { "epoch": 0.30634431283465846, "step": 944, "weight_max": 4.252122016623616e-05, "weight_mean": 9.893785318126902e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.8384989743935876e-05 }, { "epoch": 0.30666883011520363, "step": 945, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.6450174118508585e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.750345472828485e-05 }, { "epoch": 0.30666883011520363, "step": 945, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.78978096100036e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.8250933862873353e-05 }, { "epoch": 0.30666883011520363, "step": 945, "weight_max": 1.076157385426768e-09, "weight_mean": 8.332981527026107e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.1181234544087602e-10 }, { "epoch": 0.30666883011520363, "step": 945, "weight_max": 4.252122016623616e-05, "weight_mean": 9.893405149341561e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8385224393568933e-05 }, { "epoch": 0.30699334739574885, "step": 946, "weight_max": 4.252122016623616e-05, "weight_mean": 9.177227184409276e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.727191192912869e-05 }, { "epoch": 0.30699334739574885, "step": 946, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.590137071616482e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.94436597666936e-05 }, { "epoch": 0.30699334739574885, "step": 946, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0785852181434166e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357023246702738e-05 }, { "epoch": 0.30699334739574885, "step": 946, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9234359569964e-05, "weight_min": 0.0, "weight_std": 2.0628864149330184e-05 }, { "epoch": 0.307317864676294, "step": 947, "weight_max": 3.752332122530788e-05, "weight_mean": 1.622080526431091e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7764819858712144e-05 }, { "epoch": 0.307317864676294, "step": 947, "weight_max": 3.843663216684945e-05, "weight_mean": 8.980970960692503e-06, "weight_min": 0.0, "weight_std": 1.6682322893757373e-05 }, { "epoch": 0.307317864676294, "step": 947, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077201421954669e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357582585944328e-05 }, { "epoch": 0.307317864676294, "step": 947, "weight_max": 2.020880492636934e-05, "weight_mean": 2.527131982787978e-06, "weight_min": 6.29309993005478e-10, "weight_std": 7.14447514837957e-06 }, { "epoch": 0.3076423819568392, "step": 948, "weight_max": 2.6141986381844617e-05, "weight_mean": 7.487329639843665e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1203460417164024e-05 }, { "epoch": 0.3076423819568392, "step": 948, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.2297875400690828e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.705072463664692e-05 }, { "epoch": 0.3076423819568392, "step": 948, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4508354070130736e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0100955225643702e-05 }, { "epoch": 0.3076423819568392, "step": 948, "weight_max": 3.843663216684945e-05, "weight_mean": 1.191614046547329e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6838581359479576e-05 }, { "epoch": 0.3079668992373844, "step": 949, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.909695573791396e-06, "weight_min": 0.0, "weight_std": 1.5122840522963088e-05 }, { "epoch": 0.3079668992373844, "step": 949, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5409006664413027e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7856049453257583e-05 }, { "epoch": 0.3079668992373844, "step": 949, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793106088778586e-06, "weight_min": 0.0, "weight_std": 1.0726340406108648e-05 }, { "epoch": 0.3079668992373844, "step": 949, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8268654457642697e-05, "weight_min": 0.0, "weight_std": 1.9657332813949324e-05 }, { "epoch": 0.3082914165179296, "grad_norm": 3.966123104095459, "learning_rate": 8.271428571428572e-06, "loss": 0.1404, "step": 950 }, { "epoch": 0.3082914165179296, "step": 950, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5114799680304714e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0873472749372013e-05 }, { "epoch": 0.3082914165179296, "step": 950, "weight_max": 2.2490561605081894e-05, "weight_mean": 2.8119177386543015e-06, "weight_min": 0.0, "weight_std": 7.95137293607695e-06 }, { "epoch": 0.3082914165179296, "step": 950, "weight_max": 3.752332122530788e-05, "weight_mean": 7.959126378409564e-06, "weight_min": 0.0, "weight_std": 1.5045746295072604e-05 }, { "epoch": 0.3082914165179296, "step": 950, "weight_max": 2.7420132028055377e-05, "weight_mean": 3.42853104484675e-06, "weight_min": 0.0, "weight_std": 9.694070286059286e-06 }, { "epoch": 0.3086159337984748, "step": 951, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631631994328927e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9687198800966144e-05 }, { "epoch": 0.3086159337984748, "step": 951, "weight_max": 3.312147964606993e-05, "weight_mean": 6.90632987243589e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2602105925907381e-05 }, { "epoch": 0.3086159337984748, "step": 951, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4542354620061815e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.012457298405934e-05 }, { "epoch": 0.3086159337984748, "step": 951, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5270294170477428e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.107893851643894e-05 }, { "epoch": 0.30894045107901996, "step": 952, "weight_max": 3.752332122530788e-05, "weight_mean": 1.0604616363707464e-05, "weight_min": 0.0, "weight_std": 1.5928859284031205e-05 }, { "epoch": 0.30894045107901996, "step": 952, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393418294057483e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2424102351360489e-05 }, { "epoch": 0.30894045107901996, "step": 952, "weight_max": 4.187939703115262e-05, "weight_mean": 9.304827472078614e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7407281120540574e-05 }, { "epoch": 0.30894045107901996, "step": 952, "weight_max": 4.252122016623616e-05, "weight_mean": 5.319304818840465e-06, "weight_min": 0.0, "weight_std": 1.5031846487545408e-05 }, { "epoch": 0.30926496835956513, "step": 953, "weight_max": 1.3727150872000493e-05, "weight_mean": 1.7167847090604482e-06, "weight_min": 0.0, "weight_std": 4.852920937992167e-06 }, { "epoch": 0.30926496835956513, "step": 953, "weight_max": 4.187939703115262e-05, "weight_mean": 2.217392648162786e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.879214141808916e-05 }, { "epoch": 0.30926496835956513, "step": 953, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.200458872423042e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4311951417766977e-05 }, { "epoch": 0.30926496835956513, "step": 953, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8929415091406554e-05, "weight_min": 7.298507354036587e-10, "weight_std": 2.098912409564946e-05 }, { "epoch": 0.30958948564011035, "step": 954, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.404133379575796e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.5743366020615213e-05 }, { "epoch": 0.30958948564011035, "step": 954, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7659665900282562e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9099001292488538e-05 }, { "epoch": 0.30958948564011035, "step": 954, "weight_max": 4.187939703115262e-05, "weight_mean": 1.838136449805461e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9795284970314242e-05 }, { "epoch": 0.30958948564011035, "step": 954, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.57377051236108e-06, "weight_min": 0.0, "weight_std": 1.7727183148963377e-05 }, { "epoch": 0.3099140029206555, "step": 955, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4285980796557851e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9975463146693073e-05 }, { "epoch": 0.3099140029206555, "step": 955, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1558758968894836e-06, "weight_min": 0.0, "weight_std": 1.4581181858375203e-05 }, { "epoch": 0.3099140029206555, "step": 955, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3174282685213257e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8612141502671875e-05 }, { "epoch": 0.3099140029206555, "step": 955, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691081812779885e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3266227142594289e-05 }, { "epoch": 0.31023852020120074, "step": 956, "weight_max": 4.381950566312298e-05, "weight_mean": 1.948409953911323e-05, "weight_min": 0.0, "weight_std": 2.0975441657355987e-05 }, { "epoch": 0.31023852020120074, "step": 956, "weight_max": 2.952862632810138e-05, "weight_mean": 7.2827360781957395e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3485321687767282e-05 }, { "epoch": 0.31023852020120074, "step": 956, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.807364242267795e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9576891645556316e-05 }, { "epoch": 0.31023852020120074, "step": 956, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3365761333261617e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9463601347524673e-05 }, { "epoch": 0.3105630374817459, "step": 957, "weight_max": 3.340639523230493e-05, "weight_mean": 8.140676982293371e-06, "weight_min": 0.0, "weight_std": 1.5079202967172023e-05 }, { "epoch": 0.3105630374817459, "step": 957, "weight_max": 1.504423474685268e-09, "weight_mean": 6.287805831561855e-10, "weight_min": 0.0, "weight_std": 4.830106870024053e-10 }, { "epoch": 0.3105630374817459, "step": 957, "weight_max": 2.2256894226302393e-05, "weight_mean": 2.8099786959501216e-06, "weight_min": 6.586902134841921e-10, "weight_std": 7.858105163904838e-06 }, { "epoch": 0.3105630374817459, "step": 957, "weight_max": 4.187939703115262e-05, "weight_mean": 1.519230045232689e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0975903680664487e-05 }, { "epoch": 0.3108875547622911, "step": 958, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320352218201151e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.221795810124604e-05 }, { "epoch": 0.3108875547622911, "step": 958, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5467019693460315e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1348132577259094e-05 }, { "epoch": 0.3108875547622911, "step": 958, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.345358214166481e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8924220057670027e-05 }, { "epoch": 0.3108875547622911, "step": 958, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2886879630968906e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8079204892273992e-05 }, { "epoch": 0.3112120720428363, "step": 959, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.1678373084578197e-05, "weight_min": 0.0, "weight_std": 1.7233920516446233e-05 }, { "epoch": 0.3112120720428363, "step": 959, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.0587318431353197e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7758655303623527e-05 }, { "epoch": 0.3112120720428363, "step": 959, "weight_max": 3.312147964606993e-05, "weight_mean": 4.141313638683641e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1709756108757574e-05 }, { "epoch": 0.3112120720428363, "step": 959, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5214812265185174e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1103278413647786e-05 }, { "epoch": 0.31153658932338146, "grad_norm": 3.59295916557312, "learning_rate": 8.253061224489797e-06, "loss": 0.0868, "step": 960 }, { "epoch": 0.31153658932338146, "step": 960, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3855289580533281e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9164093828294426e-05 }, { "epoch": 0.31153658932338146, "step": 960, "weight_max": 3.860909814079605e-09, "weight_mean": 1.4433088058041221e-09, "weight_min": 5.864964625068581e-10, "weight_std": 1.3335736959163569e-09 }, { "epoch": 0.31153658932338146, "step": 960, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7929498830635566e-06, "weight_min": 0.0, "weight_std": 1.072640316124307e-05 }, { "epoch": 0.31153658932338146, "step": 960, "weight_max": 4.252122016623616e-05, "weight_mean": 8.646980859339237e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.656196945987176e-05 }, { "epoch": 0.3118611066039267, "step": 961, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311193364032079e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9091190551989712e-05 }, { "epoch": 0.3118611066039267, "step": 961, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5786816220497712e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1792873667436652e-05 }, { "epoch": 0.3118611066039267, "step": 961, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3295762073539663e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8959666704176925e-05 }, { "epoch": 0.3118611066039267, "step": 961, "weight_max": 3.9360926166409627e-05, "weight_mean": 2.000493805098813e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.761265957611613e-05 }, { "epoch": 0.31218562388447185, "step": 962, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.1638026990112849e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6308267731801607e-05 }, { "epoch": 0.31218562388447185, "step": 962, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.805386980588082e-06, "weight_min": 0.0, "weight_std": 1.8169388567912392e-05 }, { "epoch": 0.31218562388447185, "step": 962, "weight_max": 3.967146767536178e-05, "weight_mean": 1.888870247057639e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0244449842721224e-05 }, { "epoch": 0.31218562388447185, "step": 962, "weight_max": 2.3835209361777743e-09, "weight_mean": 8.216335944943864e-10, "weight_min": 0.0, "weight_std": 6.84777512471868e-10 }, { "epoch": 0.312510141165017, "step": 963, "weight_max": 3.671447146302853e-09, "weight_mean": 1.0312376508281318e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.0700745844971493e-09 }, { "epoch": 0.312510141165017, "step": 963, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248497589287581e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2013293599011377e-05 }, { "epoch": 0.312510141165017, "step": 963, "weight_max": 4.187939703115262e-05, "weight_mean": 9.23498737392947e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7300788385909982e-05 }, { "epoch": 0.312510141165017, "step": 963, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4335485502670053e-05, "weight_min": 0.0, "weight_std": 1.9935800082748756e-05 }, { "epoch": 0.31283465844556224, "step": 964, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5409515223582275e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2841745956393424e-05 }, { "epoch": 0.31283465844556224, "step": 964, "weight_max": 3.782653220696375e-05, "weight_mean": 9.195069651468657e-06, "weight_min": 0.0, "weight_std": 1.7033826225087978e-05 }, { "epoch": 0.31283465844556224, "step": 964, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3893037248635665e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9216253349441104e-05 }, { "epoch": 0.31283465844556224, "step": 964, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2953465557075106e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.800853897293564e-05 }, { "epoch": 0.3131591757261074, "step": 965, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.22331673791632e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7402860976289958e-05 }, { "epoch": 0.3131591757261074, "step": 965, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3447500350594055e-05, "weight_min": 8.490471659072796e-10, "weight_std": 1.8550699678598903e-05 }, { "epoch": 0.3131591757261074, "step": 965, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467161943466635e-06, "weight_min": 0.0, "weight_std": 1.2631945537577849e-05 }, { "epoch": 0.3131591757261074, "step": 965, "weight_max": 3.116124207735993e-05, "weight_mean": 3.896979706041748e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1016424650733825e-05 }, { "epoch": 0.31348369300665263, "step": 966, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965075848100241e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1212747267563827e-05 }, { "epoch": 0.31348369300665263, "step": 966, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.617220383806853e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3052514987066388e-05 }, { "epoch": 0.31348369300665263, "step": 966, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0416807526780758e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.866795173555147e-05 }, { "epoch": 0.31348369300665263, "step": 966, "weight_max": 3.752332122530788e-05, "weight_mean": 1.081731534213759e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.6887277524801902e-05 }, { "epoch": 0.3138082102871978, "step": 967, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9600080274103675e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4025546988705173e-05 }, { "epoch": 0.3138082102871978, "step": 967, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8959435692522675e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0604855308192782e-05 }, { "epoch": 0.3138082102871978, "step": 967, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7673944967100397e-05, "weight_min": 7.614652242082798e-10, "weight_std": 1.9162360331392847e-05 }, { "epoch": 0.3138082102871978, "step": 967, "weight_max": 3.662098970380612e-05, "weight_mean": 1.4237724826671183e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.640606569708325e-05 }, { "epoch": 0.31413272756774296, "step": 968, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767716291098623e-06, "weight_min": 0.0, "weight_std": 1.348079058516305e-05 }, { "epoch": 0.31413272756774296, "step": 968, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.324384336650837e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7354637748212554e-05 }, { "epoch": 0.31413272756774296, "step": 968, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3161165851634e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5033132513053715e-05 }, { "epoch": 0.31413272756774296, "step": 968, "weight_max": 8.494970643369015e-06, "weight_mean": 1.0625147979226313e-06, "weight_min": 5.864964625068581e-10, "weight_std": 3.0031656024220865e-06 }, { "epoch": 0.3144572448482882, "step": 969, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.055132639128715e-06, "weight_min": 0.0, "weight_std": 1.5010557035566308e-05 }, { "epoch": 0.3144572448482882, "step": 969, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2516868991951924e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8055638065561652e-05 }, { "epoch": 0.3144572448482882, "step": 969, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3966453479952179e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9309747585793957e-05 }, { "epoch": 0.3144572448482882, "step": 969, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3160565585130826e-06, "weight_min": 0.0, "weight_std": 1.5033156159915961e-05 }, { "epoch": 0.31478176212883335, "grad_norm": 4.628645896911621, "learning_rate": 8.23265306122449e-06, "loss": 0.145, "step": 970 }, { "epoch": 0.31478176212883335, "step": 970, "weight_max": 3.967146767536178e-05, "weight_mean": 9.244819921150338e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7173148080473766e-05 }, { "epoch": 0.31478176212883335, "step": 970, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2964395864401013e-05, "weight_min": 0.0, "weight_std": 1.8237567928736098e-05 }, { "epoch": 0.31478176212883335, "step": 970, "weight_max": 2.872879122151062e-05, "weight_mean": 3.5919372294301866e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0156823009310756e-05 }, { "epoch": 0.31478176212883335, "step": 970, "weight_max": 1.9330267605255358e-05, "weight_mean": 2.4169373773474945e-06, "weight_min": 0.0, "weight_std": 6.834017312939977e-06 }, { "epoch": 0.3151062794093786, "step": 971, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076910383650102e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357699910760857e-05 }, { "epoch": 0.3151062794093786, "step": 971, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2027308912365697e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.7699620002531447e-05 }, { "epoch": 0.3151062794093786, "step": 971, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2462472113838885e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8616039596963674e-05 }, { "epoch": 0.3151062794093786, "step": 971, "weight_max": 4.381950566312298e-05, "weight_mean": 9.069061889022123e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7268921510549262e-05 }, { "epoch": 0.31543079668992374, "step": 972, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8901293515227735e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.0312227206886746e-05 }, { "epoch": 0.31543079668992374, "step": 972, "weight_max": 3.722132169059478e-05, "weight_mean": 1.3551490155805368e-05, "weight_min": 0.0, "weight_std": 1.87225505214883e-05 }, { "epoch": 0.31543079668992374, "step": 972, "weight_max": 2.716206836339552e-05, "weight_mean": 7.754546459182166e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.095490915759001e-05 }, { "epoch": 0.31543079668992374, "step": 972, "weight_max": 3.998322426923551e-05, "weight_mean": 1.393077036482282e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9270037228125148e-05 }, { "epoch": 0.3157553139704689, "step": 973, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2388427421683446e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.757601239660289e-05 }, { "epoch": 0.3157553139704689, "step": 973, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3515473256120458e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.89788388524903e-05 }, { "epoch": 0.3157553139704689, "step": 973, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0195185495831538e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8883760276366957e-05 }, { "epoch": 0.3157553139704689, "step": 973, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8000466297962703e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9884644643752836e-05 }, { "epoch": 0.31607983125101413, "step": 974, "weight_max": 3.28377827827353e-05, "weight_mean": 7.248175279528368e-06, "weight_min": 0.0, "weight_std": 1.3575835509982426e-05 }, { "epoch": 0.31607983125101413, "step": 974, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541088856058195e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2841690477216616e-05 }, { "epoch": 0.31607983125101413, "step": 974, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5705980331404135e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1677689801435918e-05 }, { "epoch": 0.31607983125101413, "step": 974, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4515678230964113e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.013821358559653e-05 }, { "epoch": 0.3164043485315593, "step": 975, "weight_max": 2.8994179956498556e-05, "weight_mean": 1.0351312084821984e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.4300067959993612e-05 }, { "epoch": 0.3164043485315593, "step": 975, "weight_max": 4.252122016623616e-05, "weight_mean": 2.297655555594247e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.912374864332378e-05 }, { "epoch": 0.3164043485315593, "step": 975, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077547029941343e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4357443433254957e-05 }, { "epoch": 0.3164043485315593, "step": 975, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.167166364117293e-06, "weight_min": 0.0, "weight_std": 1.4576654393749777e-05 }, { "epoch": 0.3167288658121045, "step": 976, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.00360366559471e-05, "weight_min": 0.0, "weight_std": 1.8582839402370155e-05 }, { "epoch": 0.3167288658121045, "step": 976, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.783387051546015e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.9241117115598172e-05 }, { "epoch": 0.3167288658121045, "step": 976, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3717051842832007e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8386839656159282e-05 }, { "epoch": 0.3167288658121045, "step": 976, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827399268629961e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0822676813404541e-05 }, { "epoch": 0.3170533830926497, "step": 977, "weight_max": 1.8352018793521552e-09, "weight_mean": 8.554659203241499e-10, "weight_min": 6.100955851628953e-10, "weight_std": 4.1515010873460767e-10 }, { "epoch": 0.3170533830926497, "step": 977, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2452550436137244e-05, "weight_min": 0.0, "weight_std": 1.7655018382356502e-05 }, { "epoch": 0.3170533830926497, "step": 977, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.654011612117756e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8700293367146514e-05 }, { "epoch": 0.3170533830926497, "step": 977, "weight_max": 1.3301527657105794e-09, "weight_mean": 7.487183095733485e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.689471145789213e-10 }, { "epoch": 0.31737790037319485, "step": 978, "weight_max": 4.187939703115262e-05, "weight_mean": 1.386281473969575e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9362647435627878e-05 }, { "epoch": 0.31737790037319485, "step": 978, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3030025002080947e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8689515854930505e-05 }, { "epoch": 0.31737790037319485, "step": 978, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392070180387236e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9247627278673463e-05 }, { "epoch": 0.31737790037319485, "step": 978, "weight_max": 4.381950566312298e-05, "weight_mean": 9.639888958190568e-06, "weight_min": 0.0, "weight_std": 1.8022970834863372e-05 }, { "epoch": 0.3177024176537401, "step": 979, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8787708540912718e-05, "weight_min": 0.0, "weight_std": 1.7054246200132184e-05 }, { "epoch": 0.3177024176537401, "step": 979, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.469460565014742e-06, "weight_min": 0.0, "weight_std": 1.7593685697647743e-05 }, { "epoch": 0.3177024176537401, "step": 979, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315963335306151e-06, "weight_min": 0.0, "weight_std": 1.5033193449198734e-05 }, { "epoch": 0.3177024176537401, "step": 979, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3647772902913857e-05, "weight_min": 0.0, "weight_std": 1.9096445612376556e-05 }, { "epoch": 0.31802693493428524, "grad_norm": 3.5217158794403076, "learning_rate": 8.212244897959185e-06, "loss": 0.2271, "step": 980 }, { "epoch": 0.31802693493428524, "step": 980, "weight_max": 4.381950566312298e-05, "weight_mean": 1.949878060258925e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1055349861853756e-05 }, { "epoch": 0.31802693493428524, "step": 980, "weight_max": 2.3459858500274322e-09, "weight_mean": 1.050599829355292e-09, "weight_min": 5.864964625068581e-10, "weight_std": 7.410574376365275e-10 }, { "epoch": 0.31802693493428524, "step": 980, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355533049732912e-05, "weight_min": 0.0, "weight_std": 1.9195975255570374e-05 }, { "epoch": 0.31802693493428524, "step": 980, "weight_max": 3.8743506593164057e-05, "weight_mean": 2.220673195552081e-05, "weight_min": 0.0, "weight_std": 1.8485068721929565e-05 }, { "epoch": 0.31835145221483047, "step": 981, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1011818060069345e-05, "weight_min": 0.0, "weight_std": 1.6284622688544914e-05 }, { "epoch": 0.31835145221483047, "step": 981, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.498018360114656e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5745814380352385e-05 }, { "epoch": 0.31835145221483047, "step": 981, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8302120224689133e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.999062624236103e-05 }, { "epoch": 0.31835145221483047, "step": 981, "weight_max": 1.901931279135738e-09, "weight_mean": 9.676957013482479e-10, "weight_min": 6.053385570581327e-10, "weight_std": 4.789958985007559e-10 }, { "epoch": 0.31867596949537563, "step": 982, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.1801883121952415e-05, "weight_min": 8.156453290553145e-10, "weight_std": 1.83093870873563e-05 }, { "epoch": 0.31867596949537563, "step": 982, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998741133022122e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4135865967546124e-05 }, { "epoch": 0.31867596949537563, "step": 982, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6094103557406925e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8583050405140966e-05 }, { "epoch": 0.31867596949537563, "step": 982, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.580088064656593e-05, "weight_min": 7.721521200210191e-10, "weight_std": 1.708653690002393e-05 }, { "epoch": 0.3190004867759208, "step": 983, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.555409107822925e-05, "weight_min": 0.0, "weight_std": 2.1473011656780727e-05 }, { "epoch": 0.3190004867759208, "step": 983, "weight_max": 4.187939703115262e-05, "weight_mean": 9.227894224750344e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7231848687515594e-05 }, { "epoch": 0.3190004867759208, "step": 983, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248938694217941e-06, "weight_min": 0.0, "weight_std": 1.2013115338049829e-05 }, { "epoch": 0.3190004867759208, "step": 983, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4658962754765525e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0294071873649955e-05 }, { "epoch": 0.319325004056466, "step": 984, "weight_max": 2.7956259529560157e-09, "weight_mean": 1.008885863740261e-09, "weight_min": 5.864964625068581e-10, "weight_std": 7.344712060763925e-10 }, { "epoch": 0.319325004056466, "step": 984, "weight_max": 3.397988257347606e-05, "weight_mean": 4.2485194171604235e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2013284504064359e-05 }, { "epoch": 0.319325004056466, "step": 984, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316210263117682e-06, "weight_min": 0.0, "weight_std": 1.503309431427624e-05 }, { "epoch": 0.319325004056466, "step": 984, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541611815511715e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2841478564951103e-05 }, { "epoch": 0.3196495213370112, "step": 985, "weight_max": 1.9303806553239156e-08, "weight_mean": 2.8974247445745505e-09, "weight_min": 0.0, "weight_std": 6.633245952514244e-09 }, { "epoch": 0.3196495213370112, "step": 985, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.813480235403404e-06, "weight_min": 0.0, "weight_std": 1.6629213860142045e-05 }, { "epoch": 0.3196495213370112, "step": 985, "weight_max": 3.578502383305704e-09, "weight_mean": 1.057856469088847e-09, "weight_min": 0.0, "weight_std": 1.0764764635240454e-09 }, { "epoch": 0.3196495213370112, "step": 985, "weight_max": 2.056428582974945e-09, "weight_mean": 8.465755874098591e-10, "weight_min": 0.0, "weight_std": 5.781756740041999e-10 }, { "epoch": 0.3199740386175564, "step": 986, "weight_max": 1.901931279135738e-09, "weight_mean": 9.227502650865915e-10, "weight_min": 5.864964625068581e-10, "weight_std": 4.588035229513565e-10 }, { "epoch": 0.3199740386175564, "step": 986, "weight_max": 3.998322426923551e-05, "weight_mean": 5.345102181308903e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4016895875101909e-05 }, { "epoch": 0.3199740386175564, "step": 986, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.3211198317294475e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2217647963552736e-05 }, { "epoch": 0.3199740386175564, "step": 986, "weight_max": 3.632264997577295e-05, "weight_mean": 8.333425284945406e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5511512174271047e-05 }, { "epoch": 0.3202985558981016, "step": 987, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.147632226813585e-06, "weight_min": 0.0, "weight_std": 1.7069689420168288e-05 }, { "epoch": 0.3202985558981016, "step": 987, "weight_max": 4.187939703115262e-05, "weight_mean": 5.237647656031186e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4805502360104583e-05 }, { "epoch": 0.3202985558981016, "step": 987, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9481094568618573e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.09597001230577e-05 }, { "epoch": 0.3202985558981016, "step": 987, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4919947716407478e-05, "weight_min": 0.0, "weight_std": 2.060766564682126e-05 }, { "epoch": 0.32062307317864674, "step": 988, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9643975974759087e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.1163868950679898e-05 }, { "epoch": 0.32062307317864674, "step": 988, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.115345629747026e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4342565918923356e-05 }, { "epoch": 0.32062307317864674, "step": 988, "weight_max": 2.952862632810138e-05, "weight_mean": 5.620495812763693e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.1065482794947457e-05 }, { "epoch": 0.32062307317864674, "step": 988, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.0017565475427546e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7512629710836336e-05 }, { "epoch": 0.32094759045919197, "step": 989, "weight_max": 3.998322426923551e-05, "weight_mean": 1.624386277399026e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8472741430741735e-05 }, { "epoch": 0.32094759045919197, "step": 989, "weight_max": 3.722132169059478e-05, "weight_mean": 6.8313411247800104e-06, "weight_min": 0.0, "weight_std": 1.3709837730857544e-05 }, { "epoch": 0.32094759045919197, "step": 989, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.2503363905125298e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7302698324783705e-05 }, { "epoch": 0.32094759045919197, "step": 989, "weight_max": 3.397988257347606e-05, "weight_mean": 7.23781386113842e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3667607163370121e-05 }, { "epoch": 0.32127210773973713, "grad_norm": 1.0980206727981567, "learning_rate": 8.191836734693878e-06, "loss": 0.2624, "step": 990 }, { "epoch": 0.32127210773973713, "step": 990, "weight_max": 4.252122016623616e-05, "weight_mean": 9.599278200767003e-06, "weight_min": 0.0, "weight_std": 1.7909344023792073e-05 }, { "epoch": 0.32127210773973713, "step": 990, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9989234867098276e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.413579229847528e-05 }, { "epoch": 0.32127210773973713, "step": 990, "weight_max": 2.595289316786875e-09, "weight_mean": 8.294289144394895e-10, "weight_min": 0.0, "weight_std": 7.550309821802159e-10 }, { "epoch": 0.32127210773973713, "step": 990, "weight_max": 4.252122016623616e-05, "weight_mean": 7.704784366069362e-06, "weight_min": 0.0, "weight_std": 1.55768939293921e-05 }, { "epoch": 0.32159662502028236, "step": 991, "weight_max": 1.0698121499785884e-08, "weight_mean": 1.9820323160502085e-09, "weight_min": 5.679522963042416e-10, "weight_std": 3.529626368958816e-09 }, { "epoch": 0.32159662502028236, "step": 991, "weight_max": 1.1845028957679915e-08, "weight_mean": 2.106169461058016e-09, "weight_min": 0.0, "weight_std": 3.9558791797844606e-09 }, { "epoch": 0.32159662502028236, "step": 991, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.747015630826354e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9022914784727618e-05 }, { "epoch": 0.32159662502028236, "step": 991, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5993715351214632e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8023536540567875e-05 }, { "epoch": 0.3219211423008275, "step": 992, "weight_max": 3.006794759130571e-05, "weight_mean": 7.350383384618908e-06, "weight_min": 0.0, "weight_std": 1.3613017472380307e-05 }, { "epoch": 0.3219211423008275, "step": 992, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2548217455332633e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7568967450642958e-05 }, { "epoch": 0.3219211423008275, "step": 992, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0164516172371805e-05, "weight_min": 0.0, "weight_std": 2.162521741411183e-05 }, { "epoch": 0.3219211423008275, "step": 992, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076961315353401e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357679901877418e-05 }, { "epoch": 0.3222456595813727, "step": 993, "weight_max": 3.813097282545641e-05, "weight_mean": 1.359229918307392e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.879093724710401e-05 }, { "epoch": 0.3222456595813727, "step": 993, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.047291971190134e-05, "weight_min": 0.0, "weight_std": 1.9402570615056902e-05 }, { "epoch": 0.3222456595813727, "step": 993, "weight_max": 5.722029072785517e-09, "weight_mean": 1.3835848022836217e-09, "weight_min": 6.053385570581327e-10, "weight_std": 1.7651556882825048e-09 }, { "epoch": 0.3222456595813727, "step": 993, "weight_max": 4.252122016623616e-05, "weight_mean": 5.342335043678759e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5022719708213117e-05 }, { "epoch": 0.3225701768619179, "step": 994, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4527363418892492e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0234207113389857e-05 }, { "epoch": 0.3225701768619179, "step": 994, "weight_max": 3.722132169059478e-05, "weight_mean": 5.093792879051762e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3039890291111078e-05 }, { "epoch": 0.3225701768619179, "step": 994, "weight_max": 3.0339444492710754e-05, "weight_mean": 1.1041955076507293e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.5242838344420306e-05 }, { "epoch": 0.3225701768619179, "step": 994, "weight_max": 3.632264997577295e-05, "weight_mean": 5.849491117260186e-06, "weight_min": 0.0, "weight_std": 1.284645804844331e-05 }, { "epoch": 0.3228946941424631, "step": 995, "weight_max": 2.056428582974945e-09, "weight_mean": 8.825188357874936e-10, "weight_min": 0.0, "weight_std": 6.053281209617012e-10 }, { "epoch": 0.3228946941424631, "step": 995, "weight_max": 3.998322426923551e-05, "weight_mean": 2.3004613467492163e-05, "weight_min": 9.062063877074422e-10, "weight_std": 1.8144319255952723e-05 }, { "epoch": 0.3228946941424631, "step": 995, "weight_max": 2.846461757144425e-05, "weight_mean": 3.5589878280006815e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0063393347081728e-05 }, { "epoch": 0.3228946941424631, "step": 995, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.46507623969228e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4521809134748764e-05 }, { "epoch": 0.3232192114230083, "step": 996, "weight_max": 3.143761568935588e-05, "weight_mean": 3.9302049117395654e-06, "weight_min": 0.0, "weight_std": 1.1114672815892845e-05 }, { "epoch": 0.3232192114230083, "step": 996, "weight_max": 4.252122016623616e-05, "weight_mean": 1.911446997837629e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0774004951817915e-05 }, { "epoch": 0.3232192114230083, "step": 996, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0267601282976102e-05, "weight_min": 0.0, "weight_std": 1.790874739526771e-05 }, { "epoch": 0.3232192114230083, "step": 996, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998697932023788e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135883247945458e-05 }, { "epoch": 0.32354372870355347, "step": 997, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3972580745758023e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9344928659847938e-05 }, { "epoch": 0.32354372870355347, "step": 997, "weight_max": 4.187939703115262e-05, "weight_mean": 9.269990187021904e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.735362093313597e-05 }, { "epoch": 0.32354372870355347, "step": 997, "weight_max": 4.252122016623616e-05, "weight_mean": 1.32234490592964e-05, "weight_min": 0.0, "weight_std": 1.860939482867252e-05 }, { "epoch": 0.32354372870355347, "step": 997, "weight_max": 4.252122016623616e-05, "weight_mean": 9.006997061078437e-06, "weight_min": 0.0, "weight_std": 1.7033522453857586e-05 }, { "epoch": 0.32386824598409863, "step": 998, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.998991117754485e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8524737242842093e-05 }, { "epoch": 0.32386824598409863, "step": 998, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3054404917056672e-05, "weight_min": 0.0, "weight_std": 1.827845881052781e-05 }, { "epoch": 0.32386824598409863, "step": 998, "weight_max": 2.6141986381844617e-05, "weight_mean": 3.268532509537181e-06, "weight_min": 0.0, "weight_std": 9.242270607501268e-06 }, { "epoch": 0.32386824598409863, "step": 998, "weight_max": 4.187939703115262e-05, "weight_mean": 8.347188668267336e-06, "weight_min": 0.0, "weight_std": 1.610770596016664e-05 }, { "epoch": 0.32419276326464386, "step": 999, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.055187385645695e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9543627786333673e-05 }, { "epoch": 0.32419276326464386, "step": 999, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.729482371767517e-06, "weight_min": 0.0, "weight_std": 1.800537938834168e-05 }, { "epoch": 0.32419276326464386, "step": 999, "weight_max": 4.187939703115262e-05, "weight_mean": 1.562639226904139e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.156463779101614e-05 }, { "epoch": 0.32419276326464386, "step": 999, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882878783973865e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3806251445203088e-05 }, { "epoch": 0.324517280545189, "grad_norm": 0.6674030423164368, "learning_rate": 8.171428571428573e-06, "loss": 0.1499, "step": 1000 }, { "epoch": 0.324517280545189, "eval_accuracy": 0.9675, "eval_auc": 0.983657646701125, "eval_loss": 0.149508997797966, "eval_runtime": 155.5847, "eval_samples_per_second": 12.855, "eval_steps_per_second": 1.607, "step": 1000 }, { "epoch": 0.324517280545189, "step": 1000, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0806679610104766e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4356184692587703e-05 }, { "epoch": 0.324517280545189, "step": 1000, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.966430085711181e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.121219975175336e-05 }, { "epoch": 0.324517280545189, "step": 1000, "weight_max": 3.843663216684945e-05, "weight_mean": 8.396566954615992e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5760255337227136e-05 }, { "epoch": 0.324517280545189, "step": 1000, "weight_max": 4.187939703115262e-05, "weight_mean": 9.340423275716603e-06, "weight_min": 0.0, "weight_std": 1.746120506140869e-05 }, { "epoch": 0.32484179782573425, "step": 1001, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3386301361606456e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8581176846055314e-05 }, { "epoch": 0.32484179782573425, "step": 1001, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0059127564309165e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7977006791625172e-05 }, { "epoch": 0.32484179782573425, "step": 1001, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.691682523931377e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.605750003363937e-05 }, { "epoch": 0.32484179782573425, "step": 1001, "weight_max": 1.8022843217835316e-09, "weight_mean": 7.915876842901071e-10, "weight_min": 5.679522963042416e-10, "weight_std": 4.121393226697023e-10 }, { "epoch": 0.3251663151062794, "step": 1002, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0158782970393077e-05, "weight_min": 0.0, "weight_std": 1.8835780792869627e-05 }, { "epoch": 0.3251663151062794, "step": 1002, "weight_max": 4.3167921830900013e-05, "weight_mean": 7.929938874440268e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5896079275989905e-05 }, { "epoch": 0.3251663151062794, "step": 1002, "weight_max": 3.36925259034615e-05, "weight_mean": 4.243417606630828e-06, "weight_min": 0.0, "weight_std": 1.1899556739081163e-05 }, { "epoch": 0.3251663151062794, "step": 1002, "weight_max": 2.6649584469851106e-05, "weight_mean": 4.439001713762991e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.494659025222063e-06 }, { "epoch": 0.3254908323868246, "step": 1003, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5901767255854793e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.187089921790175e-05 }, { "epoch": 0.3254908323868246, "step": 1003, "weight_max": 3.3183479075660216e-08, "weight_mean": 5.062223085161577e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.1381067643867482e-08 }, { "epoch": 0.3254908323868246, "step": 1003, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.120965503621846e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.308317223447375e-05 }, { "epoch": 0.3254908323868246, "step": 1003, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311827281839214e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.909531965793576e-05 }, { "epoch": 0.3258153496673698, "step": 1004, "weight_max": 3.998322426923551e-05, "weight_mean": 5.776793841505423e-06, "weight_min": 0.0, "weight_std": 1.399193934048526e-05 }, { "epoch": 0.3258153496673698, "step": 1004, "weight_max": 4.252122016623616e-05, "weight_mean": 6.026278697390808e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4879534319334198e-05 }, { "epoch": 0.3258153496673698, "step": 1004, "weight_max": 3.572963396436535e-05, "weight_mean": 4.46700505563058e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2632010111701675e-05 }, { "epoch": 0.3258153496673698, "step": 1004, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.041182238433976e-06, "weight_min": 0.0, "weight_std": 1.6907799363252707e-05 }, { "epoch": 0.32613986694791497, "step": 1005, "weight_max": 3.998322426923551e-05, "weight_mean": 7.63671505410457e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.501238966739038e-05 }, { "epoch": 0.32613986694791497, "step": 1005, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631458280840889e-05, "weight_min": 0.0, "weight_std": 1.9687306121340953e-05 }, { "epoch": 0.32613986694791497, "step": 1005, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.74759462021757e-06, "weight_min": 0.0, "weight_std": 1.5258410712704062e-05 }, { "epoch": 0.32613986694791497, "step": 1005, "weight_max": 3.4268457966391e-05, "weight_mean": 4.2841584217967466e-06, "weight_min": 0.0, "weight_std": 1.2115486242691986e-05 }, { "epoch": 0.3264643842284602, "step": 1006, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.915904769906774e-05, "weight_min": 1.0888551171817085e-09, "weight_std": 1.773660915205255e-05 }, { "epoch": 0.3264643842284602, "step": 1006, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0035725608759094e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.858303221524693e-05 }, { "epoch": 0.3264643842284602, "step": 1006, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077446985524148e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357483451021835e-05 }, { "epoch": 0.3264643842284602, "step": 1006, "weight_max": 4.252122016623616e-05, "weight_mean": 5.31605519427103e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5033156159915961e-05 }, { "epoch": 0.32678890150900536, "step": 1007, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9730270651052706e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1294874386512674e-05 }, { "epoch": 0.32678890150900536, "step": 1007, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805788648809539e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3588913134299219e-05 }, { "epoch": 0.32678890150900536, "step": 1007, "weight_max": 4.252122016623616e-05, "weight_mean": 3.115277650067583e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.4396753613254987e-05 }, { "epoch": 0.32678890150900536, "step": 1007, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3903613762522582e-05, "weight_min": 0.0, "weight_std": 1.927578341565095e-05 }, { "epoch": 0.3271134187895505, "step": 1008, "weight_max": 4.187939703115262e-05, "weight_mean": 8.663811968290247e-06, "weight_min": 0.0, "weight_std": 1.6497853721375577e-05 }, { "epoch": 0.3271134187895505, "step": 1008, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8163977074436843e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9738805349334143e-05 }, { "epoch": 0.3271134187895505, "step": 1008, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8632270439411514e-05, "weight_min": 0.0, "weight_std": 2.000773929466959e-05 }, { "epoch": 0.3271134187895505, "step": 1008, "weight_max": 3.967146767536178e-05, "weight_mean": 4.960477326676482e-06, "weight_min": 0.0, "weight_std": 1.4025358723301906e-05 }, { "epoch": 0.32743793607009575, "step": 1009, "weight_max": 3.813097282545641e-05, "weight_mean": 8.823075404507108e-06, "weight_min": 0.0, "weight_std": 1.612694359209854e-05 }, { "epoch": 0.32743793607009575, "step": 1009, "weight_max": 3.662098970380612e-05, "weight_mean": 1.2678201528615318e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.766664172464516e-05 }, { "epoch": 0.32743793607009575, "step": 1009, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8098089640261605e-05, "weight_min": 0.0, "weight_std": 1.821506884880364e-05 }, { "epoch": 0.32743793607009575, "step": 1009, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9498773326631635e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0966183001291938e-05 }, { "epoch": 0.3277624533506409, "grad_norm": 15.795165061950684, "learning_rate": 8.151020408163266e-06, "loss": 0.1961, "step": 1010 }, { "epoch": 0.3277624533506409, "step": 1010, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392021977168042e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9247658201493323e-05 }, { "epoch": 0.3277624533506409, "step": 1010, "weight_max": 3.671447146302853e-09, "weight_mean": 1.4995407138229666e-09, "weight_min": 6.390288853630466e-10, "weight_std": 1.0134914019133134e-09 }, { "epoch": 0.3277624533506409, "step": 1010, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4926260519132484e-05, "weight_min": 0.0, "weight_std": 2.0642057279474102e-05 }, { "epoch": 0.3277624533506409, "step": 1010, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.835338121571112e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6599400623817928e-05 }, { "epoch": 0.32808697063118614, "step": 1011, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9615345081547275e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1006846509408206e-05 }, { "epoch": 0.32808697063118614, "step": 1011, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541869202512316e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2841374882555101e-05 }, { "epoch": 0.32808697063118614, "step": 1011, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.0551993707777e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.681951471255161e-05 }, { "epoch": 0.32808697063118614, "step": 1011, "weight_max": 8.490471659072796e-10, "weight_mean": 5.911802158919954e-10, "weight_min": 0.0, "weight_std": 2.507978824617396e-10 }, { "epoch": 0.3284114879117313, "step": 1012, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.465713376092026e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.02794235519832e-05 }, { "epoch": 0.3284114879117313, "step": 1012, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.5097111827344634e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8370265024714172e-05 }, { "epoch": 0.3284114879117313, "step": 1012, "weight_max": 1.7212929970256141e-09, "weight_mean": 9.495237929257883e-10, "weight_min": 0.0, "weight_std": 6.463627411079642e-10 }, { "epoch": 0.3284114879117313, "step": 1012, "weight_max": 3.813097282545641e-05, "weight_mean": 7.049939540593186e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.409036667610053e-05 }, { "epoch": 0.32873600519227647, "step": 1013, "weight_max": 4.187939703115262e-05, "weight_mean": 5.381429673434468e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4753059986105654e-05 }, { "epoch": 0.32873600519227647, "step": 1013, "weight_max": 8.155858211011946e-09, "weight_mean": 1.5758456761716388e-09, "weight_min": 5.771870759119224e-10, "weight_std": 2.6588637958013805e-09 }, { "epoch": 0.32873600519227647, "step": 1013, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.158886324352352e-06, "weight_min": 0.0, "weight_std": 1.4579968592443038e-05 }, { "epoch": 0.32873600519227647, "step": 1013, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4638964785262942e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0336676243459806e-05 }, { "epoch": 0.3290605224728217, "step": 1014, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.383524613629561e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9194038031855598e-05 }, { "epoch": 0.3290605224728217, "step": 1014, "weight_max": 1.4648394426330924e-06, "weight_mean": 1.84077919129777e-07, "weight_min": 0.0, "weight_std": 5.175065780349541e-07 }, { "epoch": 0.3290605224728217, "step": 1014, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3858370948582888e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.917901863635052e-05 }, { "epoch": 0.3290605224728217, "step": 1014, "weight_max": 4.187939703115262e-05, "weight_mean": 1.570513632032089e-05, "weight_min": 0.0, "weight_std": 2.16743464989122e-05 }, { "epoch": 0.32938503975336686, "step": 1015, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.115413387102308e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4342401300382335e-05 }, { "epoch": 0.32938503975336686, "step": 1015, "weight_max": 3.998322426923551e-05, "weight_mean": 9.841396604315378e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8224367522634566e-05 }, { "epoch": 0.32938503975336686, "step": 1015, "weight_max": 3.998322426923551e-05, "weight_mean": 1.804675775929354e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.957751010195352e-05 }, { "epoch": 0.32938503975336686, "step": 1015, "weight_max": 3.4268457966391e-05, "weight_mean": 8.38879441289464e-06, "weight_min": 0.0, "weight_std": 1.553646507090889e-05 }, { "epoch": 0.3297095570339121, "step": 1016, "weight_max": 4.187939703115262e-05, "weight_mean": 1.414192229276523e-05, "weight_min": 0.0, "weight_std": 1.9657554730656557e-05 }, { "epoch": 0.3297095570339121, "step": 1016, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5378707757918164e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9353941752342507e-05 }, { "epoch": 0.3297095570339121, "step": 1016, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.188719559460878e-06, "weight_min": 0.0, "weight_std": 1.7121694327215664e-05 }, { "epoch": 0.3297095570339121, "step": 1016, "weight_max": 3.813097282545641e-05, "weight_mean": 1.1313507457089145e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7148298866231926e-05 }, { "epoch": 0.33003407431445725, "step": 1017, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.52170194068458e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6411788237746805e-05 }, { "epoch": 0.33003407431445725, "step": 1017, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998872100259177e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.413581230735872e-05 }, { "epoch": 0.33003407431445725, "step": 1017, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.49002514971653e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.681636058492586e-05 }, { "epoch": 0.33003407431445725, "step": 1017, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.445110137865413e-05, "weight_min": 0.0, "weight_std": 2.0100598703720607e-05 }, { "epoch": 0.3303585915950024, "step": 1018, "weight_max": 4.187939703115262e-05, "weight_mean": 1.373810664517805e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.9277125829830766e-05 }, { "epoch": 0.3303585915950024, "step": 1018, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.962316653400194e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8455868485034443e-05 }, { "epoch": 0.3303585915950024, "step": 1018, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1577385420387145e-06, "weight_min": 0.0, "weight_std": 1.4580429706256837e-05 }, { "epoch": 0.3303585915950024, "step": 1018, "weight_max": 3.967146767536178e-05, "weight_mean": 9.764084097696468e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8081171219819225e-05 }, { "epoch": 0.33068310887554764, "step": 1019, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4561474017682485e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0386916730785742e-05 }, { "epoch": 0.33068310887554764, "step": 1019, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9157181668560952e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0700823370134458e-05 }, { "epoch": 0.33068310887554764, "step": 1019, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235740445641568e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4806272702116985e-05 }, { "epoch": 0.33068310887554764, "step": 1019, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396597771323286e-06, "weight_min": 0.0, "weight_std": 1.5261919543263502e-05 }, { "epoch": 0.3310076261560928, "grad_norm": 0.08098495006561279, "learning_rate": 8.13061224489796e-06, "loss": 0.1253, "step": 1020 }, { "epoch": 0.3310076261560928, "step": 1020, "weight_max": 3.36925259034615e-05, "weight_mean": 1.1934042049688287e-05, "weight_min": 0.0, "weight_std": 1.6495112504344434e-05 }, { "epoch": 0.3310076261560928, "step": 1020, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2114192941226065e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.764884109434206e-05 }, { "epoch": 0.3310076261560928, "step": 1020, "weight_max": 3.662098970380612e-05, "weight_mean": 7.690135134907905e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4578454283764586e-05 }, { "epoch": 0.3310076261560928, "step": 1020, "weight_max": 3.998322426923551e-05, "weight_mean": 9.428566045244224e-06, "weight_min": 7.508524357824342e-10, "weight_std": 1.7497393855592236e-05 }, { "epoch": 0.331332143436638, "step": 1021, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1267795343883336e-05, "weight_min": 0.0, "weight_std": 1.7659154764260165e-05 }, { "epoch": 0.331332143436638, "step": 1021, "weight_max": 4.187939703115262e-05, "weight_mean": 1.295365291298367e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8571994587546214e-05 }, { "epoch": 0.331332143436638, "step": 1021, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156531187822111e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4580917195416987e-05 }, { "epoch": 0.331332143436638, "step": 1021, "weight_max": 2.639517333591357e-05, "weight_mean": 3.3004325814545155e-06, "weight_min": 6.100955851628953e-10, "weight_std": 9.331684850621969e-06 }, { "epoch": 0.3316566607171832, "step": 1022, "weight_max": 3.602553260861896e-05, "weight_mean": 4.503892796492437e-06, "weight_min": 0.0, "weight_std": 1.2736665667034686e-05 }, { "epoch": 0.3316566607171832, "step": 1022, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6867030353751034e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8797683878801763e-05 }, { "epoch": 0.3316566607171832, "step": 1022, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235800800728612e-05, "weight_min": 0.0, "weight_std": 1.8970516975969076e-05 }, { "epoch": 0.3316566607171832, "step": 1022, "weight_max": 4.187939703115262e-05, "weight_mean": 8.40966822579503e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6180865713977255e-05 }, { "epoch": 0.33198117799772836, "step": 1023, "weight_max": 4.252122016623616e-05, "weight_mean": 9.075218258658424e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7126665625255555e-05 }, { "epoch": 0.33198117799772836, "step": 1023, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0127855603059288e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8724274923442863e-05 }, { "epoch": 0.33198117799772836, "step": 1023, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0086912880069576e-05, "weight_min": 0.0, "weight_std": 1.8736909623839892e-05 }, { "epoch": 0.33198117799772836, "step": 1023, "weight_max": 4.252122016623616e-05, "weight_mean": 1.433190846000798e-05, "weight_min": 0.0, "weight_std": 2.007029252126813e-05 }, { "epoch": 0.3323056952782736, "step": 1024, "weight_max": 1.4129907022208954e-08, "weight_mean": 2.355668327069793e-09, "weight_min": 0.0, "weight_std": 4.7675339232000624e-09 }, { "epoch": 0.3323056952782736, "step": 1024, "weight_max": 3.1715208024252206e-05, "weight_mean": 7.457626907125814e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3843848137184978e-05 }, { "epoch": 0.3323056952782736, "step": 1024, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3056349416729063e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.841238554334268e-05 }, { "epoch": 0.3323056952782736, "step": 1024, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.07696586282691e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357677173393313e-05 }, { "epoch": 0.33263021255881875, "step": 1025, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691005869972287e-06, "weight_min": 0.0, "weight_std": 1.3266258974908851e-05 }, { "epoch": 0.33263021255881875, "step": 1025, "weight_max": 4.187939703115262e-05, "weight_mean": 1.199831604026258e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7195570762851276e-05 }, { "epoch": 0.33263021255881875, "step": 1025, "weight_max": 3.9051607018336654e-05, "weight_mean": 2.0570834749378264e-05, "weight_min": 1.0385104998178463e-09, "weight_std": 1.8192204152001068e-05 }, { "epoch": 0.33263021255881875, "step": 1025, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.451569187338464e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.013820267166011e-05 }, { "epoch": 0.33295472983936397, "step": 1026, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390838724561036e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.923935087688733e-05 }, { "epoch": 0.33295472983936397, "step": 1026, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9176000932930037e-05, "weight_min": 0.0, "weight_std": 2.0643850803025998e-05 }, { "epoch": 0.33295472983936397, "step": 1026, "weight_max": 4.252122016623616e-05, "weight_mean": 1.504224019299727e-05, "weight_min": 0.0, "weight_std": 2.0796314856852405e-05 }, { "epoch": 0.33295472983936397, "step": 1026, "weight_max": 2.109268537608955e-09, "weight_mean": 1.1152296863770061e-09, "weight_min": 5.864964625068581e-10, "weight_std": 6.047556344590532e-10 }, { "epoch": 0.33327924711990914, "step": 1027, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316770511853974e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.50328678500955e-05 }, { "epoch": 0.33327924711990914, "step": 1027, "weight_max": 3.782653220696375e-05, "weight_mean": 1.611038351256866e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7410948203178123e-05 }, { "epoch": 0.33327924711990914, "step": 1027, "weight_max": 4.187939703115262e-05, "weight_mean": 1.562061152071692e-05, "weight_min": 0.0, "weight_std": 1.8625843949848786e-05 }, { "epoch": 0.33327924711990914, "step": 1027, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.257528290618211e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7451176972826943e-05 }, { "epoch": 0.3336037644004543, "step": 1028, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3085445971228182e-05, "weight_min": 0.0, "weight_std": 1.828605127229821e-05 }, { "epoch": 0.3336037644004543, "step": 1028, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.801514923106879e-06, "weight_min": 0.0, "weight_std": 1.65502424351871e-05 }, { "epoch": 0.3336037644004543, "step": 1028, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6082858564914204e-05, "weight_min": 0.0, "weight_std": 1.8666938558453694e-05 }, { "epoch": 0.3336037644004543, "step": 1028, "weight_max": 4.252122016623616e-05, "weight_mean": 2.046600275207311e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1890193238505162e-05 }, { "epoch": 0.33392828168099953, "step": 1029, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155820417712675e-06, "weight_min": 0.0, "weight_std": 1.4581203686248045e-05 }, { "epoch": 0.33392828168099953, "step": 1029, "weight_max": 3.143761568935588e-05, "weight_mean": 3.9320348150795326e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1113934306195006e-05 }, { "epoch": 0.33392828168099953, "step": 1029, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076896741229575e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357705367729068e-05 }, { "epoch": 0.33392828168099953, "step": 1029, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.439763744012453e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.7575935999047942e-05 }, { "epoch": 0.3342527989615447, "grad_norm": 0.4542214870452881, "learning_rate": 8.110204081632653e-06, "loss": 0.0508, "step": 1030 }, { "epoch": 0.3342527989615447, "step": 1030, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.879453338915482e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0224830223014578e-05 }, { "epoch": 0.3342527989615447, "step": 1030, "weight_max": 4.187939703115262e-05, "weight_mean": 9.06199602468405e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.704628266452346e-05 }, { "epoch": 0.3342527989615447, "step": 1030, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.654411769588478e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7907063011080027e-05 }, { "epoch": 0.3342527989615447, "step": 1030, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805285698239459e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3589115951617714e-05 }, { "epoch": 0.3345773162420899, "step": 1031, "weight_max": 3.1715208024252206e-05, "weight_mean": 7.36004403734114e-06, "weight_min": 0.0, "weight_std": 1.3681427844858263e-05 }, { "epoch": 0.3345773162420899, "step": 1031, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235726222163066e-05, "weight_min": 0.0, "weight_std": 1.897056426969357e-05 }, { "epoch": 0.3345773162420899, "step": 1031, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0171924259338994e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7775024389266036e-05 }, { "epoch": 0.3345773162420899, "step": 1031, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320469088270329e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2217910807521548e-05 }, { "epoch": 0.3349018335226351, "step": 1032, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0741939149738755e-05, "weight_min": 0.0, "weight_std": 1.8387709133094177e-05 }, { "epoch": 0.3349018335226351, "step": 1032, "weight_max": 2.5890016331686638e-05, "weight_mean": 3.2369794098485727e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.153209248324856e-06 }, { "epoch": 0.3349018335226351, "step": 1032, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.28913323150482e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6632009646855295e-05 }, { "epoch": 0.3349018335226351, "step": 1032, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4096322047407739e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9544537281035446e-05 }, { "epoch": 0.33522635080318025, "step": 1033, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.880962923285551e-06, "weight_min": 0.0, "weight_std": 1.6724106899346225e-05 }, { "epoch": 0.33522635080318025, "step": 1033, "weight_max": 2.538973967602942e-05, "weight_mean": 3.3323333354928764e-06, "weight_min": 6.390288853630466e-10, "weight_std": 8.92347361514112e-06 }, { "epoch": 0.33522635080318025, "step": 1033, "weight_max": 1.3301527657105794e-09, "weight_mean": 5.92246141017938e-10, "weight_min": 0.0, "weight_std": 4.3827866336165755e-10 }, { "epoch": 0.33522635080318025, "step": 1033, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074706551677082e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8654243831406347e-05 }, { "epoch": 0.3355508680837255, "step": 1034, "weight_max": 4.187939703115262e-05, "weight_mean": 8.835864719003439e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.6646115909679793e-05 }, { "epoch": 0.3355508680837255, "step": 1034, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0116931662196293e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8746784917311743e-05 }, { "epoch": 0.3355508680837255, "step": 1034, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2579156646097545e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7677761206869036e-05 }, { "epoch": 0.3355508680837255, "step": 1034, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3118623428454157e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8314807675778866e-05 }, { "epoch": 0.33587538536427064, "step": 1035, "weight_max": 1.0261104188558079e-09, "weight_mean": 8.0245643463428e-10, "weight_min": 5.958802340444436e-10, "weight_std": 1.7650650385725442e-10 }, { "epoch": 0.33587538536427064, "step": 1035, "weight_max": 4.187939703115262e-05, "weight_mean": 9.664815479482058e-06, "weight_min": 0.0, "weight_std": 1.7977174138650298e-05 }, { "epoch": 0.33587538536427064, "step": 1035, "weight_max": 4.187939703115262e-05, "weight_mean": 1.390523357258644e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.943929419212509e-05 }, { "epoch": 0.33587538536427064, "step": 1035, "weight_max": 3.6172561976854922e-06, "weight_mean": 4.544178011656186e-07, "weight_min": 5.958802340444436e-10, "weight_std": 1.2779821645381162e-06 }, { "epoch": 0.33619990264481586, "step": 1036, "weight_max": 7.790310519339982e-06, "weight_mean": 1.8402272417006316e-06, "weight_min": 5.864964625068581e-10, "weight_std": 3.411784973650356e-06 }, { "epoch": 0.33619990264481586, "step": 1036, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.4300866496341769e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.97372446564259e-05 }, { "epoch": 0.33619990264481586, "step": 1036, "weight_max": 4.252122016623616e-05, "weight_mean": 1.865631747932639e-05, "weight_min": 7.721521200210191e-10, "weight_std": 2.0283479898353107e-05 }, { "epoch": 0.33619990264481586, "step": 1036, "weight_max": 4.187939703115262e-05, "weight_mean": 9.411329301656224e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7571428543305956e-05 }, { "epoch": 0.33652441992536103, "step": 1037, "weight_max": 3.632264997577295e-05, "weight_mean": 4.540861937130103e-06, "weight_min": 0.0, "weight_std": 1.2841781426686794e-05 }, { "epoch": 0.33652441992536103, "step": 1037, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156368388270494e-06, "weight_min": 0.0, "weight_std": 1.4580982679035515e-05 }, { "epoch": 0.33652441992536103, "step": 1037, "weight_max": 3.632264997577295e-05, "weight_mean": 4.54095788882114e-06, "weight_min": 0.0, "weight_std": 1.2841743227909319e-05 }, { "epoch": 0.33652441992536103, "step": 1037, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4808229025220498e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.047410816885531e-05 }, { "epoch": 0.3368489372059062, "step": 1038, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9785253243753687e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.091614260280039e-05 }, { "epoch": 0.3368489372059062, "step": 1038, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8998012819793075e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9792343664448708e-05 }, { "epoch": 0.3368489372059062, "step": 1038, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8066883058054373e-05, "weight_min": 9.413956281179026e-10, "weight_std": 1.89692473213654e-05 }, { "epoch": 0.3368489372059062, "step": 1038, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5391204215120524e-05, "weight_min": 0.0, "weight_std": 2.1257483240333386e-05 }, { "epoch": 0.3371734544864514, "step": 1039, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.500936019117944e-06, "weight_min": 0.0, "weight_std": 1.5942423488013446e-05 }, { "epoch": 0.3371734544864514, "step": 1039, "weight_max": 8.603301404619401e-10, "weight_mean": 6.366969729221239e-10, "weight_min": 0.0, "weight_std": 2.6993476898162783e-10 }, { "epoch": 0.3371734544864514, "step": 1039, "weight_max": 3.397988257347606e-05, "weight_mean": 1.0764440048660617e-05, "weight_min": 0.0, "weight_std": 1.5227200492518023e-05 }, { "epoch": 0.3371734544864514, "step": 1039, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.72045256983256e-06, "weight_min": 0.0, "weight_std": 1.6182337276404724e-05 }, { "epoch": 0.3374979717669966, "grad_norm": 3.5498340129852295, "learning_rate": 8.089795918367348e-06, "loss": 0.1326, "step": 1040 }, { "epoch": 0.3374979717669966, "step": 1040, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4094252037466504e-05, "weight_min": 0.0, "weight_std": 1.95529773918679e-05 }, { "epoch": 0.3374979717669966, "step": 1040, "weight_max": 4.187939703115262e-05, "weight_mean": 1.430280099157244e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9913490177714266e-05 }, { "epoch": 0.3374979717669966, "step": 1040, "weight_max": 3.998322426923551e-05, "weight_mean": 9.879921890387777e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8294353139935993e-05 }, { "epoch": 0.3374979717669966, "step": 1040, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0541602932789829e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8238346456200816e-05 }, { "epoch": 0.3378224890475418, "step": 1041, "weight_max": 5.163849436939927e-06, "weight_mean": 6.466440822805453e-07, "weight_min": 6.100955851628953e-10, "weight_std": 1.8252270592711284e-06 }, { "epoch": 0.3378224890475418, "step": 1041, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156049610377522e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4581111827283166e-05 }, { "epoch": 0.3378224890475418, "step": 1041, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4032037142897025e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.967531534319278e-05 }, { "epoch": 0.3378224890475418, "step": 1041, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767875907418784e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3480726920533925e-05 }, { "epoch": 0.338147006328087, "step": 1042, "weight_max": 3.006794759130571e-05, "weight_mean": 3.759661012736615e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0630153155943844e-05 }, { "epoch": 0.338147006328087, "step": 1042, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729837201011833e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.337308367510559e-05 }, { "epoch": 0.338147006328087, "step": 1042, "weight_max": 1.56184469233267e-05, "weight_mean": 1.9531435100361705e-06, "weight_min": 5.958802340444436e-10, "weight_std": 5.521616458281642e-06 }, { "epoch": 0.338147006328087, "step": 1042, "weight_max": 4.252122016623616e-05, "weight_mean": 8.011505997274071e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5854484445299022e-05 }, { "epoch": 0.33847152360863214, "step": 1043, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074784768221434e-05, "weight_min": 0.0, "weight_std": 1.8654194718692452e-05 }, { "epoch": 0.33847152360863214, "step": 1043, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2676864318782464e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.914395579660777e-05 }, { "epoch": 0.33847152360863214, "step": 1043, "weight_max": 4.252122016623616e-05, "weight_mean": 8.647110007586889e-06, "weight_min": 0.0, "weight_std": 1.6561894881306216e-05 }, { "epoch": 0.33847152360863214, "step": 1043, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.630230326554738e-05, "weight_min": 9.178617865757133e-10, "weight_std": 1.8617087334860116e-05 }, { "epoch": 0.33879604088917736, "step": 1044, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2627742509939708e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.921264993143268e-05 }, { "epoch": 0.33879604088917736, "step": 1044, "weight_max": 3.397988257347606e-05, "weight_mean": 4.3003078644687776e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1993036423518788e-05 }, { "epoch": 0.33879604088917736, "step": 1044, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.2296470231376588e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.703955786069855e-05 }, { "epoch": 0.33879604088917736, "step": 1044, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077863988844911e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4357314285007305e-05 }, { "epoch": 0.33912055816972253, "step": 1045, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2418509868439287e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.719852298265323e-05 }, { "epoch": 0.33912055816972253, "step": 1045, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.439519999432378e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7576086975168437e-05 }, { "epoch": 0.33912055816972253, "step": 1045, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.512176802672911e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.769406662788242e-05 }, { "epoch": 0.33912055816972253, "step": 1045, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2722874998871703e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.821331534301862e-05 }, { "epoch": 0.33944507545026775, "step": 1046, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392004696768709e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.924766729644034e-05 }, { "epoch": 0.33944507545026775, "step": 1046, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9787803466897458e-05, "weight_min": 0.0, "weight_std": 2.1275449398672208e-05 }, { "epoch": 0.33944507545026775, "step": 1046, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5470985090360045e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.136890179826878e-05 }, { "epoch": 0.33944507545026775, "step": 1046, "weight_max": 1.0015337448265882e-09, "weight_mean": 6.452591794214868e-10, "weight_min": 0.0, "weight_std": 2.9498337106304007e-10 }, { "epoch": 0.3397695927308129, "step": 1047, "weight_max": 2.4160392058547586e-05, "weight_mean": 3.0206892915884964e-06, "weight_min": 6.100955851628953e-10, "weight_std": 8.541729584976565e-06 }, { "epoch": 0.3397695927308129, "step": 1047, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076984507468296e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357669897435699e-05 }, { "epoch": 0.3397695927308129, "step": 1047, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.746967978368048e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.653657636779826e-05 }, { "epoch": 0.3397695927308129, "step": 1047, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.020828656503e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7126125385402702e-05 }, { "epoch": 0.3400941100113581, "step": 1048, "weight_max": 6.100955829424493e-08, "weight_mean": 8.50483861114526e-09, "weight_min": 5.864964625068581e-10, "weight_std": 2.122907183377265e-08 }, { "epoch": 0.3400941100113581, "step": 1048, "weight_max": 4.187939703115262e-05, "weight_mean": 1.326380606769817e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.866864477051422e-05 }, { "epoch": 0.3400941100113581, "step": 1048, "weight_max": 3.813097282545641e-05, "weight_mean": 8.491553671774454e-06, "weight_min": 0.0, "weight_std": 1.5879100828897208e-05 }, { "epoch": 0.3400941100113581, "step": 1048, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7189040590892546e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.86098968697479e-05 }, { "epoch": 0.3404186272919033, "step": 1049, "weight_max": 3.998322426923551e-05, "weight_mean": 9.613618203729857e-06, "weight_min": 0.0, "weight_std": 1.781817445589695e-05 }, { "epoch": 0.3404186272919033, "step": 1049, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.506136848358437e-06, "weight_min": 0.0, "weight_std": 1.7655062038102187e-05 }, { "epoch": 0.3404186272919033, "step": 1049, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.008393135445658e-06, "weight_min": 0.0, "weight_std": 1.668550612521358e-05 }, { "epoch": 0.3404186272919033, "step": 1049, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.802096428757068e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8148835806641728e-05 }, { "epoch": 0.3407431445724485, "grad_norm": 8.043953895568848, "learning_rate": 8.069387755102041e-06, "loss": 0.1878, "step": 1050 }, { "epoch": 0.3407431445724485, "step": 1050, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8575561625766568e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.031429175985977e-05 }, { "epoch": 0.3407431445724485, "step": 1050, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3588351066573523e-05, "weight_min": 0.0, "weight_std": 1.893942680908367e-05 }, { "epoch": 0.3407431445724485, "step": 1050, "weight_max": 3.722132169059478e-05, "weight_mean": 4.6536974878108595e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3159308764443267e-05 }, { "epoch": 0.3407431445724485, "step": 1050, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5416963985189795e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2841444004152436e-05 }, { "epoch": 0.3410676618529937, "step": 1051, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4530826774716843e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9698707546922378e-05 }, { "epoch": 0.3410676618529937, "step": 1051, "weight_max": 3.632264997577295e-05, "weight_mean": 8.100241757347248e-06, "weight_min": 0.0, "weight_std": 1.514057021267945e-05 }, { "epoch": 0.3410676618529937, "step": 1051, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4512180971214548e-05, "weight_min": 7.298507354036587e-10, "weight_std": 2.04156222025631e-05 }, { "epoch": 0.3410676618529937, "step": 1051, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076949491922278e-06, "weight_min": 0.0, "weight_std": 1.4357683539856225e-05 }, { "epoch": 0.34139217913353886, "step": 1052, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.716554814076517e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8136195649276488e-05 }, { "epoch": 0.34139217913353886, "step": 1052, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3886231499782298e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9010843971045688e-05 }, { "epoch": 0.34139217913353886, "step": 1052, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156113275006646e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4581086361431517e-05 }, { "epoch": 0.34139217913353886, "step": 1052, "weight_max": 4.252122016623616e-05, "weight_mean": 5.330190560925985e-06, "weight_min": 0.0, "weight_std": 1.5027500921860337e-05 }, { "epoch": 0.34171669641408403, "step": 1053, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430592071003048e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.2527656508609653e-05 }, { "epoch": 0.34171669641408403, "step": 1053, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0769231165759265e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357695363287348e-05 }, { "epoch": 0.34171669641408403, "step": 1053, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.289993613492697e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7295578800258227e-05 }, { "epoch": 0.34171669641408403, "step": 1053, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2812641216441989e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.737780803523492e-05 }, { "epoch": 0.34204121369462925, "step": 1054, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274558007949963e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.90388600458391e-05 }, { "epoch": 0.34204121369462925, "step": 1054, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0197930350841489e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.8902381270891055e-05 }, { "epoch": 0.34204121369462925, "step": 1054, "weight_max": 3.602553260861896e-05, "weight_mean": 6.787806341890246e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3390097592491657e-05 }, { "epoch": 0.34204121369462925, "step": 1054, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2410080671543255e-05, "weight_min": 0.0, "weight_std": 1.740078369039111e-05 }, { "epoch": 0.3423657309751744, "step": 1055, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315738690114813e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5033284398668911e-05 }, { "epoch": 0.3423657309751744, "step": 1055, "weight_max": 3.782653220696375e-05, "weight_mean": 4.7292619456129614e-06, "weight_min": 0.0, "weight_std": 1.3373316505749244e-05 }, { "epoch": 0.3423657309751744, "step": 1055, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320729658502387e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2217806215630844e-05 }, { "epoch": 0.3423657309751744, "step": 1055, "weight_max": 4.252122016623616e-05, "weight_mean": 9.635547030484304e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7966543964575976e-05 }, { "epoch": 0.34269024825571964, "step": 1056, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0784861842694227e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6510948626091704e-05 }, { "epoch": 0.34269024825571964, "step": 1056, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5309458831325173e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.113055597874336e-05 }, { "epoch": 0.34269024825571964, "step": 1056, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5435905879712664e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1342750187614e-05 }, { "epoch": 0.34269024825571964, "step": 1056, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.570089448767249e-06, "weight_min": 0.0, "weight_std": 1.587944643688388e-05 }, { "epoch": 0.3430147655362648, "step": 1057, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.1190450095455162e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.5793151760590263e-05 }, { "epoch": 0.3430147655362648, "step": 1057, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550660590524785e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9535338651621714e-05 }, { "epoch": 0.3430147655362648, "step": 1057, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4054068742552772e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9479570255498402e-05 }, { "epoch": 0.3430147655362648, "step": 1057, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959520538250217e-06, "weight_min": 0.0, "weight_std": 1.4025744349055458e-05 }, { "epoch": 0.34333928281681, "step": 1058, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3632495210913476e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8809520042850636e-05 }, { "epoch": 0.34333928281681, "step": 1058, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.941448423778638e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.3783579561277293e-05 }, { "epoch": 0.34333928281681, "step": 1058, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0636118531692773e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9684433937072754e-05 }, { "epoch": 0.34333928281681, "step": 1058, "weight_max": 3.662098970380612e-05, "weight_mean": 8.103554137051105e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5170883671089541e-05 }, { "epoch": 0.3436638000973552, "step": 1059, "weight_max": 3.998322426923551e-05, "weight_mean": 9.976583896786906e-06, "weight_min": 0.0, "weight_std": 1.832991256378591e-05 }, { "epoch": 0.3436638000973552, "step": 1059, "weight_max": 3.752332122530788e-05, "weight_mean": 8.889819582691416e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5548394003417343e-05 }, { "epoch": 0.3436638000973552, "step": 1059, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4492215086647775e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.032278644037433e-05 }, { "epoch": 0.3436638000973552, "step": 1059, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1542999345692806e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.834633258113172e-05 }, { "epoch": 0.34398831737790037, "grad_norm": 0.5212376117706299, "learning_rate": 8.048979591836736e-06, "loss": 0.098, "step": 1060 }, { "epoch": 0.34398831737790037, "step": 1060, "weight_max": 4.252122016623616e-05, "weight_mean": 1.554752998345066e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.146223960153293e-05 }, { "epoch": 0.34398831737790037, "step": 1060, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.441384463571012e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.7527781892567873e-05 }, { "epoch": 0.34398831737790037, "step": 1060, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315549969964195e-06, "weight_min": 0.0, "weight_std": 1.5033361705718562e-05 }, { "epoch": 0.34398831737790037, "step": 1060, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.846879038377665e-06, "weight_min": 0.0, "weight_std": 1.8257100236951374e-05 }, { "epoch": 0.3443128346584456, "step": 1061, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3348448192118667e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9092316506430507e-05 }, { "epoch": 0.3443128346584456, "step": 1061, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5237359548336826e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1062996893306263e-05 }, { "epoch": 0.3443128346584456, "step": 1061, "weight_max": 1.2874567190124253e-08, "weight_mean": 2.173409896499834e-09, "weight_min": 0.0, "weight_std": 4.333431391501108e-09 }, { "epoch": 0.3443128346584456, "step": 1061, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1197390449524391e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.7132440916611813e-05 }, { "epoch": 0.34463735193899075, "step": 1062, "weight_max": 3.662098970380612e-05, "weight_mean": 9.08135643840069e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.681482171989046e-05 }, { "epoch": 0.34463735193899075, "step": 1062, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3507949663326144e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9936589524149895e-05 }, { "epoch": 0.34463735193899075, "step": 1062, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5865680325077847e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.189661972806789e-05 }, { "epoch": 0.34463735193899075, "step": 1062, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390773240942508e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9239390894654207e-05 }, { "epoch": 0.3449618692195359, "step": 1063, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236232027527876e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806074432271998e-05 }, { "epoch": 0.3449618692195359, "step": 1063, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.476439117861446e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.76343437487958e-05 }, { "epoch": 0.3449618692195359, "step": 1063, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.405362000805326e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4504321370623074e-05 }, { "epoch": 0.3449618692195359, "step": 1063, "weight_max": 4.3167921830900013e-05, "weight_mean": 6.378959369612858e-06, "weight_min": 0.0, "weight_std": 1.5117211660253815e-05 }, { "epoch": 0.34528638650008114, "step": 1064, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.168987627257593e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4322809875011444e-05 }, { "epoch": 0.34528638650008114, "step": 1064, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.562841862323694e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1579206077149138e-05 }, { "epoch": 0.34528638650008114, "step": 1064, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2753767805406824e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8088727301801555e-05 }, { "epoch": 0.34528638650008114, "step": 1064, "weight_max": 1.2064879095774472e-09, "weight_mean": 7.424764691954522e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.0078645690535524e-10 }, { "epoch": 0.3456109037806263, "step": 1065, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3632586160383653e-05, "weight_min": 0.0, "weight_std": 1.886252357508056e-05 }, { "epoch": 0.3456109037806263, "step": 1065, "weight_max": 2.690521432668902e-05, "weight_mean": 3.36397897626739e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.512095857644454e-06 }, { "epoch": 0.3456109037806263, "step": 1065, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.868897015810944e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0143719666521065e-05 }, { "epoch": 0.3456109037806263, "step": 1065, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.843293810263276e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8236745745525695e-05 }, { "epoch": 0.34593542106117153, "step": 1066, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.1511951925058383e-05, "weight_min": 0.0, "weight_std": 1.61605239554774e-05 }, { "epoch": 0.34593542106117153, "step": 1066, "weight_max": 3.632264997577295e-05, "weight_mean": 1.216281179949874e-05, "weight_min": 0.0, "weight_std": 1.6894256987143308e-05 }, { "epoch": 0.34593542106117153, "step": 1066, "weight_max": 3.967146767536178e-05, "weight_mean": 1.1111980711575598e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.791994145605713e-05 }, { "epoch": 0.34593542106117153, "step": 1066, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156147381057963e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.458107271901099e-05 }, { "epoch": 0.3462599383417167, "step": 1067, "weight_max": 4.252122016623616e-05, "weight_mean": 7.705040843575262e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.557675022922922e-05 }, { "epoch": 0.3462599383417167, "step": 1067, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9521668946254067e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1012823708588257e-05 }, { "epoch": 0.3462599383417167, "step": 1067, "weight_max": 2.343743108212948e-05, "weight_mean": 2.930373511844664e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.286103366117459e-06 }, { "epoch": 0.3462599383417167, "step": 1067, "weight_max": 4.381950566312298e-05, "weight_mean": 5.478000730363419e-06, "weight_min": 0.0, "weight_std": 1.549230728414841e-05 }, { "epoch": 0.34658445562226187, "step": 1068, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.1299915058771148e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.6169467926374637e-05 }, { "epoch": 0.34658445562226187, "step": 1068, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920926130580483e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3915861927671358e-05 }, { "epoch": 0.34658445562226187, "step": 1068, "weight_max": 3.662098970380612e-05, "weight_mean": 6.539975856867386e-06, "weight_min": 0.0, "weight_std": 1.3324885003385134e-05 }, { "epoch": 0.34658445562226187, "step": 1068, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.66968946158886e-06, "weight_min": 0.0, "weight_std": 1.691682882665191e-05 }, { "epoch": 0.3469089729028071, "step": 1069, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653588348446647e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.315935151069425e-05 }, { "epoch": 0.3469089729028071, "step": 1069, "weight_max": 3.602553260861896e-05, "weight_mean": 4.5040187615086325e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2736614735331386e-05 }, { "epoch": 0.3469089729028071, "step": 1069, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0550414774334058e-05, "weight_min": 0.0, "weight_std": 2.2003447156748734e-05 }, { "epoch": 0.3469089729028071, "step": 1069, "weight_max": 3.967146767536178e-05, "weight_mean": 1.711747063382063e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8682345398701727e-05 }, { "epoch": 0.34723349018335226, "grad_norm": 15.456985473632812, "learning_rate": 8.02857142857143e-06, "loss": 0.2725, "step": 1070 }, { "epoch": 0.34723349018335226, "step": 1070, "weight_max": 3.722132169059478e-05, "weight_mean": 7.321944849536521e-06, "weight_min": 0.0, "weight_std": 1.4201096746546682e-05 }, { "epoch": 0.34723349018335226, "step": 1070, "weight_max": 3.662098970380612e-05, "weight_mean": 9.044761100085452e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6747033441788517e-05 }, { "epoch": 0.34723349018335226, "step": 1070, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396639153332217e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.526190317235887e-05 }, { "epoch": 0.34723349018335226, "step": 1070, "weight_max": 1.2744457711377777e-09, "weight_mean": 7.156411019337838e-10, "weight_min": 0.0, "weight_std": 3.8612368857826596e-10 }, { "epoch": 0.3475580074638975, "step": 1071, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4766771528229583e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.040224899246823e-05 }, { "epoch": 0.3475580074638975, "step": 1071, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0114865290233865e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.873225482995622e-05 }, { "epoch": 0.3475580074638975, "step": 1071, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.85826065717265e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9902598069165833e-05 }, { "epoch": 0.3475580074638975, "step": 1071, "weight_max": 3.255531191825867e-05, "weight_mean": 9.876479452941567e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4068745258555282e-05 }, { "epoch": 0.34788252474444264, "step": 1072, "weight_max": 4.252122016623616e-05, "weight_mean": 8.776384675002191e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.672468351898715e-05 }, { "epoch": 0.34788252474444264, "step": 1072, "weight_max": 3.998322426923551e-05, "weight_mean": 9.464631148148328e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.756063284119591e-05 }, { "epoch": 0.34788252474444264, "step": 1072, "weight_max": 3.662098970380612e-05, "weight_mean": 7.418810128001496e-06, "weight_min": 0.0, "weight_std": 1.4228908185032196e-05 }, { "epoch": 0.34788252474444264, "step": 1072, "weight_max": 4.187939703115262e-05, "weight_mean": 1.906693069031462e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.054564538411796e-05 }, { "epoch": 0.3482070420249878, "step": 1073, "weight_max": 2.820166446326766e-05, "weight_mean": 7.407867997244466e-06, "weight_min": 0.0, "weight_std": 1.19818596431287e-05 }, { "epoch": 0.3482070420249878, "step": 1073, "weight_max": 4.187939703115262e-05, "weight_mean": 8.165051440300886e-06, "weight_min": 0.0, "weight_std": 1.5900819562375546e-05 }, { "epoch": 0.3482070420249878, "step": 1073, "weight_max": 3.967146767536178e-05, "weight_mean": 1.0816929716384038e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.782251820259262e-05 }, { "epoch": 0.3482070420249878, "step": 1073, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4998453480075113e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0716835933853872e-05 }, { "epoch": 0.34853155930553303, "step": 1074, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.84387328469893e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3697520444111433e-05 }, { "epoch": 0.34853155930553303, "step": 1074, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9585495465435088e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1179550458327867e-05 }, { "epoch": 0.34853155930553303, "step": 1074, "weight_max": 4.187939703115262e-05, "weight_mean": 1.492662249802379e-05, "weight_min": 0.0, "weight_std": 2.0645935364882462e-05 }, { "epoch": 0.34853155930553303, "step": 1074, "weight_max": 3.843663216684945e-05, "weight_mean": 9.609892913431395e-06, "weight_min": 0.0, "weight_std": 1.7792250218917616e-05 }, { "epoch": 0.3488560765860782, "step": 1075, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.032262485357933e-06, "weight_min": 0.0, "weight_std": 1.536634772492107e-05 }, { "epoch": 0.3488560765860782, "step": 1075, "weight_max": 3.602553260861896e-05, "weight_mean": 8.683027772349305e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5916542906779796e-05 }, { "epoch": 0.3488560765860782, "step": 1075, "weight_max": 4.187939703115262e-05, "weight_mean": 1.97673052753089e-05, "weight_min": 0.0, "weight_std": 2.1169473257032223e-05 }, { "epoch": 0.3488560765860782, "step": 1075, "weight_max": 2.86272097582696e-06, "weight_mean": 4.298287024084857e-07, "weight_min": 0.0, "weight_std": 1.0031784540842636e-06 }, { "epoch": 0.3491805938666234, "step": 1076, "weight_max": 3.843663216684945e-05, "weight_mean": 9.05538399820216e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6801393940113485e-05 }, { "epoch": 0.3491805938666234, "step": 1076, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390744137112051e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.923940726555884e-05 }, { "epoch": 0.3491805938666234, "step": 1076, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0732742339314427e-05, "weight_min": 0.0, "weight_std": 1.7119700714829378e-05 }, { "epoch": 0.3491805938666234, "step": 1076, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2520910786406603e-05, "weight_min": 0.0, "weight_std": 1.7731410480337217e-05 }, { "epoch": 0.3495051111471686, "step": 1077, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.859820809448138e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.640253140067216e-05 }, { "epoch": 0.3495051111471686, "step": 1077, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7859867512015626e-05, "weight_min": 0.0, "weight_std": 2.0740795662277378e-05 }, { "epoch": 0.3495051111471686, "step": 1077, "weight_max": 5.107872766529908e-06, "weight_mean": 6.389934696926503e-07, "weight_min": 0.0, "weight_std": 1.8056998669635504e-06 }, { "epoch": 0.3495051111471686, "step": 1077, "weight_max": 4.187939703115262e-05, "weight_mean": 9.925915946951136e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8414641090203077e-05 }, { "epoch": 0.34982962842771376, "step": 1078, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7933750718366355e-06, "weight_min": 0.0, "weight_std": 1.0726232176239137e-05 }, { "epoch": 0.34982962842771376, "step": 1078, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.551279092382174e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.143830897693988e-05 }, { "epoch": 0.34982962842771376, "step": 1078, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0114814358530566e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8732287571765482e-05 }, { "epoch": 0.34982962842771376, "step": 1078, "weight_max": 3.752332122530788e-05, "weight_mean": 9.12037467060145e-06, "weight_min": 0.0, "weight_std": 1.6895419321372174e-05 }, { "epoch": 0.350154145708259, "step": 1079, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4841017875587568e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0600946299964562e-05 }, { "epoch": 0.350154145708259, "step": 1079, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.3088261766824871e-05, "weight_min": 0.0, "weight_std": 1.8145547073800117e-05 }, { "epoch": 0.350154145708259, "step": 1079, "weight_max": 3.813097282545641e-05, "weight_mean": 6.7193200266046915e-06, "weight_min": 0.0, "weight_std": 1.381878973916173e-05 }, { "epoch": 0.350154145708259, "step": 1079, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235878107778262e-05, "weight_min": 0.0, "weight_std": 1.8970469682244584e-05 }, { "epoch": 0.35047866298880415, "grad_norm": 4.807929039001465, "learning_rate": 8.008163265306124e-06, "loss": 0.0938, "step": 1080 }, { "epoch": 0.35047866298880415, "step": 1080, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.404364391230047e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.7517486412543803e-05 }, { "epoch": 0.35047866298880415, "step": 1080, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.331718062327127e-06, "weight_min": 0.0, "weight_std": 9.421841241419315e-06 }, { "epoch": 0.35047866298880415, "step": 1080, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.996821972890757e-06, "weight_min": 0.0, "weight_std": 1.8512528185965493e-05 }, { "epoch": 0.35047866298880415, "step": 1080, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2849886843468994e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8170308976550587e-05 }, { "epoch": 0.35080318026934937, "step": 1081, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235716344031971e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806283616053406e-05 }, { "epoch": 0.35080318026934937, "step": 1081, "weight_max": 4.187939703115262e-05, "weight_mean": 1.458031874790322e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.017745646298863e-05 }, { "epoch": 0.35080318026934937, "step": 1081, "weight_max": 3.782653220696375e-05, "weight_mean": 1.7108235624618828e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.848191641329322e-05 }, { "epoch": 0.35080318026934937, "step": 1081, "weight_max": 3.632264997577295e-05, "weight_mean": 8.788556442596018e-06, "weight_min": 0.0, "weight_std": 1.628346035431605e-05 }, { "epoch": 0.35112769754989454, "step": 1082, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.6206406144192442e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7964892322197556e-05 }, { "epoch": 0.35112769754989454, "step": 1082, "weight_max": 3.998322426923551e-05, "weight_mean": 8.82550830283435e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.6529826098121703e-05 }, { "epoch": 0.35112769754989454, "step": 1082, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9972365407738835e-05, "weight_min": 0.0, "weight_std": 2.142985977116041e-05 }, { "epoch": 0.35112769754989454, "step": 1082, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.3567347347561736e-06, "weight_min": 0.0, "weight_std": 1.2320843779889401e-05 }, { "epoch": 0.3514522148304397, "step": 1083, "weight_max": 2.5141427613561973e-05, "weight_mean": 3.1522101835435024e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.88502017915016e-06 }, { "epoch": 0.3514522148304397, "step": 1083, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320248081057798e-06, "weight_min": 0.0, "weight_std": 1.2217999938002322e-05 }, { "epoch": 0.3514522148304397, "step": 1083, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.375349165755324e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9014960344065912e-05 }, { "epoch": 0.3514522148304397, "step": 1083, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0561417841236107e-05, "weight_min": 0.0, "weight_std": 2.2059526600060053e-05 }, { "epoch": 0.3517767321109849, "step": 1084, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4079829270485789e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.94651547644753e-05 }, { "epoch": 0.3517767321109849, "step": 1084, "weight_max": 4.252122016623616e-05, "weight_mean": 8.458430784230586e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6334823158103973e-05 }, { "epoch": 0.3517767321109849, "step": 1084, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0664641851908527e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.2105215975898318e-05 }, { "epoch": 0.3517767321109849, "step": 1084, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5270206858986057e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1079009457025677e-05 }, { "epoch": 0.3521012493915301, "step": 1085, "weight_max": 4.252122016623616e-05, "weight_mean": 9.017679985845461e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7027094145305455e-05 }, { "epoch": 0.3521012493915301, "step": 1085, "weight_max": 3.813097282545641e-05, "weight_mean": 9.233142918674275e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.710710239422042e-05 }, { "epoch": 0.3521012493915301, "step": 1085, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4597459085052833e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0418903659447096e-05 }, { "epoch": 0.3521012493915301, "step": 1085, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5040010111988522e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0786417735507712e-05 }, { "epoch": 0.3524257666720753, "step": 1086, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7026242858264595e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8668733900994994e-05 }, { "epoch": 0.3524257666720753, "step": 1086, "weight_max": 9.458863914346693e-09, "weight_mean": 2.1602426514277795e-09, "weight_min": 0.0, "weight_std": 3.08083736300091e-09 }, { "epoch": 0.3524257666720753, "step": 1086, "weight_max": 3.752332122530788e-05, "weight_mean": 1.285545113205444e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.791207068890799e-05 }, { "epoch": 0.3524257666720753, "step": 1086, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311900950910058e-05, "weight_min": 0.0, "weight_std": 1.9095274183200672e-05 }, { "epoch": 0.3527502839526205, "step": 1087, "weight_max": 3.967146767536178e-05, "weight_mean": 8.010107194422744e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5381237972178496e-05 }, { "epoch": 0.3527502839526205, "step": 1087, "weight_max": 1.504423474685268e-09, "weight_mean": 1.0152767515592132e-09, "weight_min": 5.771870759119224e-10, "weight_std": 3.590242270146149e-10 }, { "epoch": 0.3527502839526205, "step": 1087, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076979959994787e-06, "weight_min": 0.0, "weight_std": 1.4357671716425102e-05 }, { "epoch": 0.3527502839526205, "step": 1087, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.767233677848708e-06, "weight_min": 0.0, "weight_std": 1.810293542803265e-05 }, { "epoch": 0.35307480123316565, "step": 1088, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.436317794898059e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2525355487014167e-05 }, { "epoch": 0.35307480123316565, "step": 1088, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.359423655841965e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2319757843215484e-05 }, { "epoch": 0.35307480123316565, "step": 1088, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0276391549268737e-05, "weight_min": 0.0, "weight_std": 1.7408903659088537e-05 }, { "epoch": 0.35307480123316565, "step": 1088, "weight_max": 3.813097282545641e-05, "weight_mean": 8.836492270347662e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6427984519395977e-05 }, { "epoch": 0.35339931851371087, "step": 1089, "weight_max": 3.752332122530788e-05, "weight_mean": 7.107386409188621e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.402823363605421e-05 }, { "epoch": 0.35339931851371087, "step": 1089, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.850214357546065e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9794390027527697e-05 }, { "epoch": 0.35339931851371087, "step": 1089, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615628313331399e-06, "weight_min": 0.0, "weight_std": 1.3053157999820542e-05 }, { "epoch": 0.35339931851371087, "step": 1089, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.241078259947244e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.715631879051216e-05 }, { "epoch": 0.35372383579425604, "grad_norm": 2.8107166290283203, "learning_rate": 7.987755102040818e-06, "loss": 0.1831, "step": 1090 }, { "epoch": 0.35372383579425604, "step": 1090, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233744433207903e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.895372406579554e-05 }, { "epoch": 0.35372383579425604, "step": 1090, "weight_max": 3.998322426923551e-05, "weight_mean": 7.3344331212865654e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.472318672313122e-05 }, { "epoch": 0.35372383579425604, "step": 1090, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4300763875071425e-06, "weight_min": 0.0, "weight_std": 1.2527864782896359e-05 }, { "epoch": 0.35372383579425604, "step": 1090, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8928831195808016e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.044611937890295e-05 }, { "epoch": 0.35404835307480126, "step": 1091, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8411499695503153e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9792572857113555e-05 }, { "epoch": 0.35404835307480126, "step": 1091, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4042334441910498e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9414041162235662e-05 }, { "epoch": 0.35404835307480126, "step": 1091, "weight_max": 4.252122016623616e-05, "weight_mean": 9.108231097343378e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7174968888866715e-05 }, { "epoch": 0.35404835307480126, "step": 1091, "weight_max": 4.187939703115262e-05, "weight_mean": 1.26854765767348e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7901274986797944e-05 }, { "epoch": 0.3543728703553464, "step": 1092, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3388161278271582e-05, "weight_min": 0.0, "weight_std": 1.859007534221746e-05 }, { "epoch": 0.3543728703553464, "step": 1092, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.610760227777064e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.223046431026887e-05 }, { "epoch": 0.3543728703553464, "step": 1092, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.9928469555452466e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.6660686014802195e-05 }, { "epoch": 0.3543728703553464, "step": 1092, "weight_max": 3.998322426923551e-05, "weight_mean": 1.457421421946492e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.013934135902673e-05 }, { "epoch": 0.3546973876358916, "step": 1093, "weight_max": 7.829137893544669e-10, "weight_mean": 6.41089736852507e-10, "weight_min": 5.679522963042416e-10, "weight_std": 6.686766418351198e-11 }, { "epoch": 0.3546973876358916, "step": 1093, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074814781546593e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8654176528798416e-05 }, { "epoch": 0.3546973876358916, "step": 1093, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235463959252229e-06, "weight_min": 0.0, "weight_std": 1.4806385479460005e-05 }, { "epoch": 0.3546973876358916, "step": 1093, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4583130905521102e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0191560906823725e-05 }, { "epoch": 0.3550219049164368, "step": 1094, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4536501112161204e-05, "weight_min": 0.0, "weight_std": 2.009427953453269e-05 }, { "epoch": 0.3550219049164368, "step": 1094, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0117030797118787e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8746723071672022e-05 }, { "epoch": 0.3550219049164368, "step": 1094, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1780911336245481e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.708766103547532e-05 }, { "epoch": 0.3550219049164368, "step": 1094, "weight_max": 4.252122016623616e-05, "weight_mean": 1.538991455163341e-05, "weight_min": 0.0, "weight_std": 2.1250618374324404e-05 }, { "epoch": 0.355346422196982, "step": 1095, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4079956599744037e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9247499949415214e-05 }, { "epoch": 0.355346422196982, "step": 1095, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3714640772377606e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9163440811098553e-05 }, { "epoch": 0.355346422196982, "step": 1095, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0769594963639975e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.435768081137212e-05 }, { "epoch": 0.355346422196982, "step": 1095, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4049597666598856e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9456227164482698e-05 }, { "epoch": 0.3556709394775272, "step": 1096, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3658319403475616e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8313749023946002e-05 }, { "epoch": 0.3556709394775272, "step": 1096, "weight_max": 2.6141986381844617e-05, "weight_mean": 4.030495347251417e-06, "weight_min": 6.196655966128617e-10, "weight_std": 9.180182132695336e-06 }, { "epoch": 0.3556709394775272, "step": 1096, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.484175845689606e-05, "weight_min": 0.0, "weight_std": 2.049780778179411e-05 }, { "epoch": 0.3556709394775272, "step": 1096, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.006457730720285e-06, "weight_min": 0.0, "weight_std": 1.6854835848789662e-05 }, { "epoch": 0.35599545675807237, "step": 1097, "weight_max": 1.3004798347537871e-05, "weight_mean": 1.6261558357655304e-06, "weight_min": 0.0, "weight_std": 4.597665792971384e-06 }, { "epoch": 0.35599545675807237, "step": 1097, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.403379408468027e-06, "weight_min": 0.0, "weight_std": 1.751809031702578e-05 }, { "epoch": 0.35599545675807237, "step": 1097, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8431452190270647e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0173631128272973e-05 }, { "epoch": 0.35599545675807237, "step": 1097, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.121804861119017e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.434010573575506e-05 }, { "epoch": 0.35631997403861754, "step": 1098, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.031500667333603e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6842357581481338e-05 }, { "epoch": 0.35631997403861754, "step": 1098, "weight_max": 2.109268537608955e-09, "weight_mean": 9.143463208793889e-10, "weight_min": 5.679522963042416e-10, "weight_std": 5.157615445838815e-10 }, { "epoch": 0.35631997403861754, "step": 1098, "weight_max": 3.813097282545641e-05, "weight_mean": 9.27014752960531e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7172756997752003e-05 }, { "epoch": 0.35631997403861754, "step": 1098, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.6527974366908893e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8511573216528632e-05 }, { "epoch": 0.35664449131916276, "step": 1099, "weight_max": 4.252122016623616e-05, "weight_mean": 9.819347724260297e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.826212610467337e-05 }, { "epoch": 0.35664449131916276, "step": 1099, "weight_max": 8.490471659072796e-10, "weight_mean": 7.046282446410146e-10, "weight_min": 6.053385570581327e-10, "weight_std": 9.153507118941917e-11 }, { "epoch": 0.35664449131916276, "step": 1099, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077757123217452e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357357031258289e-05 }, { "epoch": 0.35664449131916276, "step": 1099, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0443518476677127e-05, "weight_min": 0.0, "weight_std": 1.593918386788573e-05 }, { "epoch": 0.3569690085997079, "grad_norm": 22.317380905151367, "learning_rate": 7.967346938775511e-06, "loss": 0.1445, "step": 1100 }, { "epoch": 0.3569690085997079, "eval_accuracy": 0.969, "eval_auc": 0.9930655014322063, "eval_loss": 0.18540063500404358, "eval_runtime": 156.2347, "eval_samples_per_second": 12.801, "eval_steps_per_second": 1.6, "step": 1100 }, { "epoch": 0.3569690085997079, "step": 1100, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4237624782253988e-05, "weight_min": 0.0, "weight_std": 1.944068935699761e-05 }, { "epoch": 0.3569690085997079, "step": 1100, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5040339349070564e-05, "weight_min": 0.0, "weight_std": 2.07881039386848e-05 }, { "epoch": 0.3569690085997079, "step": 1100, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0773919610946905e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.435750618838938e-05 }, { "epoch": 0.3569690085997079, "step": 1100, "weight_max": 1.1946053746214602e-06, "weight_mean": 1.5006288833774306e-07, "weight_min": 0.0, "weight_std": 4.220591733883339e-07 }, { "epoch": 0.35729352588025315, "step": 1101, "weight_max": 3.998322426923551e-05, "weight_mean": 9.427725672139786e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7497912267572246e-05 }, { "epoch": 0.35729352588025315, "step": 1101, "weight_max": 4.187939703115262e-05, "weight_mean": 1.982313369808253e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.129670428985264e-05 }, { "epoch": 0.35729352588025315, "step": 1101, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156174665899016e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581060895579867e-05 }, { "epoch": 0.35729352588025315, "step": 1101, "weight_max": 3.006794759130571e-05, "weight_mean": 3.759166020245175e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0630352335283533e-05 }, { "epoch": 0.3576180431607983, "step": 1102, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155621566809714e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8815166185959242e-05 }, { "epoch": 0.3576180431607983, "step": 1102, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7740992916515097e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9364961190149188e-05 }, { "epoch": 0.3576180431607983, "step": 1102, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0773196562658995e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357534382725134e-05 }, { "epoch": 0.3576180431607983, "step": 1102, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.377919215883594e-05, "weight_min": 0.0, "weight_std": 1.905431599880103e-05 }, { "epoch": 0.3579425604413435, "step": 1103, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0076018043037038e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8662467482499778e-05 }, { "epoch": 0.3579425604413435, "step": 1103, "weight_max": 3.0339444492710754e-05, "weight_mean": 7.3511992013663985e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3619289347843733e-05 }, { "epoch": 0.3579425604413435, "step": 1103, "weight_max": 3.813097282545641e-05, "weight_mean": 9.457242413191125e-06, "weight_min": 0.0, "weight_std": 1.7511038095108233e-05 }, { "epoch": 0.3579425604413435, "step": 1103, "weight_max": 3.967146767536178e-05, "weight_mean": 7.426226147799753e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.380375942972023e-05 }, { "epoch": 0.3582670777218887, "step": 1104, "weight_max": 2.440382377244532e-05, "weight_mean": 3.0513378987961914e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.627706847619265e-06 }, { "epoch": 0.3582670777218887, "step": 1104, "weight_max": 6.1986669130931205e-09, "weight_mean": 1.5203713843447986e-09, "weight_min": 0.0, "weight_std": 2.0096475594755248e-09 }, { "epoch": 0.3582670777218887, "step": 1104, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805184744327562e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3589156878879294e-05 }, { "epoch": 0.3582670777218887, "step": 1104, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6535317627131008e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1277586711221375e-05 }, { "epoch": 0.35859159500243387, "step": 1105, "weight_max": 3.312147964606993e-05, "weight_mean": 4.141087629250251e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1709847058227751e-05 }, { "epoch": 0.35859159500243387, "step": 1105, "weight_max": 3.36925259034615e-05, "weight_mean": 6.822748673585011e-06, "weight_min": 0.0, "weight_std": 1.3086433682474308e-05 }, { "epoch": 0.35859159500243387, "step": 1105, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.238820555561688e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5564666682621464e-05 }, { "epoch": 0.35859159500243387, "step": 1105, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3763573406322394e-05, "weight_min": 0.0, "weight_std": 1.9126491679344326e-05 }, { "epoch": 0.3589161122829791, "step": 1106, "weight_max": 3.998322426923551e-05, "weight_mean": 8.825265467748977e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.652997343626339e-05 }, { "epoch": 0.3589161122829791, "step": 1106, "weight_max": 2.952862632810138e-05, "weight_mean": 3.694924771480146e-06, "weight_min": 0.0, "weight_std": 1.043839529302204e-05 }, { "epoch": 0.3589161122829791, "step": 1106, "weight_max": 3.116124207735993e-05, "weight_mean": 3.8966986721789e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1016539247066248e-05 }, { "epoch": 0.3589161122829791, "step": 1106, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3547148884972557e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.974051701836288e-05 }, { "epoch": 0.35924062956352426, "step": 1107, "weight_max": 3.998322426923551e-05, "weight_mean": 8.928496754379012e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6687195966369472e-05 }, { "epoch": 0.35924062956352426, "step": 1107, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0699208360165358e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.688593147264328e-05 }, { "epoch": 0.35924062956352426, "step": 1107, "weight_max": 3.813097282545641e-05, "weight_mean": 8.03851980890613e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.521481044619577e-05 }, { "epoch": 0.35924062956352426, "step": 1107, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4004903277964331e-05, "weight_min": 0.0, "weight_std": 1.967432581295725e-05 }, { "epoch": 0.3595651468440694, "step": 1108, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315720954968128e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5033292584121227e-05 }, { "epoch": 0.3595651468440694, "step": 1108, "weight_max": 3.28377827827353e-05, "weight_mean": 4.1056855479837395e-06, "weight_min": 0.0, "weight_std": 1.160952069767518e-05 }, { "epoch": 0.3595651468440694, "step": 1108, "weight_max": 4.252122016623616e-05, "weight_mean": 8.973359399533365e-06, "weight_min": 0.0, "weight_std": 1.6987722119665705e-05 }, { "epoch": 0.3595651468440694, "step": 1108, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4617789929616265e-05, "weight_min": 0.0, "weight_std": 2.0224335457896814e-05 }, { "epoch": 0.35988966412461465, "step": 1109, "weight_max": 3.752332122530788e-05, "weight_mean": 9.09109803615138e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6825761122163385e-05 }, { "epoch": 0.35988966412461465, "step": 1109, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.887578926282004e-06, "weight_min": 0.0, "weight_std": 1.729512405290734e-05 }, { "epoch": 0.35988966412461465, "step": 1109, "weight_max": 1.0261104188558079e-09, "weight_mean": 5.569044114750454e-10, "weight_min": 0.0, "weight_std": 3.783348079267057e-10 }, { "epoch": 0.35988966412461465, "step": 1109, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9211967052542605e-06, "weight_min": 0.0, "weight_std": 1.3915752788307145e-05 }, { "epoch": 0.3602141814051598, "grad_norm": 5.436362266540527, "learning_rate": 7.946938775510204e-06, "loss": 0.0674, "step": 1110 }, { "epoch": 0.3602141814051598, "step": 1110, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3355899682210293e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8725153495324776e-05 }, { "epoch": 0.3602141814051598, "step": 1110, "weight_max": 2.319888153579086e-05, "weight_mean": 2.900555045926012e-06, "weight_min": 6.053385570581327e-10, "weight_std": 8.20176319393795e-06 }, { "epoch": 0.3602141814051598, "step": 1110, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2293033250898588e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9356137272552587e-05 }, { "epoch": 0.3602141814051598, "step": 1110, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.016489270317834e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.162409873562865e-05 }, { "epoch": 0.36053869868570504, "step": 1111, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5549758245470002e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1476591427926905e-05 }, { "epoch": 0.36053869868570504, "step": 1111, "weight_max": 1.9187997857272876e-09, "weight_mean": 8.441807253234401e-10, "weight_min": 5.679522963042416e-10, "weight_std": 4.4491418882408595e-10 }, { "epoch": 0.36053869868570504, "step": 1111, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.485763257544022e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7413405657862313e-05 }, { "epoch": 0.36053869868570504, "step": 1111, "weight_max": 3.397988257347606e-05, "weight_mean": 9.14190968615003e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3828358532919083e-05 }, { "epoch": 0.3608632159662502, "step": 1112, "weight_max": 3.662098970380612e-05, "weight_mean": 7.69042526371777e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4578279660781845e-05 }, { "epoch": 0.3608632159662502, "step": 1112, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861330242216354e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0919650776486378e-05 }, { "epoch": 0.3608632159662502, "step": 1112, "weight_max": 3.967146767536178e-05, "weight_mean": 7.485572950827191e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4803236808802467e-05 }, { "epoch": 0.3608632159662502, "step": 1112, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3167177611612715e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.503288876847364e-05 }, { "epoch": 0.3611877332467954, "step": 1113, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5309216905734502e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.112673973897472e-05 }, { "epoch": 0.3611877332467954, "step": 1113, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.526654523331672e-05, "weight_min": 0.0, "weight_std": 2.102421785821207e-05 }, { "epoch": 0.3611877332467954, "step": 1113, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3874938051449135e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0243671315256506e-05 }, { "epoch": 0.3611877332467954, "step": 1113, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4311108316178434e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.98004454432521e-05 }, { "epoch": 0.3615122505273406, "step": 1114, "weight_max": 3.255531191825867e-05, "weight_mean": 1.5227818948915228e-05, "weight_min": 0.0, "weight_std": 1.6375302948290482e-05 }, { "epoch": 0.3615122505273406, "step": 1114, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2834738299716264e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.790095484466292e-05 }, { "epoch": 0.3615122505273406, "step": 1114, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.317187343374826e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8596387235447764e-05 }, { "epoch": 0.3615122505273406, "step": 1114, "weight_max": 4.187939703115262e-05, "weight_mean": 9.062176104635e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7046171706169844e-05 }, { "epoch": 0.36183676780788576, "step": 1115, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156146926310612e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.458107271901099e-05 }, { "epoch": 0.36183676780788576, "step": 1115, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.044982107297983e-06, "weight_min": 0.0, "weight_std": 1.6751417206251062e-05 }, { "epoch": 0.36183676780788576, "step": 1115, "weight_max": 4.252122016623616e-05, "weight_mean": 2.003725603572093e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1446658138302155e-05 }, { "epoch": 0.36183676780788576, "step": 1115, "weight_max": 9.914649545805787e-09, "weight_mean": 1.9354600144794176e-09, "weight_min": 5.864964625068581e-10, "weight_std": 3.22756621429221e-09 }, { "epoch": 0.362161285088431, "step": 1116, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156023689778522e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4581121831724886e-05 }, { "epoch": 0.362161285088431, "step": 1116, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274839041812811e-05, "weight_min": 0.0, "weight_std": 1.903868542285636e-05 }, { "epoch": 0.362161285088431, "step": 1116, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.680635230324697e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.8079505025525577e-05 }, { "epoch": 0.362161285088431, "step": 1116, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3941247743787244e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9493782019708306e-05 }, { "epoch": 0.36248580236897615, "step": 1117, "weight_max": 3.632264997577295e-05, "weight_mean": 1.2504496226029005e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7311875126324594e-05 }, { "epoch": 0.36248580236897615, "step": 1117, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.919091482704971e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.055964978353586e-05 }, { "epoch": 0.36248580236897615, "step": 1117, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4926677977200598e-05, "weight_min": 0.0, "weight_std": 2.0645891709136777e-05 }, { "epoch": 0.36248580236897615, "step": 1117, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.641834938316606e-06, "weight_min": 0.0, "weight_std": 1.6006126315915026e-05 }, { "epoch": 0.3628103196495213, "step": 1118, "weight_max": 3.967146767536178e-05, "weight_mean": 9.574574505677447e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.774249176378362e-05 }, { "epoch": 0.3628103196495213, "step": 1118, "weight_max": 1.1534618815645103e-09, "weight_mean": 7.498261456184707e-10, "weight_min": 5.771870759119224e-10, "weight_std": 2.1304699671098604e-10 }, { "epoch": 0.3628103196495213, "step": 1118, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2390141718206e-06, "weight_min": 0.0, "weight_std": 1.4804953934799414e-05 }, { "epoch": 0.3628103196495213, "step": 1118, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070052455062978e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1509783689689357e-05 }, { "epoch": 0.36313483693006654, "step": 1119, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.209009835496545e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.866869024524931e-05 }, { "epoch": 0.36313483693006654, "step": 1119, "weight_max": 7.449420081684366e-06, "weight_mean": 9.319101650362427e-07, "weight_min": 5.771870759119224e-10, "weight_std": 2.6334719223086722e-06 }, { "epoch": 0.36313483693006654, "step": 1119, "weight_max": 3.22740561387036e-05, "weight_mean": 4.034750872961013e-06, "weight_min": 0.0, "weight_std": 1.1410402294131927e-05 }, { "epoch": 0.36313483693006654, "step": 1119, "weight_max": 3.397988257347606e-05, "weight_mean": 4.397974407766014e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1960184565396048e-05 }, { "epoch": 0.3634593542106117, "grad_norm": 20.099864959716797, "learning_rate": 7.926530612244897e-06, "loss": 0.0979, "step": 1120 }, { "epoch": 0.3634593542106117, "step": 1120, "weight_max": 1.2217544735904085e-06, "weight_mean": 1.5346546433647745e-07, "weight_min": 5.771870759119224e-10, "weight_std": 4.3165414354007225e-07 }, { "epoch": 0.3634593542106117, "step": 1120, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4368870324688032e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.996945502469316e-05 }, { "epoch": 0.3634593542106117, "step": 1120, "weight_max": 9.062063877074422e-10, "weight_mean": 6.447518630103843e-10, "weight_min": 0.0, "weight_std": 2.9048052851976536e-10 }, { "epoch": 0.3634593542106117, "step": 1120, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861679033434484e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0919509804807603e-05 }, { "epoch": 0.36378387149115693, "step": 1121, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691957201430341e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3265874258650001e-05 }, { "epoch": 0.36378387149115693, "step": 1121, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0001918781199493e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8545497368904762e-05 }, { "epoch": 0.36378387149115693, "step": 1121, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6050718841142952e-05, "weight_min": 0.0, "weight_std": 2.1578189262072556e-05 }, { "epoch": 0.36378387149115693, "step": 1121, "weight_max": 4.252122016623616e-05, "weight_mean": 1.815857285691891e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0559051336022094e-05 }, { "epoch": 0.3641083887717021, "step": 1122, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3145639968570322e-05, "weight_min": 0.0, "weight_std": 1.860561314970255e-05 }, { "epoch": 0.3641083887717021, "step": 1122, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8808186723617837e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0435185433598235e-05 }, { "epoch": 0.3641083887717021, "step": 1122, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3157841648499016e-06, "weight_min": 0.0, "weight_std": 1.5033266208774876e-05 }, { "epoch": 0.3641083887717021, "step": 1122, "weight_max": 4.252122016623616e-05, "weight_mean": 9.747787771630101e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.8140468455385417e-05 }, { "epoch": 0.36443290605224726, "step": 1123, "weight_max": 3.843663216684945e-05, "weight_mean": 9.272089300793596e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7180589566123672e-05 }, { "epoch": 0.36443290605224726, "step": 1123, "weight_max": 3.752332122530788e-05, "weight_mean": 1.8018283299170434e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.928271922224667e-05 }, { "epoch": 0.36443290605224726, "step": 1123, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076783509139204e-06, "weight_min": 0.0, "weight_std": 1.4357750842464156e-05 }, { "epoch": 0.36443290605224726, "step": 1123, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.023997538140975e-05, "weight_min": 0.0, "weight_std": 1.894085835374426e-05 }, { "epoch": 0.3647574233327925, "step": 1124, "weight_max": 4.252122016623616e-05, "weight_mean": 9.176563253276981e-06, "weight_min": 0.0, "weight_std": 1.7272313925786875e-05 }, { "epoch": 0.3647574233327925, "step": 1124, "weight_max": 4.187939703115262e-05, "weight_mean": 1.812851769500412e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.995984530367423e-05 }, { "epoch": 0.3647574233327925, "step": 1124, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920788796880515e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3915917406848166e-05 }, { "epoch": 0.3647574233327925, "step": 1124, "weight_max": 5.3206110628423176e-09, "weight_mean": 2.0763226693532033e-09, "weight_min": 6.29309993005478e-10, "weight_std": 2.0611243822798997e-09 }, { "epoch": 0.36508194061333765, "step": 1125, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0279029083903879e-05, "weight_min": 0.0, "weight_std": 1.9060917111346498e-05 }, { "epoch": 0.36508194061333765, "step": 1125, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7599508282728493e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.015083191508893e-05 }, { "epoch": 0.36508194061333765, "step": 1125, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4429209588561207e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0323501303209923e-05 }, { "epoch": 0.36508194061333765, "step": 1125, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3368326108320616e-05, "weight_min": 0.0, "weight_std": 1.8476608602213673e-05 }, { "epoch": 0.3654064578938829, "step": 1126, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8398415704723448e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0525081708910875e-05 }, { "epoch": 0.3654064578938829, "step": 1126, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0442899767658673e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.193407817685511e-05 }, { "epoch": 0.3654064578938829, "step": 1126, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.2967255315743387e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7929018213180825e-05 }, { "epoch": 0.3654064578938829, "step": 1126, "weight_max": 3.572963396436535e-05, "weight_mean": 4.467219696380198e-06, "weight_min": 0.0, "weight_std": 1.2631922800210305e-05 }, { "epoch": 0.36573097517442804, "step": 1127, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.4627981777885e-06, "weight_min": 0.0, "weight_std": 1.614530810911674e-05 }, { "epoch": 0.36573097517442804, "step": 1127, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3030593436269555e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8292232198291458e-05 }, { "epoch": 0.36573097517442804, "step": 1127, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5468169294763356e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1355201170081273e-05 }, { "epoch": 0.36573097517442804, "step": 1127, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.360049832845107e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.551421701151412e-05 }, { "epoch": 0.3660554924549732, "step": 1128, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.211585731653031e-05, "weight_min": 0.0, "weight_std": 1.7702630429994315e-05 }, { "epoch": 0.3660554924549732, "step": 1128, "weight_max": 4.3167921830900013e-05, "weight_mean": 7.65870208851993e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5237340448948089e-05 }, { "epoch": 0.3660554924549732, "step": 1128, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.6110298020066693e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.888856422738172e-05 }, { "epoch": 0.3660554924549732, "step": 1128, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541093858279055e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2841687748732511e-05 }, { "epoch": 0.36638000973551843, "step": 1129, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.450476065656403e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.038522325165104e-05 }, { "epoch": 0.36638000973551843, "step": 1129, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311155165254604e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9091214198851958e-05 }, { "epoch": 0.36638000973551843, "step": 1129, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4925382856745273e-05, "weight_min": 0.0, "weight_std": 2.0639048671000637e-05 }, { "epoch": 0.36638000973551843, "step": 1129, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0769112931448035e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4357699001266155e-05 }, { "epoch": 0.3667045270160636, "grad_norm": 0.10484948009252548, "learning_rate": 7.906122448979592e-06, "loss": 0.1292, "step": 1130 }, { "epoch": 0.3667045270160636, "step": 1130, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9669869288918562e-05, "weight_min": 0.0, "weight_std": 2.11448896152433e-05 }, { "epoch": 0.3667045270160636, "step": 1130, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076901288703084e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357703548739664e-05 }, { "epoch": 0.3667045270160636, "step": 1130, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.078132289781934e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357206055137794e-05 }, { "epoch": 0.3667045270160636, "step": 1130, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156971383257769e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.458073893445544e-05 }, { "epoch": 0.3670290442966088, "step": 1131, "weight_max": 4.252122016623616e-05, "weight_mean": 5.339339168131119e-06, "weight_min": 0.0, "weight_std": 1.5023889318399597e-05 }, { "epoch": 0.3670290442966088, "step": 1131, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.4477471268037334e-05, "weight_min": 0.0, "weight_std": 1.8202201317762956e-05 }, { "epoch": 0.3670290442966088, "step": 1131, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0777084652509075e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.435737794963643e-05 }, { "epoch": 0.3670290442966088, "step": 1131, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3529952159151435e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.9941313439630903e-05 }, { "epoch": 0.367353561577154, "step": 1132, "weight_max": 3.967146767536178e-05, "weight_mean": 9.653511369833723e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.7873779142973945e-05 }, { "epoch": 0.367353561577154, "step": 1132, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156866336619714e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4580781680706423e-05 }, { "epoch": 0.367353561577154, "step": 1132, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155714335269295e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8815109797287732e-05 }, { "epoch": 0.367353561577154, "step": 1132, "weight_max": 4.187939703115262e-05, "weight_mean": 1.039132803271059e-05, "weight_min": 0.0, "weight_std": 1.923904892464634e-05 }, { "epoch": 0.36767807885769915, "step": 1133, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9653747585834935e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1059575374238193e-05 }, { "epoch": 0.36767807885769915, "step": 1133, "weight_max": 2.5561155414521863e-09, "weight_mean": 1.0774199310503718e-09, "weight_min": 0.0, "weight_std": 7.740961205371377e-10 }, { "epoch": 0.36767807885769915, "step": 1133, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4119490515440702e-05, "weight_min": 0.0, "weight_std": 1.952855018316768e-05 }, { "epoch": 0.36767807885769915, "step": 1133, "weight_max": 3.998322426923551e-05, "weight_mean": 8.393649295612704e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.591409818502143e-05 }, { "epoch": 0.3680025961382444, "step": 1134, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8719147192314267e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0560857592499815e-05 }, { "epoch": 0.3680025961382444, "step": 1134, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.648893865232822e-06, "weight_min": 0.0, "weight_std": 1.786984648788348e-05 }, { "epoch": 0.3680025961382444, "step": 1134, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077257355878828e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357559848576784e-05 }, { "epoch": 0.3680025961382444, "step": 1134, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1374891982995905e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.86706347449217e-05 }, { "epoch": 0.36832711341878954, "step": 1135, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0516959921224043e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.045643122983165e-05 }, { "epoch": 0.36832711341878954, "step": 1135, "weight_max": 3.662098970380612e-05, "weight_mean": 8.612449164502323e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.598802009539213e-05 }, { "epoch": 0.36832711341878954, "step": 1135, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.888389412371907e-06, "weight_min": 0.0, "weight_std": 1.3804036825604271e-05 }, { "epoch": 0.36832711341878954, "step": 1135, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156014140084153e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4581126379198395e-05 }, { "epoch": 0.36865163069933476, "step": 1136, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.368233021523338e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.2316237189224921e-05 }, { "epoch": 0.36865163069933476, "step": 1136, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2808932297048159e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7950675101019442e-05 }, { "epoch": 0.36865163069933476, "step": 1136, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998417352908291e-06, "weight_min": 0.0, "weight_std": 1.413599693478318e-05 }, { "epoch": 0.36865163069933476, "step": 1136, "weight_max": 2.8994179956498556e-05, "weight_mean": 5.457221959659364e-06, "weight_min": 0.0, "weight_std": 1.0805130841617938e-05 }, { "epoch": 0.36897614797987993, "step": 1137, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.31156137085054e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.726690970826894e-05 }, { "epoch": 0.36897614797987993, "step": 1137, "weight_max": 3.255531191825867e-05, "weight_mean": 4.06998151447624e-06, "weight_min": 0.0, "weight_std": 1.1509812793519814e-05 }, { "epoch": 0.36897614797987993, "step": 1137, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155863163963659e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4581187315343414e-05 }, { "epoch": 0.36897614797987993, "step": 1137, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.00466292881174e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1496993213077076e-05 }, { "epoch": 0.3693006652604251, "step": 1138, "weight_max": 3.602553260861896e-05, "weight_mean": 8.896378858480603e-06, "weight_min": 0.0, "weight_std": 1.6473353753099218e-05 }, { "epoch": 0.3693006652604251, "step": 1138, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.3933514461969025e-06, "weight_min": 0.0, "weight_std": 1.242412872670684e-05 }, { "epoch": 0.3693006652604251, "step": 1138, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105382686248049e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1609642569965217e-05 }, { "epoch": 0.3693006652604251, "step": 1138, "weight_max": 4.252122016623616e-05, "weight_mean": 2.043282802333124e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1848427422810346e-05 }, { "epoch": 0.3696251825409703, "step": 1139, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235453045315808e-06, "weight_min": 0.0, "weight_std": 1.4806389117438812e-05 }, { "epoch": 0.3696251825409703, "step": 1139, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274607120663859e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.903882912301924e-05 }, { "epoch": 0.3696251825409703, "step": 1139, "weight_max": 3.967146767536178e-05, "weight_mean": 4.95969652547501e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4025674317963421e-05 }, { "epoch": 0.3696251825409703, "step": 1139, "weight_max": 3.998322426923551e-05, "weight_mean": 7.000539881119039e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4456640201387927e-05 }, { "epoch": 0.3699496998215155, "grad_norm": 4.3816914558410645, "learning_rate": 7.885714285714286e-06, "loss": 0.0373, "step": 1140 }, { "epoch": 0.3699496998215155, "step": 1140, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920818810205674e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3915904673922341e-05 }, { "epoch": 0.3699496998215155, "step": 1140, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.54795723373536e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.759155020408798e-05 }, { "epoch": 0.3699496998215155, "step": 1140, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9213626728742383e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9204833733965643e-05 }, { "epoch": 0.3699496998215155, "step": 1140, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4438059224630706e-05, "weight_min": 0.0, "weight_std": 2.0031533495057374e-05 }, { "epoch": 0.3702742171020607, "step": 1141, "weight_max": 3.998322426923551e-05, "weight_mean": 9.918502655636985e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8365055439062417e-05 }, { "epoch": 0.3702742171020607, "step": 1141, "weight_max": 2.020880492636934e-05, "weight_mean": 4.1799976315815e-06, "weight_min": 6.29309993005478e-10, "weight_std": 7.156974334066035e-06 }, { "epoch": 0.3702742171020607, "step": 1141, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.696274901216384e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8000529962591827e-05 }, { "epoch": 0.3702742171020607, "step": 1141, "weight_max": 3.632264997577295e-05, "weight_mean": 8.610581971879583e-06, "weight_min": 0.0, "weight_std": 1.597339905856643e-05 }, { "epoch": 0.3705987343826059, "step": 1142, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3335627954802476e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.85974822670687e-05 }, { "epoch": 0.3705987343826059, "step": 1142, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3109749488648959e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8267239283886738e-05 }, { "epoch": 0.3705987343826059, "step": 1142, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.77790046413429e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.526065716461744e-05 }, { "epoch": 0.3705987343826059, "step": 1142, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.726280061120633e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.883030381577555e-05 }, { "epoch": 0.37092325166315104, "step": 1143, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2962536857230589e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.822755984903779e-05 }, { "epoch": 0.37092325166315104, "step": 1143, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9287539544166066e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.074824988085311e-05 }, { "epoch": 0.37092325166315104, "step": 1143, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7799236957216635e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9997896743007004e-05 }, { "epoch": 0.37092325166315104, "step": 1143, "weight_max": 4.187939703115262e-05, "weight_mean": 6.646012025157688e-06, "weight_min": 0.0, "weight_std": 1.4773838302062359e-05 }, { "epoch": 0.37124776894369627, "step": 1144, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.469343806093093e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0320881958468817e-05 }, { "epoch": 0.37124776894369627, "step": 1144, "weight_max": 4.252122016623616e-05, "weight_mean": 9.56310577748809e-06, "weight_min": 0.0, "weight_std": 1.785282074706629e-05 }, { "epoch": 0.37124776894369627, "step": 1144, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3047141439747065e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9444258214207366e-05 }, { "epoch": 0.37124776894369627, "step": 1144, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4794699382036924e-05, "weight_min": 0.0, "weight_std": 2.0575143935275264e-05 }, { "epoch": 0.37157228622424143, "step": 1145, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4470884707407095e-05, "weight_min": 0.0, "weight_std": 2.005213355005253e-05 }, { "epoch": 0.37157228622424143, "step": 1145, "weight_max": 3.143761568935588e-05, "weight_mean": 1.039498056343291e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.4670100426883437e-05 }, { "epoch": 0.37157228622424143, "step": 1145, "weight_max": 3.722132169059478e-05, "weight_mean": 8.150974281306844e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5147373233048711e-05 }, { "epoch": 0.37157228622424143, "step": 1145, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.1704312530346215e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8842898498405702e-05 }, { "epoch": 0.37189680350478665, "step": 1146, "weight_max": 4.187939703115262e-05, "weight_mean": 8.735809387871996e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5645720850443467e-05 }, { "epoch": 0.37189680350478665, "step": 1146, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.091379989607958e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4351906429510564e-05 }, { "epoch": 0.37189680350478665, "step": 1146, "weight_max": 3.998322426923551e-05, "weight_mean": 2.4678854970261455e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.043704444076866e-05 }, { "epoch": 0.37189680350478665, "step": 1146, "weight_max": 3.3183479075660216e-08, "weight_mean": 4.808299092218249e-09, "weight_min": 0.0, "weight_std": 1.1482213402302932e-08 }, { "epoch": 0.3722213207853318, "step": 1147, "weight_max": 2.1563215341302566e-05, "weight_mean": 2.6989537218469195e-06, "weight_min": 6.053385570581327e-10, "weight_std": 7.622316843480803e-06 }, { "epoch": 0.3722213207853318, "step": 1147, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9651357433758676e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.119934833899606e-05 }, { "epoch": 0.3722213207853318, "step": 1147, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.390194125240669e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5542378605459817e-05 }, { "epoch": 0.3722213207853318, "step": 1147, "weight_max": 3.4558255720185116e-05, "weight_mean": 7.28059194443631e-06, "weight_min": 0.0, "weight_std": 1.3788332580588758e-05 }, { "epoch": 0.372545838065877, "step": 1148, "weight_max": 3.722132169059478e-05, "weight_mean": 8.44584883452626e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5744613847346045e-05 }, { "epoch": 0.372545838065877, "step": 1148, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998666099709226e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4135895980871283e-05 }, { "epoch": 0.372545838065877, "step": 1148, "weight_max": 4.252122016623616e-05, "weight_mean": 1.539338154543657e-05, "weight_min": 0.0, "weight_std": 2.127153675246518e-05 }, { "epoch": 0.372545838065877, "step": 1148, "weight_max": 4.252122016623616e-05, "weight_mean": 1.015860561892623e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.883588993223384e-05 }, { "epoch": 0.3728703553464222, "step": 1149, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3815396414429415e-05, "weight_min": 0.0, "weight_std": 1.9251589037594385e-05 }, { "epoch": 0.3728703553464222, "step": 1149, "weight_max": 2.674530374946471e-09, "weight_mean": 1.210059163980759e-09, "weight_min": 5.771870759119224e-10, "weight_std": 8.316269894947936e-10 }, { "epoch": 0.3728703553464222, "step": 1149, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3772344573226292e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9819388398900628e-05 }, { "epoch": 0.3728703553464222, "step": 1149, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.95810296444688e-06, "weight_min": 0.0, "weight_std": 1.8442660802975297e-05 }, { "epoch": 0.3731948726269674, "grad_norm": 2.6238369941711426, "learning_rate": 7.86530612244898e-06, "loss": 0.1719, "step": 1150 }, { "epoch": 0.3731948726269674, "step": 1150, "weight_max": 3.632264997577295e-05, "weight_mean": 9.00703435036121e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6677307939971797e-05 }, { "epoch": 0.3731948726269674, "step": 1150, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396534106694162e-06, "weight_min": 0.0, "weight_std": 1.5261945009115152e-05 }, { "epoch": 0.3731948726269674, "step": 1150, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.97418010229012e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.8549901142250746e-05 }, { "epoch": 0.3731948726269674, "step": 1150, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3425724066328257e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9766082914429717e-05 }, { "epoch": 0.3735193899075126, "step": 1151, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1022338842158206e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7892876712721772e-05 }, { "epoch": 0.3735193899075126, "step": 1151, "weight_max": 2.5141427613561973e-05, "weight_mean": 3.1432707601197762e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.888598131306935e-06 }, { "epoch": 0.3735193899075126, "step": 1151, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315746875567129e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033281670184806e-05 }, { "epoch": 0.3735193899075126, "step": 1151, "weight_max": 2.1598909825115697e-06, "weight_mean": 2.716399194468977e-07, "weight_min": 6.100955851628953e-10, "weight_std": 7.629709557477327e-07 }, { "epoch": 0.37384390718805777, "step": 1152, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232349268335383e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8945553165394813e-05 }, { "epoch": 0.37384390718805777, "step": 1152, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.693953764624894e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4608250239689369e-05 }, { "epoch": 0.37384390718805777, "step": 1152, "weight_max": 4.187939703115262e-05, "weight_mean": 9.984085409087129e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.846735176513903e-05 }, { "epoch": 0.37384390718805777, "step": 1152, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3853717064193916e-05, "weight_min": 0.0, "weight_std": 1.9308961782371625e-05 }, { "epoch": 0.37416842446860293, "step": 1153, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5036824152048212e-05, "weight_min": 0.0, "weight_std": 2.0768608010257594e-05 }, { "epoch": 0.37416842446860293, "step": 1153, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5308572983485647e-05, "weight_min": 0.0, "weight_std": 2.1127270883880556e-05 }, { "epoch": 0.37416842446860293, "step": 1153, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.300438144535292e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4477543118118774e-05 }, { "epoch": 0.37416842446860293, "step": 1153, "weight_max": 4.252122016623616e-05, "weight_mean": 1.963345130207017e-05, "weight_min": 0.0, "weight_std": 2.0570054402924143e-05 }, { "epoch": 0.37449294174914816, "step": 1154, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0246805433998816e-05, "weight_min": 0.0, "weight_std": 2.172709719161503e-05 }, { "epoch": 0.37449294174914816, "step": 1154, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.903416528482921e-06, "weight_min": 0.0, "weight_std": 1.66996251209639e-05 }, { "epoch": 0.37449294174914816, "step": 1154, "weight_max": 3.5141503758495674e-05, "weight_mean": 7.9842811828712e-06, "weight_min": 0.0, "weight_std": 1.4881814422551543e-05 }, { "epoch": 0.37449294174914816, "step": 1154, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4116310921963304e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.951164085767232e-05 }, { "epoch": 0.3748174590296933, "step": 1155, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2356340347614605e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.480631635786267e-05 }, { "epoch": 0.3748174590296933, "step": 1155, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805598564416869e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3588989531854168e-05 }, { "epoch": 0.3748174590296933, "step": 1155, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.357280431577237e-06, "weight_min": 0.0, "weight_std": 1.2203428013890516e-05 }, { "epoch": 0.3748174590296933, "step": 1155, "weight_max": 3.813097282545641e-05, "weight_mean": 4.766882284457097e-06, "weight_min": 0.0, "weight_std": 1.3481128917192109e-05 }, { "epoch": 0.37514197631023855, "step": 1156, "weight_max": 3.813097282545641e-05, "weight_mean": 1.531449379399419e-05, "weight_min": 0.0, "weight_std": 1.7369520719512366e-05 }, { "epoch": 0.37514197631023855, "step": 1156, "weight_max": 1.4685788585211412e-08, "weight_mean": 2.5148860771651016e-09, "weight_min": 6.053385570581327e-10, "weight_std": 4.923800922540522e-09 }, { "epoch": 0.37514197631023855, "step": 1156, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.439194875274552e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9930150301661342e-05 }, { "epoch": 0.37514197631023855, "step": 1156, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2519235396175645e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4799778000451624e-05 }, { "epoch": 0.3754664935907837, "step": 1157, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.943615294905612e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.3906822459830437e-05 }, { "epoch": 0.3754664935907837, "step": 1157, "weight_max": 3.4268457966391e-05, "weight_mean": 4.2849042074522e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.21151851999457e-05 }, { "epoch": 0.3754664935907837, "step": 1157, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843629994866205e-06, "weight_min": 0.0, "weight_std": 1.3697617760044523e-05 }, { "epoch": 0.3754664935907837, "step": 1157, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.183258963370463e-06, "weight_min": 0.0, "weight_std": 1.457032067264663e-05 }, { "epoch": 0.3757910108713289, "step": 1158, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3753577150055207e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9071994756814092e-05 }, { "epoch": 0.3757910108713289, "step": 1158, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.754485310812015e-05, "weight_min": 0.0, "weight_std": 1.9236533262301236e-05 }, { "epoch": 0.3757910108713289, "step": 1158, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4666600691271015e-05, "weight_min": 0.0, "weight_std": 2.0482919353526086e-05 }, { "epoch": 0.3757910108713289, "step": 1158, "weight_max": 3.4558255720185116e-05, "weight_mean": 7.432212441926822e-06, "weight_min": 0.0, "weight_std": 1.4000504052091856e-05 }, { "epoch": 0.3761155281518741, "step": 1159, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235523985902546e-06, "weight_min": 0.0, "weight_std": 1.4806360923103057e-05 }, { "epoch": 0.3761155281518741, "step": 1159, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392089279775973e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.924761636473704e-05 }, { "epoch": 0.3761155281518741, "step": 1159, "weight_max": 3.0886087188264355e-05, "weight_mean": 5.937558853474911e-06, "weight_min": 0.0, "weight_std": 1.16361552500166e-05 }, { "epoch": 0.3761155281518741, "step": 1159, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4976138118072413e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.075791053357534e-05 }, { "epoch": 0.37644004543241927, "grad_norm": 1.8805943727493286, "learning_rate": 7.844897959183674e-06, "loss": 0.0725, "step": 1160 }, { "epoch": 0.37644004543241927, "step": 1160, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3809446500090417e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.922202318382915e-05 }, { "epoch": 0.37644004543241927, "step": 1160, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235575372353196e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806340004724916e-05 }, { "epoch": 0.37644004543241927, "step": 1160, "weight_max": 3.967146767536178e-05, "weight_mean": 1.393758793710731e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.9305556634208187e-05 }, { "epoch": 0.37644004543241927, "step": 1160, "weight_max": 4.252122016623616e-05, "weight_mean": 8.874008017301094e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6853589841048233e-05 }, { "epoch": 0.3767645627129645, "step": 1161, "weight_max": 1.2699998386267453e-08, "weight_mean": 2.636266982491975e-09, "weight_min": 5.864964625068581e-10, "weight_std": 4.164575351239819e-09 }, { "epoch": 0.3767645627129645, "step": 1161, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2007307304884307e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6112360754050314e-05 }, { "epoch": 0.3767645627129645, "step": 1161, "weight_max": 1.0812275519356263e-07, "weight_mean": 1.4370120204887371e-08, "weight_min": 5.771870759119224e-10, "weight_std": 3.7885712345087086e-08 }, { "epoch": 0.3767645627129645, "step": 1161, "weight_max": 1.504423474685268e-09, "weight_mean": 6.213494163631594e-10, "weight_min": 0.0, "weight_std": 4.925203023198321e-10 }, { "epoch": 0.37708907999350966, "step": 1162, "weight_max": 3.722132169059478e-05, "weight_mean": 5.060512648924487e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3044576917309314e-05 }, { "epoch": 0.37708907999350966, "step": 1162, "weight_max": 3.397988257347606e-05, "weight_mean": 1.2476367373892572e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.5819412510609254e-05 }, { "epoch": 0.37708907999350966, "step": 1162, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2668998351728078e-05, "weight_min": 0.0, "weight_std": 1.8193437426816672e-05 }, { "epoch": 0.37708907999350966, "step": 1162, "weight_max": 3.486749999837002e-09, "weight_mean": 1.012288475266132e-09, "weight_min": 0.0, "weight_std": 1.119506931601677e-09 }, { "epoch": 0.3774135972740548, "step": 1163, "weight_max": 2.2961557988310233e-05, "weight_mean": 2.8708329864457482e-06, "weight_min": 6.053385570581327e-10, "weight_std": 8.117878678604029e-06 }, { "epoch": 0.3774135972740548, "step": 1163, "weight_max": 5.220130105954013e-08, "weight_mean": 8.338552959230583e-09, "weight_min": 6.29309993005478e-10, "weight_std": 1.8033070148248953e-08 }, { "epoch": 0.3774135972740548, "step": 1163, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.827459168969654e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6365995179512538e-05 }, { "epoch": 0.3774135972740548, "step": 1163, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.587228305463213e-06, "weight_min": 0.0, "weight_std": 1.7813548765843734e-05 }, { "epoch": 0.37773811455460005, "step": 1164, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.435069862054661e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7830599972512573e-05 }, { "epoch": 0.37773811455460005, "step": 1164, "weight_max": 4.252122016623616e-05, "weight_mean": 6.747221505065681e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4999889572209213e-05 }, { "epoch": 0.37773811455460005, "step": 1164, "weight_max": 3.312147964606993e-05, "weight_mean": 4.1557318581908476e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1704002645274159e-05 }, { "epoch": 0.37773811455460005, "step": 1164, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2091421012883075e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.834291651903186e-05 }, { "epoch": 0.3780626318351452, "step": 1165, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284284841560293e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2115435310988687e-05 }, { "epoch": 0.3780626318351452, "step": 1165, "weight_max": 3.632264997577295e-05, "weight_mean": 8.645538400742225e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.603430246177595e-05 }, { "epoch": 0.3780626318351452, "step": 1165, "weight_max": 3.722132169059478e-05, "weight_mean": 1.6114787285914645e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.756019628373906e-05 }, { "epoch": 0.3780626318351452, "step": 1165, "weight_max": 4.252122016623616e-05, "weight_mean": 1.402828456775751e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9652248738566414e-05 }, { "epoch": 0.37838714911569044, "step": 1166, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077242349216249e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4357566215039697e-05 }, { "epoch": 0.37838714911569044, "step": 1166, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.418331521563232e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9735045498237014e-05 }, { "epoch": 0.37838714911569044, "step": 1166, "weight_max": 3.967146767536178e-05, "weight_mean": 8.902583431336097e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.659650297369808e-05 }, { "epoch": 0.37838714911569044, "step": 1166, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1801006621681154e-05, "weight_min": 0.0, "weight_std": 1.8354081475990824e-05 }, { "epoch": 0.3787116663962356, "step": 1167, "weight_max": 4.187939703115262e-05, "weight_mean": 1.813800736272242e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.996045102714561e-05 }, { "epoch": 0.3787116663962356, "step": 1167, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4691949218104128e-05, "weight_min": 0.0, "weight_std": 2.031449184869416e-05 }, { "epoch": 0.3787116663962356, "step": 1167, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8297150745638646e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9803705072263256e-05 }, { "epoch": 0.3787116663962356, "step": 1167, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2897204214823432e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8004242519964464e-05 }, { "epoch": 0.37903618367678077, "step": 1168, "weight_max": 3.0886087188264355e-05, "weight_mean": 7.503641882067313e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3701942407351453e-05 }, { "epoch": 0.37903618367678077, "step": 1168, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.559201155032497e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.1534995539695956e-05 }, { "epoch": 0.37903618367678077, "step": 1168, "weight_max": 2.563926318543963e-05, "weight_mean": 3.205738494216348e-06, "weight_min": 6.196655966128617e-10, "weight_std": 9.064512596523855e-06 }, { "epoch": 0.37903618367678077, "step": 1168, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.311523172073066e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7266933355131187e-05 }, { "epoch": 0.379360700957326, "step": 1169, "weight_max": 2.846461757144425e-05, "weight_mean": 3.575373511921498e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0056881365017034e-05 }, { "epoch": 0.379360700957326, "step": 1169, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.640839041618165e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6001093172235414e-05 }, { "epoch": 0.379360700957326, "step": 1169, "weight_max": 4.381950566312298e-05, "weight_mean": 7.240108061523642e-06, "weight_min": 0.0, "weight_std": 1.5581505067530088e-05 }, { "epoch": 0.379360700957326, "step": 1169, "weight_max": 3.967146767536178e-05, "weight_mean": 9.46259569900576e-06, "weight_min": 0.0, "weight_std": 1.7547235984238796e-05 }, { "epoch": 0.37968521823787116, "grad_norm": 3.342078685760498, "learning_rate": 7.824489795918367e-06, "loss": 0.1259, "step": 1170 }, { "epoch": 0.37968521823787116, "step": 1170, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4538944014930166e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0106224837945774e-05 }, { "epoch": 0.37968521823787116, "step": 1170, "weight_max": 1.344265809777312e-09, "weight_mean": 8.00756239094369e-10, "weight_min": 5.771870759119224e-10, "weight_std": 2.3766652534895627e-10 }, { "epoch": 0.37968521823787116, "step": 1170, "weight_max": 4.187939703115262e-05, "weight_mean": 1.711695222184062e-05, "weight_min": 0.0, "weight_std": 1.8688135241973214e-05 }, { "epoch": 0.37968521823787116, "step": 1170, "weight_max": 3.998322426923551e-05, "weight_mean": 5.178139417694183e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4072114026930649e-05 }, { "epoch": 0.3800097355184164, "step": 1171, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235632670519408e-06, "weight_min": 0.0, "weight_std": 1.480631635786267e-05 }, { "epoch": 0.3800097355184164, "step": 1171, "weight_max": 4.187939703115262e-05, "weight_mean": 1.952283491846174e-05, "weight_min": 8.490471659072796e-10, "weight_std": 2.0428753487067297e-05 }, { "epoch": 0.3800097355184164, "step": 1171, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5258589883160312e-05, "weight_min": 0.0, "weight_std": 2.118270640494302e-05 }, { "epoch": 0.3800097355184164, "step": 1171, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4131285752227996e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9599934603320435e-05 }, { "epoch": 0.38033425279896155, "step": 1172, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2515871276264079e-05, "weight_min": 0.0, "weight_std": 1.8642413124325685e-05 }, { "epoch": 0.38033425279896155, "step": 1172, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998702024749946e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4135881428956054e-05 }, { "epoch": 0.38033425279896155, "step": 1172, "weight_max": 3.998322426923551e-05, "weight_mean": 5.335303740139352e-06, "weight_min": 0.0, "weight_std": 1.401314148097299e-05 }, { "epoch": 0.38033425279896155, "step": 1172, "weight_max": 3.4558255720185116e-05, "weight_mean": 8.607848030806053e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5419886040035635e-05 }, { "epoch": 0.3806587700795067, "step": 1173, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.4304210859991144e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2527725630206987e-05 }, { "epoch": 0.3806587700795067, "step": 1173, "weight_max": 4.187939703115262e-05, "weight_mean": 1.894514571176842e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0385492462082766e-05 }, { "epoch": 0.3806587700795067, "step": 1173, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4590157661587e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0235493138898164e-05 }, { "epoch": 0.3806587700795067, "step": 1173, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8842823919840157e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.03181534743635e-05 }, { "epoch": 0.38098328736005194, "step": 1174, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077190962765599e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357587133417837e-05 }, { "epoch": 0.38098328736005194, "step": 1174, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.937569873523898e-06, "weight_min": 0.0, "weight_std": 1.6750753275118768e-05 }, { "epoch": 0.38098328736005194, "step": 1174, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3470158592099324e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9026596419280395e-05 }, { "epoch": 0.38098328736005194, "step": 1174, "weight_max": 8.239723001679522e-07, "weight_mean": 1.0407222106323388e-07, "weight_min": 6.100955851628953e-10, "weight_std": 2.908842020588054e-07 }, { "epoch": 0.3813078046405971, "step": 1175, "weight_max": 3.116124207735993e-05, "weight_mean": 7.132474365789676e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.327908285020385e-05 }, { "epoch": 0.3813078046405971, "step": 1175, "weight_max": 2.9469142646121327e-06, "weight_mean": 3.689226559799863e-07, "weight_min": 0.0, "weight_std": 1.0416659961265395e-06 }, { "epoch": 0.3813078046405971, "step": 1175, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.468950358685106e-05, "weight_min": 0.0, "weight_std": 2.0296893126214854e-05 }, { "epoch": 0.3813078046405971, "step": 1175, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155970939085819e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4581142750103027e-05 }, { "epoch": 0.3816323219211423, "step": 1176, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.6061681637656875e-05, "weight_min": 0.0, "weight_std": 2.1609912437270395e-05 }, { "epoch": 0.3816323219211423, "step": 1176, "weight_max": 1.5802011921195458e-09, "weight_mean": 9.250438193220134e-10, "weight_min": 6.100955851628953e-10, "weight_std": 3.276001414587171e-10 }, { "epoch": 0.3816323219211423, "step": 1176, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311668120266404e-05, "weight_min": 0.0, "weight_std": 1.9095417883363552e-05 }, { "epoch": 0.3816323219211423, "step": 1176, "weight_max": 3.22740561387036e-05, "weight_mean": 4.035433903482044e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.141012671723729e-05 }, { "epoch": 0.3819568392016875, "step": 1177, "weight_max": 3.0886084800840763e-09, "weight_mean": 1.1235813390797489e-09, "weight_min": 5.958802340444436e-10, "weight_std": 8.32128033145807e-10 }, { "epoch": 0.3819568392016875, "step": 1177, "weight_max": 3.662098970380612e-05, "weight_mean": 1.6794887415017e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8038157577393577e-05 }, { "epoch": 0.3819568392016875, "step": 1177, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0472988833498675e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.940252877830062e-05 }, { "epoch": 0.3819568392016875, "step": 1177, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.08791298570577e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.721580520097632e-05 }, { "epoch": 0.38228135648223266, "step": 1178, "weight_max": 4.187939703115262e-05, "weight_mean": 9.776189472177066e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.8160506442654878e-05 }, { "epoch": 0.38228135648223266, "step": 1178, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316141596267698e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5033121599117294e-05 }, { "epoch": 0.38228135648223266, "step": 1178, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8056845874525607e-05, "weight_min": 0.0, "weight_std": 2.0255914932931773e-05 }, { "epoch": 0.38228135648223266, "step": 1178, "weight_max": 3.22740561387036e-05, "weight_mean": 9.442094778933097e-06, "weight_min": 0.0, "weight_std": 1.3703371223527938e-05 }, { "epoch": 0.3826058737627779, "step": 1179, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9364730330125894e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3909641893405933e-05 }, { "epoch": 0.3826058737627779, "step": 1179, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.284644258703338e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.802528822736349e-05 }, { "epoch": 0.3826058737627779, "step": 1179, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3454995496431366e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.880290437838994e-05 }, { "epoch": 0.3826058737627779, "step": 1179, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0194680726272054e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8884073142544366e-05 }, { "epoch": 0.38293039104332305, "grad_norm": 1.255717396736145, "learning_rate": 7.804081632653062e-06, "loss": 0.1329, "step": 1180 }, { "epoch": 0.38293039104332305, "step": 1180, "weight_max": 4.187939703115262e-05, "weight_mean": 1.562575198477134e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1565168935921974e-05 }, { "epoch": 0.38293039104332305, "step": 1180, "weight_max": 3.4849264920921996e-05, "weight_mean": 7.98091605247464e-06, "weight_min": 0.0, "weight_std": 1.4859213479212485e-05 }, { "epoch": 0.38293039104332305, "step": 1180, "weight_max": 2.7939937353949063e-05, "weight_mean": 3.493469421300688e-06, "weight_min": 0.0, "weight_std": 9.877864613372367e-06 }, { "epoch": 0.38293039104332305, "step": 1180, "weight_max": 4.187939703115262e-05, "weight_mean": 6.99764223099919e-06, "weight_min": 7.614652242082798e-10, "weight_std": 1.4932396879885346e-05 }, { "epoch": 0.38325490832386827, "step": 1181, "weight_max": 4.252122016623616e-05, "weight_mean": 9.385035809827968e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7579497580300085e-05 }, { "epoch": 0.38325490832386827, "step": 1181, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.847031494951807e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.6676418454153463e-05 }, { "epoch": 0.38325490832386827, "step": 1181, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.458334554627072e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0191382645862177e-05 }, { "epoch": 0.38325490832386827, "step": 1181, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1566475222934969e-05, "weight_min": 0.0, "weight_std": 1.8450027710059658e-05 }, { "epoch": 0.38357942560441344, "step": 1182, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076908564660698e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4357699910760857e-05 }, { "epoch": 0.38357942560441344, "step": 1182, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3466205928125419e-05, "weight_min": 0.0, "weight_std": 1.8891101717599668e-05 }, { "epoch": 0.38357942560441344, "step": 1182, "weight_max": 3.199401908204891e-05, "weight_mean": 3.999723503511632e-06, "weight_min": 0.0, "weight_std": 1.1311403795843944e-05 }, { "epoch": 0.38357942560441344, "step": 1182, "weight_max": 3.4849264920921996e-05, "weight_mean": 6.407889031834202e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.284658083022805e-05 }, { "epoch": 0.3839039428849586, "step": 1183, "weight_max": 6.345375709315704e-07, "weight_mean": 8.011026153553757e-08, "weight_min": 5.864964625068581e-10, "weight_std": 2.2402312538361002e-07 }, { "epoch": 0.3839039428849586, "step": 1183, "weight_max": 2.0431491066119634e-05, "weight_mean": 4.092080416739918e-06, "weight_min": 0.0, "weight_std": 7.88147281127749e-06 }, { "epoch": 0.3839039428849586, "step": 1183, "weight_max": 4.381950566312298e-05, "weight_mean": 9.582716302247718e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7983491488848813e-05 }, { "epoch": 0.3839039428849586, "step": 1183, "weight_max": 1.3050325264885032e-08, "weight_mean": 2.167544810305344e-09, "weight_min": 0.0, "weight_std": 4.419292487511939e-09 }, { "epoch": 0.3842284601655038, "step": 1184, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4947860108804889e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0513558411039412e-05 }, { "epoch": 0.3842284601655038, "step": 1184, "weight_max": 4.252122016623616e-05, "weight_mean": 1.000606789602898e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8574493878986686e-05 }, { "epoch": 0.3842284601655038, "step": 1184, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0871371159737464e-05, "weight_min": 0.0, "weight_std": 1.6150239389389753e-05 }, { "epoch": 0.3842284601655038, "step": 1184, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.252882591681555e-06, "weight_min": 0.0, "weight_std": 1.7238931832253e-05 }, { "epoch": 0.384552977446049, "step": 1185, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2515311027527787e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.7140917407232337e-05 }, { "epoch": 0.384552977446049, "step": 1185, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4042778275324963e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.941951813932974e-05 }, { "epoch": 0.384552977446049, "step": 1185, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1187361451447941e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7709282474243082e-05 }, { "epoch": 0.384552977446049, "step": 1185, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5096901734068524e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.093999501084909e-05 }, { "epoch": 0.3848774947265942, "step": 1186, "weight_max": 3.967146767536178e-05, "weight_mean": 1.5475097825401463e-05, "weight_min": 0.0, "weight_std": 1.864877231128048e-05 }, { "epoch": 0.3848774947265942, "step": 1186, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882013399765128e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3806600691168569e-05 }, { "epoch": 0.3848774947265942, "step": 1186, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.086954726924887e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4442131941905245e-05 }, { "epoch": 0.3848774947265942, "step": 1186, "weight_max": 4.187939703115262e-05, "weight_mean": 9.888063686958048e-06, "weight_min": 0.0, "weight_std": 1.8350245227338746e-05 }, { "epoch": 0.3852020120071394, "step": 1187, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.767908184381668e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.650166632316541e-05 }, { "epoch": 0.3852020120071394, "step": 1187, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0840147297130898e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9382114260224625e-05 }, { "epoch": 0.3852020120071394, "step": 1187, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861318418785231e-06, "weight_min": 0.0, "weight_std": 1.0919656233454589e-05 }, { "epoch": 0.3852020120071394, "step": 1187, "weight_max": 3.909020662717921e-09, "weight_mean": 1.2250785941247955e-09, "weight_min": 6.29309993005478e-10, "weight_std": 1.136230998177723e-09 }, { "epoch": 0.38552652928768455, "step": 1188, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.01664316188544e-06, "weight_min": 0.0, "weight_std": 1.69220566021977e-05 }, { "epoch": 0.38552652928768455, "step": 1188, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3795853192277718e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9150287698721513e-05 }, { "epoch": 0.38552652928768455, "step": 1188, "weight_max": 4.447596802492626e-05, "weight_mean": 1.497990706411656e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0782657884410582e-05 }, { "epoch": 0.38552652928768455, "step": 1188, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.367623533762526e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7460601156926714e-05 }, { "epoch": 0.38585104656822977, "step": 1189, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0090087926073465e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8734957848209888e-05 }, { "epoch": 0.38585104656822977, "step": 1189, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767153313878225e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3481018868333194e-05 }, { "epoch": 0.38585104656822977, "step": 1189, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8414066289551556e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.021898762905039e-05 }, { "epoch": 0.38585104656822977, "step": 1189, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396813321567606e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.5261832231772132e-05 }, { "epoch": 0.38617556384877494, "grad_norm": 0.1814596801996231, "learning_rate": 7.783673469387755e-06, "loss": 0.1154, "step": 1190 }, { "epoch": 0.38617556384877494, "step": 1190, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8258957425132394e-05, "weight_min": 7.829137893544669e-10, "weight_std": 1.9878596503986046e-05 }, { "epoch": 0.38617556384877494, "step": 1190, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3229418073024135e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.846989653131459e-05 }, { "epoch": 0.38617556384877494, "step": 1190, "weight_max": 3.4268457966391e-05, "weight_mean": 4.292716312193079e-06, "weight_min": 0.0, "weight_std": 1.2112051990698092e-05 }, { "epoch": 0.38617556384877494, "step": 1190, "weight_max": 3.752332122530788e-05, "weight_mean": 1.745949703035876e-05, "weight_min": 1.1273959543700585e-09, "weight_std": 1.517704367870465e-05 }, { "epoch": 0.38650008112932016, "step": 1191, "weight_max": 1.2475846233428456e-05, "weight_mean": 1.5606045735694352e-06, "weight_min": 0.0, "weight_std": 4.410424025991233e-06 }, { "epoch": 0.38650008112932016, "step": 1191, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4697907317895442e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.034803583228495e-05 }, { "epoch": 0.38650008112932016, "step": 1191, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0479997399670538e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.939821413543541e-05 }, { "epoch": 0.38650008112932016, "step": 1191, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8694536265684292e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0313622371759266e-05 }, { "epoch": 0.38682459840986533, "step": 1192, "weight_max": 3.813097282545641e-05, "weight_mean": 4.768508460983867e-06, "weight_min": 0.0, "weight_std": 1.3480472262017429e-05 }, { "epoch": 0.38682459840986533, "step": 1192, "weight_max": 4.187939703115262e-05, "weight_mean": 9.926152415573597e-06, "weight_min": 0.0, "weight_std": 1.8414495571050793e-05 }, { "epoch": 0.38682459840986533, "step": 1192, "weight_max": 3.998322426923551e-05, "weight_mean": 5.07482172906748e-06, "weight_min": 0.0, "weight_std": 1.4106735761743039e-05 }, { "epoch": 0.38682459840986533, "step": 1192, "weight_max": 4.252122016623616e-05, "weight_mean": 9.132447303272784e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7160420611617155e-05 }, { "epoch": 0.3871491156904105, "step": 1193, "weight_max": 4.252122016623616e-05, "weight_mean": 8.940977750171442e-06, "weight_min": 0.0, "weight_std": 1.6941867215791717e-05 }, { "epoch": 0.3871491156904105, "step": 1193, "weight_max": 3.998322426923551e-05, "weight_mean": 1.0078145351144485e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6646079529891722e-05 }, { "epoch": 0.3871491156904105, "step": 1193, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767286554852035e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.348096520814579e-05 }, { "epoch": 0.3871491156904105, "step": 1193, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5455337890889496e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6862346456036903e-05 }, { "epoch": 0.3874736329709557, "step": 1194, "weight_max": 3.967146767536178e-05, "weight_mean": 8.820217772154137e-06, "weight_min": 0.0, "weight_std": 1.6498508557560854e-05 }, { "epoch": 0.3874736329709557, "step": 1194, "weight_max": 3.967146767536178e-05, "weight_mean": 9.207324183080345e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7114152797148563e-05 }, { "epoch": 0.3874736329709557, "step": 1194, "weight_max": 3.843663216684945e-05, "weight_mean": 2.2091528080636635e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.852628520282451e-05 }, { "epoch": 0.3874736329709557, "step": 1194, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.8945610109949484e-06, "weight_min": 0.0, "weight_std": 1.4707485206599813e-05 }, { "epoch": 0.3877981502515009, "step": 1195, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1471674042695668e-05, "weight_min": 7.829137893544669e-10, "weight_std": 1.816427720768843e-05 }, { "epoch": 0.3877981502515009, "step": 1195, "weight_max": 1.1273959543700585e-09, "weight_mean": 6.778674288554498e-10, "weight_min": 0.0, "weight_std": 3.2321550991198933e-10 }, { "epoch": 0.3877981502515009, "step": 1195, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3712205145566259e-05, "weight_min": 0.0, "weight_std": 1.91695180546958e-05 }, { "epoch": 0.3877981502515009, "step": 1195, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0315994586562738e-05, "weight_min": 0.0, "weight_std": 1.9106299077975564e-05 }, { "epoch": 0.3881226675320461, "step": 1196, "weight_max": 3.22740561387036e-05, "weight_mean": 4.035353413200937e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1410159459046554e-05 }, { "epoch": 0.3881226675320461, "step": 1196, "weight_max": 3.255531191825867e-05, "weight_mean": 8.105525012069847e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5004127817519475e-05 }, { "epoch": 0.3881226675320461, "step": 1196, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0392044714535587e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.924764183058869e-05 }, { "epoch": 0.3881226675320461, "step": 1196, "weight_max": 2.4894337911973707e-05, "weight_mean": 3.112486638201517e-06, "weight_min": 0.0, "weight_std": 8.80119750945596e-06 }, { "epoch": 0.3884471848125913, "step": 1197, "weight_max": 3.632264997577295e-05, "weight_mean": 4.541112957667792e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2841680472774897e-05 }, { "epoch": 0.3884471848125913, "step": 1197, "weight_max": 4.187939703115262e-05, "weight_mean": 9.235020115738735e-06, "weight_min": 0.0, "weight_std": 1.7300770196015947e-05 }, { "epoch": 0.3884471848125913, "step": 1197, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3795420272799674e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.914970198413357e-05 }, { "epoch": 0.3884471848125913, "step": 1197, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882171651843237e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3806536117044743e-05 }, { "epoch": 0.38877170209313644, "step": 1198, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.8435331336804666e-06, "weight_min": 0.0, "weight_std": 1.36976577778114e-05 }, { "epoch": 0.38877170209313644, "step": 1198, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8795526557369158e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0649045836762525e-05 }, { "epoch": 0.38877170209313644, "step": 1198, "weight_max": 4.187939703115262e-05, "weight_mean": 1.403329679305898e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9526945834513754e-05 }, { "epoch": 0.38877170209313644, "step": 1198, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.26150944197434e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.829173015721608e-05 }, { "epoch": 0.38909621937368166, "step": 1199, "weight_max": 3.006794759130571e-05, "weight_mean": 3.758967068279162e-06, "weight_min": 0.0, "weight_std": 1.063043328031199e-05 }, { "epoch": 0.38909621937368166, "step": 1199, "weight_max": 3.722132169059478e-05, "weight_mean": 8.411599992541596e-06, "weight_min": 0.0, "weight_std": 1.5691071894252673e-05 }, { "epoch": 0.38909621937368166, "step": 1199, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.7507463780930266e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8908378478954546e-05 }, { "epoch": 0.38909621937368166, "step": 1199, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920725586998742e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3915942872699816e-05 }, { "epoch": 0.38942073665422683, "grad_norm": 3.4759020805358887, "learning_rate": 7.763265306122449e-06, "loss": 0.1587, "step": 1200 }, { "epoch": 0.38942073665422683, "eval_accuracy": 0.923, "eval_auc": 0.9925439263253908, "eval_loss": 0.32146748900413513, "eval_runtime": 156.1018, "eval_samples_per_second": 12.812, "eval_steps_per_second": 1.602, "step": 1200 }, { "epoch": 0.38942073665422683, "step": 1200, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471085261087865e-05, "weight_min": 0.0, "weight_std": 1.939015055540949e-05 }, { "epoch": 0.38942073665422683, "step": 1200, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578258995024953e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.2947218237968627e-05 }, { "epoch": 0.38942073665422683, "step": 1200, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.311421308666468e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7266995200770907e-05 }, { "epoch": 0.38942073665422683, "step": 1200, "weight_max": 4.252122016623616e-05, "weight_mean": 8.129333764372859e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5969573723850772e-05 }, { "epoch": 0.38974525393477205, "step": 1201, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0780418961076066e-05, "weight_min": 0.0, "weight_std": 1.811748734326102e-05 }, { "epoch": 0.38974525393477205, "step": 1201, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.7445536185696255e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5152545529417694e-05 }, { "epoch": 0.38974525393477205, "step": 1201, "weight_max": 4.187939703115262e-05, "weight_mean": 6.315856808214448e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4684324014524464e-05 }, { "epoch": 0.38974525393477205, "step": 1201, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3717064575757831e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.902441908896435e-05 }, { "epoch": 0.3900697712153172, "step": 1202, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0632124940457288e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.968689502973575e-05 }, { "epoch": 0.3900697712153172, "step": 1202, "weight_max": 3.782653220696375e-05, "weight_mean": 8.693486961419694e-06, "weight_min": 0.0, "weight_std": 1.6177971701836213e-05 }, { "epoch": 0.3900697712153172, "step": 1202, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2892485756310634e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.7980572010856122e-05 }, { "epoch": 0.3900697712153172, "step": 1202, "weight_max": 3.722132169059478e-05, "weight_mean": 1.2425796739989892e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7321290215477347e-05 }, { "epoch": 0.3903942884958624, "step": 1203, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4277491320390254e-05, "weight_min": 2.021574685429073e-09, "weight_std": 2.020082320086658e-05 }, { "epoch": 0.3903942884958624, "step": 1203, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1592134796956088e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.820503530325368e-05 }, { "epoch": 0.3903942884958624, "step": 1203, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1668804290820844e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.6871334082679823e-05 }, { "epoch": 0.3903942884958624, "step": 1203, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8392303900327533e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.984351547434926e-05 }, { "epoch": 0.3907188057764076, "step": 1204, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4267917322285939e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9876652004313655e-05 }, { "epoch": 0.3907188057764076, "step": 1204, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235609478404513e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.480632636230439e-05 }, { "epoch": 0.3907188057764076, "step": 1204, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235785010881955e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4806255421717651e-05 }, { "epoch": 0.3907188057764076, "step": 1204, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077391051599989e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357505278894678e-05 }, { "epoch": 0.3910433230569528, "step": 1205, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578078915074002e-06, "weight_min": 0.0, "weight_std": 1.294729099754477e-05 }, { "epoch": 0.3910433230569528, "step": 1205, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.92392233456485e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.838880780269392e-05 }, { "epoch": 0.3910433230569528, "step": 1205, "weight_max": 3.22740561387036e-05, "weight_mean": 7.793312761350535e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4441065104620066e-05 }, { "epoch": 0.3910433230569528, "step": 1205, "weight_max": 3.3513551933594954e-09, "weight_mean": 1.250951453535265e-09, "weight_min": 6.053385570581327e-10, "weight_std": 9.739662409913308e-10 }, { "epoch": 0.391367840337498, "step": 1206, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248090135661187e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.20134591270471e-05 }, { "epoch": 0.391367840337498, "step": 1206, "weight_max": 4.381950566312298e-05, "weight_mean": 1.505033560533775e-05, "weight_min": 0.0, "weight_std": 2.1001716959290206e-05 }, { "epoch": 0.391367840337498, "step": 1206, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4879278751323e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0538973330985755e-05 }, { "epoch": 0.391367840337498, "step": 1206, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.0851508452324197e-05, "weight_min": 0.0, "weight_std": 1.8498587451176718e-05 }, { "epoch": 0.39169235761804316, "step": 1207, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2356799642439e-06, "weight_min": 0.0, "weight_std": 1.4806297258473933e-05 }, { "epoch": 0.39169235761804316, "step": 1207, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.0594701961963437e-05, "weight_min": 0.0, "weight_std": 1.545838858874049e-05 }, { "epoch": 0.39169235761804316, "step": 1207, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998472377337748e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4135975106910337e-05 }, { "epoch": 0.39169235761804316, "step": 1207, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5389363397844136e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1238893168629147e-05 }, { "epoch": 0.39201687489858833, "step": 1208, "weight_max": 4.252122016623616e-05, "weight_mean": 6.919435691088438e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.506919306848431e-05 }, { "epoch": 0.39201687489858833, "step": 1208, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070189788762946e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1509727301017847e-05 }, { "epoch": 0.39201687489858833, "step": 1208, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.583292583352886e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6315727407345548e-05 }, { "epoch": 0.39201687489858833, "step": 1208, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315778707881691e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5033268937258981e-05 }, { "epoch": 0.39234139217913355, "step": 1209, "weight_max": 3.116124207735993e-05, "weight_mean": 3.896021098626079e-06, "weight_min": 0.0, "weight_std": 1.1016813004971482e-05 }, { "epoch": 0.39234139217913355, "step": 1209, "weight_max": 3.312147964606993e-05, "weight_mean": 6.229463906493038e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.2320002497290261e-05 }, { "epoch": 0.39234139217913355, "step": 1209, "weight_max": 3.662098970380612e-05, "weight_mean": 1.695578612270765e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8278078641742468e-05 }, { "epoch": 0.39234139217913355, "step": 1209, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5398734831251204e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1295058104442433e-05 }, { "epoch": 0.3926659094596787, "grad_norm": 2.711230754852295, "learning_rate": 7.742857142857144e-06, "loss": 0.0933, "step": 1210 }, { "epoch": 0.3926659094596787, "step": 1210, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396741016738815e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.526186133560259e-05 }, { "epoch": 0.3926659094596787, "step": 1210, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2876200344180688e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.858153882494662e-05 }, { "epoch": 0.3926659094596787, "step": 1210, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3136334018781781e-05, "weight_min": 0.0, "weight_std": 1.711463846731931e-05 }, { "epoch": 0.3926659094596787, "step": 1210, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3961458535050042e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.943841198226437e-05 }, { "epoch": 0.39299042674022394, "step": 1211, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.780079951975495e-06, "weight_min": 0.0, "weight_std": 1.6582627722527832e-05 }, { "epoch": 0.39299042674022394, "step": 1211, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396626875153743e-06, "weight_min": 0.0, "weight_std": 1.5261906810337678e-05 }, { "epoch": 0.39299042674022394, "step": 1211, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.8663979062694125e-05, "weight_min": 0.0, "weight_std": 2.0015144400531426e-05 }, { "epoch": 0.39299042674022394, "step": 1211, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0520772523013875e-05, "weight_min": 9.178617865757133e-10, "weight_std": 1.851262823038269e-05 }, { "epoch": 0.3933149440207691, "step": 1212, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6685464288457297e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.833275928220246e-05 }, { "epoch": 0.3933149440207691, "step": 1212, "weight_max": 3.843663216684945e-05, "weight_mean": 1.4076288607611787e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9429069652687758e-05 }, { "epoch": 0.3933149440207691, "step": 1212, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.47960208778386e-05, "weight_min": 0.0, "weight_std": 1.9545721443137154e-05 }, { "epoch": 0.3933149440207691, "step": 1212, "weight_max": 4.252122016623616e-05, "weight_mean": 1.904966848087497e-05, "weight_min": 0.0, "weight_std": 2.046124791377224e-05 }, { "epoch": 0.3936394613013143, "step": 1213, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235727257968392e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806279068579897e-05 }, { "epoch": 0.3936394613013143, "step": 1213, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236120159679558e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4806119907007087e-05 }, { "epoch": 0.3936394613013143, "step": 1213, "weight_max": 1.3301527657105794e-09, "weight_mean": 8.856397837320173e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.737124138452174e-10 }, { "epoch": 0.3936394613013143, "step": 1213, "weight_max": 3.0339444492710754e-05, "weight_mean": 9.068163308256771e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.3037882126809563e-05 }, { "epoch": 0.3939639785818595, "step": 1214, "weight_max": 4.252122016623616e-05, "weight_mean": 9.674407010606956e-06, "weight_min": 0.0, "weight_std": 1.802289261831902e-05 }, { "epoch": 0.3939639785818595, "step": 1214, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.6389654926315416e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3043884791841265e-05 }, { "epoch": 0.3939639785818595, "step": 1214, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930620550818276e-06, "weight_min": 0.0, "weight_std": 1.1114504559373017e-05 }, { "epoch": 0.3939639785818595, "step": 1214, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.579871402820572e-06, "weight_min": 0.0, "weight_std": 1.7779813788365573e-05 }, { "epoch": 0.39428849586240466, "step": 1215, "weight_max": 4.252122016623616e-05, "weight_mean": 1.481226172472816e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.049733120657038e-05 }, { "epoch": 0.39428849586240466, "step": 1215, "weight_max": 3.843663216684945e-05, "weight_mean": 9.345778380520642e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7312097043031827e-05 }, { "epoch": 0.39428849586240466, "step": 1215, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.8858331613009796e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4212439054972492e-05 }, { "epoch": 0.39428849586240466, "step": 1215, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315945600159466e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5033200725156348e-05 }, { "epoch": 0.3946130131429499, "step": 1216, "weight_max": 3.782653220696375e-05, "weight_mean": 6.359928192978259e-06, "weight_min": 0.0, "weight_std": 1.3113495697325561e-05 }, { "epoch": 0.3946130131429499, "step": 1216, "weight_max": 3.998322426923551e-05, "weight_mean": 1.0007978744397406e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8501150407246314e-05 }, { "epoch": 0.3946130131429499, "step": 1216, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316259375831578e-06, "weight_min": 0.0, "weight_std": 1.5033074305392802e-05 }, { "epoch": 0.3946130131429499, "step": 1216, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.8387629501812626e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.070863800123334e-05 }, { "epoch": 0.39493753042349505, "step": 1217, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.487940699007595e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0542860511341132e-05 }, { "epoch": 0.39493753042349505, "step": 1217, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550953447818756e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9535160390660167e-05 }, { "epoch": 0.39493753042349505, "step": 1217, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3159446906647645e-06, "weight_min": 0.0, "weight_std": 1.503320163465105e-05 }, { "epoch": 0.39493753042349505, "step": 1217, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0115109944308642e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8732105672825128e-05 }, { "epoch": 0.3952620477040402, "step": 1218, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3537770428229123e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.896128560474608e-05 }, { "epoch": 0.3952620477040402, "step": 1218, "weight_max": 4.187939703115262e-05, "weight_mean": 1.835398688854184e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.995189450099133e-05 }, { "epoch": 0.3952620477040402, "step": 1218, "weight_max": 4.187939703115262e-05, "weight_mean": 9.925785889208782e-06, "weight_min": 0.0, "weight_std": 1.841471930674743e-05 }, { "epoch": 0.3952620477040402, "step": 1218, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844883733312599e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.3697112990485039e-05 }, { "epoch": 0.39558656498458544, "step": 1219, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.692057574284263e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.797157892724499e-05 }, { "epoch": 0.39558656498458544, "step": 1219, "weight_max": 4.187939703115262e-05, "weight_mean": 1.928805431816727e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0706471332232468e-05 }, { "epoch": 0.39558656498458544, "step": 1219, "weight_max": 3.813097282545641e-05, "weight_mean": 9.345162652607542e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7305703295278363e-05 }, { "epoch": 0.39558656498458544, "step": 1219, "weight_max": 3.340639523230493e-05, "weight_mean": 1.4499544704449363e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.6004165445338003e-05 }, { "epoch": 0.3959110822651306, "grad_norm": 7.597581386566162, "learning_rate": 7.722448979591837e-06, "loss": 0.1778, "step": 1220 }, { "epoch": 0.3959110822651306, "step": 1220, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4295799701358192e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.986496317840647e-05 }, { "epoch": 0.3959110822651306, "step": 1220, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0711898539739195e-05, "weight_min": 0.0, "weight_std": 1.9833469195873477e-05 }, { "epoch": 0.3959110822651306, "step": 1220, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.0967541786376387e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8716018530540168e-05 }, { "epoch": 0.3959110822651306, "step": 1220, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.883062213892117e-05, "weight_min": 7.721521200210191e-10, "weight_std": 2.0254412447684444e-05 }, { "epoch": 0.39623559954567583, "step": 1221, "weight_max": 4.252122016623616e-05, "weight_mean": 1.794425224943552e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9651422917377204e-05 }, { "epoch": 0.39623559954567583, "step": 1221, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3975011825095862e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9032046111533418e-05 }, { "epoch": 0.39623559954567583, "step": 1221, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471024324942846e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9390188754186966e-05 }, { "epoch": 0.39623559954567583, "step": 1221, "weight_max": 2.7797480015578913e-06, "weight_mean": 3.492361315693415e-07, "weight_min": 0.0, "weight_std": 9.8207726750843e-07 }, { "epoch": 0.396560116826221, "step": 1222, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4800554708926938e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.813948074413929e-05 }, { "epoch": 0.396560116826221, "step": 1222, "weight_max": 4.187939703115262e-05, "weight_mean": 5.237805453361943e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4805441423959564e-05 }, { "epoch": 0.396560116826221, "step": 1222, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9338371203048155e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0991379642509855e-05 }, { "epoch": 0.396560116826221, "step": 1222, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393918970890809e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2423901353031397e-05 }, { "epoch": 0.39688463410676617, "step": 1223, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.082148608285934e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6817721188999712e-05 }, { "epoch": 0.39688463410676617, "step": 1223, "weight_max": 4.252122016623616e-05, "weight_mean": 9.671792213339359e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8024495147983544e-05 }, { "epoch": 0.39688463410676617, "step": 1223, "weight_max": 4.4065102677848245e-09, "weight_mean": 1.3036782764430654e-09, "weight_min": 0.0, "weight_std": 1.3236937101979152e-09 }, { "epoch": 0.39688463410676617, "step": 1223, "weight_max": 4.187939703115262e-05, "weight_mean": 1.039110793499276e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9239183529862203e-05 }, { "epoch": 0.3972091513873114, "step": 1224, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.461789543100167e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0224249965394847e-05 }, { "epoch": 0.3972091513873114, "step": 1224, "weight_max": 1.0954876415780745e-05, "weight_mean": 1.3701487660000566e-06, "weight_min": 6.100955851628953e-10, "weight_std": 3.872814431815641e-06 }, { "epoch": 0.3972091513873114, "step": 1224, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.853399589890614e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.010501157201361e-05 }, { "epoch": 0.3972091513873114, "step": 1224, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.257559961231891e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7850978110800497e-05 }, { "epoch": 0.39753366866785655, "step": 1225, "weight_max": 3.967146767536178e-05, "weight_mean": 1.963369140867144e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.769143455021549e-05 }, { "epoch": 0.39753366866785655, "step": 1225, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2583705029101111e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7762702555046417e-05 }, { "epoch": 0.39753366866785655, "step": 1225, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155931830813643e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4581159121007659e-05 }, { "epoch": 0.39753366866785655, "step": 1225, "weight_max": 3.0339444492710754e-05, "weight_mean": 1.0876641681534238e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.5024052117951214e-05 }, { "epoch": 0.3978581859484018, "step": 1226, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.497059181740042e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7593023585504852e-05 }, { "epoch": 0.3978581859484018, "step": 1226, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4441988241742365e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0049130398547277e-05 }, { "epoch": 0.3978581859484018, "step": 1226, "weight_max": 3.662098970380612e-05, "weight_mean": 9.805925401451532e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5603734937030822e-05 }, { "epoch": 0.3978581859484018, "step": 1226, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0239542461931705e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.899527524074074e-05 }, { "epoch": 0.39818270322894694, "step": 1227, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5140918549150229e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.102101279888302e-05 }, { "epoch": 0.39818270322894694, "step": 1227, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.69186112342868e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.797170079953503e-05 }, { "epoch": 0.39818270322894694, "step": 1227, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.998219158849679e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.1622470740112476e-05 }, { "epoch": 0.39818270322894694, "step": 1227, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.843852240010165e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.823640377551783e-05 }, { "epoch": 0.3985072205094921, "step": 1228, "weight_max": 3.4558255720185116e-05, "weight_mean": 1.530410736449994e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.6625488569843583e-05 }, { "epoch": 0.3985072205094921, "step": 1228, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5468696801690385e-05, "weight_min": 0.0, "weight_std": 2.1358864614740014e-05 }, { "epoch": 0.3985072205094921, "step": 1228, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.85127849667333e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5458570487680845e-05 }, { "epoch": 0.3985072205094921, "step": 1228, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.193261576001532e-06, "weight_min": 0.0, "weight_std": 1.7021395251504146e-05 }, { "epoch": 0.39883173779003733, "step": 1229, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.856552808021661e-05, "weight_min": 0.0, "weight_std": 2.0131761630182154e-05 }, { "epoch": 0.39883173779003733, "step": 1229, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0394314813311212e-05, "weight_min": 0.0, "weight_std": 1.9264301954535767e-05 }, { "epoch": 0.39883173779003733, "step": 1229, "weight_max": 2.021574685429073e-09, "weight_mean": 9.968151859496288e-10, "weight_min": 5.958802340444436e-10, "weight_std": 4.550378407408573e-10 }, { "epoch": 0.39883173779003733, "step": 1229, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653676114685368e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3159316040400881e-05 }, { "epoch": 0.3991562550705825, "grad_norm": 3.018324851989746, "learning_rate": 7.702040816326532e-06, "loss": 0.1191, "step": 1230 }, { "epoch": 0.3991562550705825, "step": 1230, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0696556273614988e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7367874534102157e-05 }, { "epoch": 0.3991562550705825, "step": 1230, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0159295672783628e-05, "weight_min": 7.508524357824342e-10, "weight_std": 2.1596133592538536e-05 }, { "epoch": 0.3991562550705825, "step": 1230, "weight_max": 3.782653220696375e-05, "weight_mean": 4.730441560241161e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3372840840020217e-05 }, { "epoch": 0.3991562550705825, "step": 1230, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0018159628089052e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.7988977560889907e-05 }, { "epoch": 0.3994807723511277, "step": 1231, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5414137780899182e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.1179759642109275e-05 }, { "epoch": 0.3994807723511277, "step": 1231, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8935023035737686e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0482319087022915e-05 }, { "epoch": 0.3994807723511277, "step": 1231, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.27497967495583e-06, "weight_min": 0.0, "weight_std": 1.7203617971972562e-05 }, { "epoch": 0.3994807723511277, "step": 1231, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076914931123611e-06, "weight_min": 0.0, "weight_std": 1.4357698091771454e-05 }, { "epoch": 0.3998052896316729, "step": 1232, "weight_max": 3.998322426923551e-05, "weight_mean": 2.2042540877009742e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.857951247075107e-05 }, { "epoch": 0.3998052896316729, "step": 1232, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5037052435218357e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0766101442859508e-05 }, { "epoch": 0.3998052896316729, "step": 1232, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.733557817526162e-06, "weight_min": 0.0, "weight_std": 1.806377804314252e-05 }, { "epoch": 0.3998052896316729, "step": 1232, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.1050321518268902e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.689166310825385e-05 }, { "epoch": 0.40012980691221806, "step": 1233, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4443756299442612e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.00583253899822e-05 }, { "epoch": 0.40012980691221806, "step": 1233, "weight_max": 4.252122016623616e-05, "weight_mean": 5.348294962459477e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5020346836536191e-05 }, { "epoch": 0.40012980691221806, "step": 1233, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5946794519550167e-05, "weight_min": 0.0, "weight_std": 2.200975541200023e-05 }, { "epoch": 0.40012980691221806, "step": 1233, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5709616491221823e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.169873368984554e-05 }, { "epoch": 0.4004543241927633, "step": 1234, "weight_max": 3.4849264920921996e-05, "weight_mean": 5.8091072787647136e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2418297046679072e-05 }, { "epoch": 0.4004543241927633, "step": 1234, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9900038751075044e-05, "weight_min": 2.674530374946471e-09, "weight_std": 2.107819091179408e-05 }, { "epoch": 0.4004543241927633, "step": 1234, "weight_max": 4.187939703115262e-05, "weight_mean": 1.848107422119938e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.014126039284747e-05 }, { "epoch": 0.4004543241927633, "step": 1234, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9595637392485514e-06, "weight_min": 0.0, "weight_std": 1.4025727068656124e-05 }, { "epoch": 0.40077884147330844, "step": 1235, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.003721081360709e-05, "weight_min": 0.0, "weight_std": 1.8593407730804756e-05 }, { "epoch": 0.40077884147330844, "step": 1235, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3402599506662227e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.924070602399297e-05 }, { "epoch": 0.40077884147330844, "step": 1235, "weight_max": 3.782653220696375e-05, "weight_mean": 5.453822268464137e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3237014172773343e-05 }, { "epoch": 0.40077884147330844, "step": 1235, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0768271648848895e-06, "weight_min": 0.0, "weight_std": 1.4357733562064823e-05 }, { "epoch": 0.40110335875385367, "step": 1236, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.464862159802578e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7515378203825094e-05 }, { "epoch": 0.40110335875385367, "step": 1236, "weight_max": 3.782653220696375e-05, "weight_mean": 8.288194294436835e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5544994312222116e-05 }, { "epoch": 0.40110335875385367, "step": 1236, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3343423233891372e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8940623704111204e-05 }, { "epoch": 0.40110335875385367, "step": 1236, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3236093764135148e-05, "weight_min": 0.0, "weight_std": 1.8356446162215434e-05 }, { "epoch": 0.40142787603439883, "step": 1237, "weight_max": 4.381950566312298e-05, "weight_mean": 8.173549758794252e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.6259793483186513e-05 }, { "epoch": 0.40142787603439883, "step": 1237, "weight_max": 3.662098970380612e-05, "weight_mean": 1.4869693586661015e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.6525955288670957e-05 }, { "epoch": 0.40142787603439883, "step": 1237, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3636449693876784e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9189117665519007e-05 }, { "epoch": 0.40142787603439883, "step": 1237, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9051352865062654e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0468925868044607e-05 }, { "epoch": 0.401752393314944, "step": 1238, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077161404187791e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.435759895684896e-05 }, { "epoch": 0.401752393314944, "step": 1238, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0321045010641683e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9157305359840393e-05 }, { "epoch": 0.401752393314944, "step": 1238, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.793050796550233e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9290857380838133e-05 }, { "epoch": 0.401752393314944, "step": 1238, "weight_max": 3.397988257347606e-05, "weight_mean": 8.212422471842729e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.521717058494687e-05 }, { "epoch": 0.4020769105954892, "step": 1239, "weight_max": 3.967146767536178e-05, "weight_mean": 1.733703538775444e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.877304384834133e-05 }, { "epoch": 0.4020769105954892, "step": 1239, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.787732778931968e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.7314872820861638e-05 }, { "epoch": 0.4020769105954892, "step": 1239, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8447826732881367e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0111514459131286e-05 }, { "epoch": 0.4020769105954892, "step": 1239, "weight_max": 4.252122016623616e-05, "weight_mean": 1.323199558100896e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.891511601570528e-05 }, { "epoch": 0.4024014278760344, "grad_norm": 0.1553173065185547, "learning_rate": 7.681632653061225e-06, "loss": 0.0943, "step": 1240 }, { "epoch": 0.4024014278760344, "step": 1240, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5043408893689048e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0804594896617346e-05 }, { "epoch": 0.4024014278760344, "step": 1240, "weight_max": 2.820166446326766e-05, "weight_mean": 3.6401957004272845e-06, "weight_min": 0.0, "weight_std": 9.929501175065525e-06 }, { "epoch": 0.4024014278760344, "step": 1240, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0422661034681369e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.857694769569207e-05 }, { "epoch": 0.4024014278760344, "step": 1240, "weight_max": 5.490859766865697e-09, "weight_mean": 1.3176190138963761e-09, "weight_min": 0.0, "weight_std": 1.7367891569364247e-09 }, { "epoch": 0.4027259451565796, "step": 1241, "weight_max": 4.252122016623616e-05, "weight_mean": 1.784601408871822e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.973506914509926e-05 }, { "epoch": 0.4027259451565796, "step": 1241, "weight_max": 5.548205450622845e-09, "weight_mean": 1.337596811090691e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.7078574110485079e-09 }, { "epoch": 0.4027259451565796, "step": 1241, "weight_max": 4.252122016623616e-05, "weight_mean": 1.015893849398708e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.883568438643124e-05 }, { "epoch": 0.4027259451565796, "step": 1241, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5075956980581395e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0819541532546282e-05 }, { "epoch": 0.4030504624371248, "step": 1242, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.696171218820382e-06, "weight_min": 0.0, "weight_std": 1.8000591808231547e-05 }, { "epoch": 0.4030504624371248, "step": 1242, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393569270177977e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2424042324710172e-05 }, { "epoch": 0.4030504624371248, "step": 1242, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.85273565875832e-05, "weight_min": 7.829137893544669e-10, "weight_std": 1.9863693523802795e-05 }, { "epoch": 0.4030504624371248, "step": 1242, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5230167264235206e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.102087455568835e-05 }, { "epoch": 0.40337497971766995, "step": 1243, "weight_max": 3.4849264920921996e-05, "weight_mean": 5.645480996463448e-06, "weight_min": 0.0, "weight_std": 1.2339301065367181e-05 }, { "epoch": 0.40337497971766995, "step": 1243, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5508940123254433e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.848231113399379e-05 }, { "epoch": 0.40337497971766995, "step": 1243, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0792675311677158e-05, "weight_min": 0.0, "weight_std": 1.9982437152066268e-05 }, { "epoch": 0.40337497971766995, "step": 1243, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.996821063396055e-06, "weight_min": 0.0, "weight_std": 1.8512528185965493e-05 }, { "epoch": 0.40369949699821517, "step": 1244, "weight_max": 4.252122016623616e-05, "weight_mean": 9.572147973813117e-06, "weight_min": 0.0, "weight_std": 1.694236198090948e-05 }, { "epoch": 0.40369949699821517, "step": 1244, "weight_max": 3.782653220696375e-05, "weight_mean": 5.979266461508814e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.333496857114369e-05 }, { "epoch": 0.40369949699821517, "step": 1244, "weight_max": 3.998322426923551e-05, "weight_mean": 9.996432709158398e-06, "weight_min": 0.0, "weight_std": 1.8508249922888353e-05 }, { "epoch": 0.40369949699821517, "step": 1244, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3992167623655405e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8507635104469955e-05 }, { "epoch": 0.40402401427876034, "step": 1245, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921337222185684e-06, "weight_min": 7.298507354036587e-10, "weight_std": 1.3915695490140934e-05 }, { "epoch": 0.40402401427876034, "step": 1245, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959437774232356e-06, "weight_min": 0.0, "weight_std": 1.4025778000359423e-05 }, { "epoch": 0.40402401427876034, "step": 1245, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.986609277082607e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1506017219508067e-05 }, { "epoch": 0.40402401427876034, "step": 1245, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.2392596292775124e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9105713363387622e-05 }, { "epoch": 0.40434853155930556, "step": 1246, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0121035302290693e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8769071175483987e-05 }, { "epoch": 0.40434853155930556, "step": 1246, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.011483709589811e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8732274838839658e-05 }, { "epoch": 0.40434853155930556, "step": 1246, "weight_max": 3.967146767536178e-05, "weight_mean": 9.87946714303689e-06, "weight_min": 0.0, "weight_std": 1.8292374079464935e-05 }, { "epoch": 0.40434853155930556, "step": 1246, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4196864867699333e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9689718101290055e-05 }, { "epoch": 0.4046730488398507, "step": 1247, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.805376976146363e-06, "weight_min": 0.0, "weight_std": 1.8169394024880603e-05 }, { "epoch": 0.4046730488398507, "step": 1247, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827252385235624e-06, "weight_min": 0.0, "weight_std": 1.0822735930560157e-05 }, { "epoch": 0.4046730488398507, "step": 1247, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4917141015757807e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0589877749443986e-05 }, { "epoch": 0.4046730488398507, "step": 1247, "weight_max": 4.252122016623616e-05, "weight_mean": 1.039218477671966e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9247558157076128e-05 }, { "epoch": 0.4049975661203959, "step": 1248, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.545123248244636e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7715758076519705e-05 }, { "epoch": 0.4049975661203959, "step": 1248, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3064543963992037e-05, "weight_min": 0.0, "weight_std": 1.7564307199791074e-05 }, { "epoch": 0.4049975661203959, "step": 1248, "weight_max": 4.252122016623616e-05, "weight_mean": 9.385251360072289e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.757936661306303e-05 }, { "epoch": 0.4049975661203959, "step": 1248, "weight_max": 4.252122016623616e-05, "weight_mean": 1.912074367282912e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0663301256718114e-05 }, { "epoch": 0.4053220834009411, "step": 1249, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2344396964181215e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.7600683349883184e-05 }, { "epoch": 0.4053220834009411, "step": 1249, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1170555808348581e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.631727718631737e-05 }, { "epoch": 0.4053220834009411, "step": 1249, "weight_max": 3.632264997577295e-05, "weight_mean": 9.007392691273708e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6677087842253968e-05 }, { "epoch": 0.4053220834009411, "step": 1249, "weight_max": 4.252122016623616e-05, "weight_mean": 1.063107447407674e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9683024220285006e-05 }, { "epoch": 0.4056466006814863, "grad_norm": 0.16458776593208313, "learning_rate": 7.661224489795919e-06, "loss": 0.0736, "step": 1250 }, { "epoch": 0.4056466006814863, "step": 1250, "weight_max": 3.397988257347606e-05, "weight_mean": 8.889481250662357e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.458493534300942e-05 }, { "epoch": 0.4056466006814863, "step": 1250, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.856939530232921e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.6911775674088858e-05 }, { "epoch": 0.4056466006814863, "step": 1250, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3241161468613427e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8386179363005795e-05 }, { "epoch": 0.4056466006814863, "step": 1250, "weight_max": 3.006794759130571e-05, "weight_mean": 3.7600275391014293e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0630005817802157e-05 }, { "epoch": 0.4059711179620315, "step": 1251, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.877803338691592e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6527686966583133e-05 }, { "epoch": 0.4059711179620315, "step": 1251, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5707089914940298e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1684794774046168e-05 }, { "epoch": 0.4059711179620315, "step": 1251, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235589014773723e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806334547756705e-05 }, { "epoch": 0.4059711179620315, "step": 1251, "weight_max": 4.381950566312298e-05, "weight_mean": 9.036031769937836e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.722616252664011e-05 }, { "epoch": 0.40629563524257667, "step": 1252, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.360383046441711e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7413451132597402e-05 }, { "epoch": 0.40629563524257667, "step": 1252, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153704351978377e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8802798877004534e-05 }, { "epoch": 0.40629563524257667, "step": 1252, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4878502042847686e-05, "weight_min": 0.0, "weight_std": 2.053961725323461e-05 }, { "epoch": 0.40629563524257667, "step": 1252, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4292739251686726e-05, "weight_min": 0.0, "weight_std": 1.985893140954431e-05 }, { "epoch": 0.40662015252312184, "step": 1253, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5631128917448223e-05, "weight_min": 0.0, "weight_std": 1.8262324374518357e-05 }, { "epoch": 0.40662015252312184, "step": 1253, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3612949260277674e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8910166545538232e-05 }, { "epoch": 0.40662015252312184, "step": 1253, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212150088278577e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1911871297343168e-05 }, { "epoch": 0.40662015252312184, "step": 1253, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233406101178844e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8953935068566352e-05 }, { "epoch": 0.40694466980366706, "step": 1254, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9984628276433796e-06, "weight_min": 0.0, "weight_std": 1.4135978744889144e-05 }, { "epoch": 0.40694466980366706, "step": 1254, "weight_max": 3.22740561387036e-05, "weight_mean": 4.035246092826128e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1410202205297537e-05 }, { "epoch": 0.40694466980366706, "step": 1254, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.936905371432658e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8488868590793572e-05 }, { "epoch": 0.40694466980366706, "step": 1254, "weight_max": 3.662098970380612e-05, "weight_mean": 1.3217899322626181e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8252954760100693e-05 }, { "epoch": 0.4072691870842122, "step": 1255, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.966489202866796e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1212176104891114e-05 }, { "epoch": 0.4072691870842122, "step": 1255, "weight_max": 4.381950566312298e-05, "weight_mean": 2.0562656573019922e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.2066331439418718e-05 }, { "epoch": 0.4072691870842122, "step": 1255, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6466596207465045e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.797861659724731e-05 }, { "epoch": 0.4072691870842122, "step": 1255, "weight_max": 4.252122016623616e-05, "weight_mean": 1.047104706231039e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9390174202271737e-05 }, { "epoch": 0.40759370436475745, "step": 1256, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729313332063612e-06, "weight_min": 0.0, "weight_std": 1.3373296496865805e-05 }, { "epoch": 0.40759370436475745, "step": 1256, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.05522067315178e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9543422240531072e-05 }, { "epoch": 0.40759370436475745, "step": 1256, "weight_max": 2.4894337911973707e-05, "weight_mean": 5.556029918807326e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0384736924606841e-05 }, { "epoch": 0.40759370436475745, "step": 1256, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3937924197525717e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0109780962229706e-05 }, { "epoch": 0.4079182216453026, "step": 1257, "weight_max": 3.2068148811958963e-06, "weight_mean": 4.038672045680869e-07, "weight_min": 5.958802340444436e-10, "weight_std": 1.132578063334222e-06 }, { "epoch": 0.4079182216453026, "step": 1257, "weight_max": 3.312147964606993e-05, "weight_mean": 4.14202622778248e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.1709467798937112e-05 }, { "epoch": 0.4079182216453026, "step": 1257, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.39384257333586e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2423931366356555e-05 }, { "epoch": 0.4079182216453026, "step": 1257, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.668180609878618e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6613912521279417e-05 }, { "epoch": 0.4082427389258478, "step": 1258, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274780834151898e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.903872180264443e-05 }, { "epoch": 0.4082427389258478, "step": 1258, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.0114024007634725e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.532618625788018e-05 }, { "epoch": 0.4082427389258478, "step": 1258, "weight_max": 4.252122016623616e-05, "weight_mean": 9.564784704707563e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.785179483704269e-05 }, { "epoch": 0.4082427389258478, "step": 1258, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5786064977874048e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.178941576858051e-05 }, { "epoch": 0.408567256206393, "step": 1259, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.64931405178504e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7869588191388175e-05 }, { "epoch": 0.408567256206393, "step": 1259, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3053007023700047e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9170633095200174e-05 }, { "epoch": 0.408567256206393, "step": 1259, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2202593097754288e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7133561414084397e-05 }, { "epoch": 0.408567256206393, "step": 1259, "weight_max": 1.3727149417164242e-09, "weight_mean": 7.95971288880537e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.5463250952206806e-10 }, { "epoch": 0.40889177348693817, "grad_norm": 9.716146469116211, "learning_rate": 7.640816326530612e-06, "loss": 0.0629, "step": 1260 }, { "epoch": 0.40889177348693817, "step": 1260, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3472111934097484e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.839548713178374e-05 }, { "epoch": 0.40889177348693817, "step": 1260, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0402898371685296e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1888121409574524e-05 }, { "epoch": 0.40889177348693817, "step": 1260, "weight_max": 3.572963396436535e-05, "weight_mean": 1.2772552508977242e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.7454214685130864e-05 }, { "epoch": 0.40889177348693817, "step": 1260, "weight_max": 2.926079287135508e-05, "weight_mean": 3.939495854865527e-06, "weight_min": 0.0, "weight_std": 1.0261624993290752e-05 }, { "epoch": 0.4092162907674834, "step": 1261, "weight_max": 4.187939703115262e-05, "weight_mean": 9.775768376130145e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.816076473915018e-05 }, { "epoch": 0.4092162907674834, "step": 1261, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9602804210735485e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.402543784934096e-05 }, { "epoch": 0.4092162907674834, "step": 1261, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4586364159185905e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0367790057207458e-05 }, { "epoch": 0.4092162907674834, "step": 1261, "weight_max": 3.5434961318969727e-05, "weight_mean": 7.761263077554759e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4559921510226559e-05 }, { "epoch": 0.40954080804802856, "step": 1262, "weight_max": 3.0339444492710754e-05, "weight_mean": 6.858360393380281e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2630071068997495e-05 }, { "epoch": 0.40954080804802856, "step": 1262, "weight_max": 4.187939703115262e-05, "weight_mean": 1.974280530703254e-05, "weight_min": 7.403143098549947e-10, "weight_std": 2.0366807802929543e-05 }, { "epoch": 0.40954080804802856, "step": 1262, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.427562534052413e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5622097635059617e-05 }, { "epoch": 0.40954080804802856, "step": 1262, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0078374543809332e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.867907303676475e-05 }, { "epoch": 0.4098653253285737, "step": 1263, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5043617167975754e-05, "weight_min": 0.0, "weight_std": 2.079517798847519e-05 }, { "epoch": 0.4098653253285737, "step": 1263, "weight_max": 3.722132169059478e-05, "weight_mean": 1.4174197531247046e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7487313016317785e-05 }, { "epoch": 0.4098653253285737, "step": 1263, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.732917189656291e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.9009987227036618e-05 }, { "epoch": 0.4098653253285737, "step": 1263, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.6110752565437e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7801685316953808e-05 }, { "epoch": 0.41018984260911895, "step": 1264, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393874405650422e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2423919542925432e-05 }, { "epoch": 0.41018984260911895, "step": 1264, "weight_max": 2.202445102739148e-05, "weight_mean": 2.7535727440408664e-06, "weight_min": 0.0, "weight_std": 7.78661069489317e-06 }, { "epoch": 0.41018984260911895, "step": 1264, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5077981515787542e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0833680537180044e-05 }, { "epoch": 0.41018984260911895, "step": 1264, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6326528566423804e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.020605643338058e-05 }, { "epoch": 0.4105143598896641, "step": 1265, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3294455129653215e-05, "weight_min": 0.0, "weight_std": 1.854221773101017e-05 }, { "epoch": 0.4105143598896641, "step": 1265, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3893556570110377e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9378954675630666e-05 }, { "epoch": 0.4105143598896641, "step": 1265, "weight_max": 1.1534618815645103e-09, "weight_mean": 7.699342274847254e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.0395388156124739e-10 }, { "epoch": 0.4105143598896641, "step": 1265, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.232138159452006e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.49207289723563e-05 }, { "epoch": 0.41083887717020934, "step": 1266, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.5014602467999794e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7423917597625405e-05 }, { "epoch": 0.41083887717020934, "step": 1266, "weight_max": 3.967146767536178e-05, "weight_mean": 9.170461453322787e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.6975240214378573e-05 }, { "epoch": 0.41083887717020934, "step": 1266, "weight_max": 4.252122016623616e-05, "weight_mean": 7.926006219349802e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.577284092491027e-05 }, { "epoch": 0.41083887717020934, "step": 1266, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2286664968996774e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7307203961536288e-05 }, { "epoch": 0.4111633944507545, "step": 1267, "weight_max": 2.7956259529560157e-09, "weight_mean": 1.0815579543077547e-09, "weight_min": 5.771870759119224e-10, "weight_std": 7.485109754234998e-10 }, { "epoch": 0.4111633944507545, "step": 1267, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5868772607063875e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1914120225119404e-05 }, { "epoch": 0.4111633944507545, "step": 1267, "weight_max": 4.187939703115262e-05, "weight_mean": 2.650951864779927e-05, "weight_min": 8.378387983398738e-10, "weight_std": 1.7472686522523873e-05 }, { "epoch": 0.4111633944507545, "step": 1267, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.542928637529258e-06, "weight_min": 0.0, "weight_std": 1.771710958564654e-05 }, { "epoch": 0.41148791173129967, "step": 1268, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8378104869043455e-05, "weight_min": 0.0, "weight_std": 1.9918346879421733e-05 }, { "epoch": 0.41148791173129967, "step": 1268, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3088232662994415e-05, "weight_min": 0.0, "weight_std": 1.8468528651283123e-05 }, { "epoch": 0.41148791173129967, "step": 1268, "weight_max": 4.252122016623616e-05, "weight_mean": 1.458322731195949e-05, "weight_min": 0.0, "weight_std": 2.019633393501863e-05 }, { "epoch": 0.41148791173129967, "step": 1268, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.4370600183610804e-05, "weight_min": 0.0, "weight_std": 1.619134854990989e-05 }, { "epoch": 0.4118124290118449, "step": 1269, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7609409042051993e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8989685486303642e-05 }, { "epoch": 0.4118124290118449, "step": 1269, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315903308655834e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033218915050384e-05 }, { "epoch": 0.4118124290118449, "step": 1269, "weight_max": 3.340639523230493e-05, "weight_mean": 4.177371920377482e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1810308933490887e-05 }, { "epoch": 0.4118124290118449, "step": 1269, "weight_max": 3.722132169059478e-05, "weight_mean": 1.2831115782319102e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7924790881806985e-05 }, { "epoch": 0.41213694629239006, "grad_norm": 11.681173324584961, "learning_rate": 7.620408163265307e-06, "loss": 0.2046, "step": 1270 }, { "epoch": 0.41213694629239006, "step": 1270, "weight_max": 4.187939703115262e-05, "weight_mean": 1.555593917146325e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1451207430800423e-05 }, { "epoch": 0.41213694629239006, "step": 1270, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4054603525437415e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.772222458384931e-05 }, { "epoch": 0.41213694629239006, "step": 1270, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3688189937965944e-05, "weight_min": 0.0, "weight_std": 1.9174758563167416e-05 }, { "epoch": 0.41213694629239006, "step": 1270, "weight_max": 3.9360926166409627e-05, "weight_mean": 5.129089004185516e-06, "weight_min": 0.0, "weight_std": 1.3844010027241893e-05 }, { "epoch": 0.4124614635729353, "step": 1271, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4491301044472493e-05, "weight_min": 7.937498991417158e-10, "weight_std": 2.0155472157057375e-05 }, { "epoch": 0.4124614635729353, "step": 1271, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3234520338301081e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9053568394156173e-05 }, { "epoch": 0.4124614635729353, "step": 1271, "weight_max": 4.187939703115262e-05, "weight_mean": 1.011699168884661e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.874674671853427e-05 }, { "epoch": 0.4124614635729353, "step": 1271, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.4187791748554446e-05, "weight_min": 0.0, "weight_std": 1.959090332093183e-05 }, { "epoch": 0.41278598085348045, "step": 1272, "weight_max": 3.397988257347606e-05, "weight_mean": 4.251876362104667e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2011932994937524e-05 }, { "epoch": 0.41278598085348045, "step": 1272, "weight_max": 2.714597657771378e-09, "weight_mean": 1.0282097395730716e-09, "weight_min": 0.0, "weight_std": 8.111291083245931e-10 }, { "epoch": 0.41278598085348045, "step": 1272, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4847177226329222e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.0527426386252046e-05 }, { "epoch": 0.41278598085348045, "step": 1272, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0311534424545243e-05, "weight_min": 0.0, "weight_std": 1.9095499737886712e-05 }, { "epoch": 0.4131104981340256, "step": 1273, "weight_max": 4.252122016623616e-05, "weight_mean": 6.069019036658574e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4876382920192555e-05 }, { "epoch": 0.4131104981340256, "step": 1273, "weight_max": 3.9051607018336654e-05, "weight_mean": 2.127196239598561e-05, "weight_min": 8.490471659072796e-10, "weight_std": 1.7768996258382685e-05 }, { "epoch": 0.4131104981340256, "step": 1273, "weight_max": 3.397988257347606e-05, "weight_mean": 6.719000793964369e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3006650988245383e-05 }, { "epoch": 0.4131104981340256, "step": 1273, "weight_max": 2.3677199351368472e-05, "weight_mean": 3.6542853649734752e-06, "weight_min": 6.196655966128617e-10, "weight_std": 8.31606757856207e-06 }, { "epoch": 0.41343501541457084, "step": 1274, "weight_max": 2.639517333591357e-05, "weight_mean": 3.2997559173963964e-06, "weight_min": 0.0, "weight_std": 9.3319576990325e-06 }, { "epoch": 0.41343501541457084, "step": 1274, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0086931069963612e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.873689870990347e-05 }, { "epoch": 0.41343501541457084, "step": 1274, "weight_max": 1.868417864869798e-09, "weight_mean": 8.0910556032876e-10, "weight_min": 0.0, "weight_std": 5.163263150365083e-10 }, { "epoch": 0.41343501541457084, "step": 1274, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.726060867847991e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0534525245020632e-05 }, { "epoch": 0.413759532695116, "step": 1275, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4577632100554183e-05, "weight_min": 0.0, "weight_std": 2.0160940039204434e-05 }, { "epoch": 0.413759532695116, "step": 1275, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3924644008511677e-05, "weight_min": 0.0, "weight_std": 1.9394416085560806e-05 }, { "epoch": 0.413759532695116, "step": 1275, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5045421352842823e-05, "weight_min": 0.0, "weight_std": 2.0811408830923028e-05 }, { "epoch": 0.413759532695116, "step": 1275, "weight_max": 4.381950566312298e-05, "weight_mean": 1.674826671660412e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9207927834941074e-05 }, { "epoch": 0.41408404997566123, "step": 1276, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.69636039371835e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8000477211899124e-05 }, { "epoch": 0.41408404997566123, "step": 1276, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.3209752220718656e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2217707080708351e-05 }, { "epoch": 0.41408404997566123, "step": 1276, "weight_max": 4.252122016623616e-05, "weight_mean": 2.360803773626685e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.987648465728853e-05 }, { "epoch": 0.41408404997566123, "step": 1276, "weight_max": 4.8733774349329906e-08, "weight_mean": 7.797002155029986e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.685765127490413e-08 }, { "epoch": 0.4144085672562064, "step": 1277, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.361292541143484e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7390377252013423e-05 }, { "epoch": 0.4144085672562064, "step": 1277, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.323640940943733e-06, "weight_min": 7.403143098549947e-10, "weight_std": 1.5441062714671716e-05 }, { "epoch": 0.4144085672562064, "step": 1277, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0394422133686021e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.926423647091724e-05 }, { "epoch": 0.4144085672562064, "step": 1277, "weight_max": 4.3167921830900013e-05, "weight_mean": 7.73243391449796e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5740048183943145e-05 }, { "epoch": 0.41473308453675156, "step": 1278, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0313067832612433e-05, "weight_min": 0.0, "weight_std": 1.909455568238627e-05 }, { "epoch": 0.41473308453675156, "step": 1278, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0395775461802259e-05, "weight_min": 0.0, "weight_std": 1.9263401554781012e-05 }, { "epoch": 0.41473308453675156, "step": 1278, "weight_max": 2.563926318543963e-05, "weight_mean": 3.2121151889441535e-06, "weight_min": 5.771870759119224e-10, "weight_std": 9.06195509742247e-06 }, { "epoch": 0.41473308453675156, "step": 1278, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.4753147954470478e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.064521686406806e-05 }, { "epoch": 0.4150576018172968, "step": 1279, "weight_max": 3.9051607018336654e-05, "weight_mean": 5.192662229092093e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3708527148992289e-05 }, { "epoch": 0.4150576018172968, "step": 1279, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4881963579682633e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.055884397123009e-05 }, { "epoch": 0.4150576018172968, "step": 1279, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4438071048061829e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0031524400110357e-05 }, { "epoch": 0.4150576018172968, "step": 1279, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0919172382273246e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7328953617834486e-05 }, { "epoch": 0.41538211909784195, "grad_norm": 4.504613876342773, "learning_rate": 7.600000000000001e-06, "loss": 0.1371, "step": 1280 }, { "epoch": 0.41538211909784195, "step": 1280, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.0554005712037906e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.5483004972338676e-05 }, { "epoch": 0.41538211909784195, "step": 1280, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0588253644527867e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.952143975358922e-05 }, { "epoch": 0.41538211909784195, "step": 1280, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.157511168363271e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4326902601169422e-05 }, { "epoch": 0.41538211909784195, "step": 1280, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3778762877336703e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9047402020078152e-05 }, { "epoch": 0.4157066363783872, "step": 1281, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843452188652009e-06, "weight_min": 0.0, "weight_std": 1.3697690519620664e-05 }, { "epoch": 0.4157066363783872, "step": 1281, "weight_max": 3.662098970380612e-05, "weight_mean": 8.302942660520785e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.5480345609830692e-05 }, { "epoch": 0.4157066363783872, "step": 1281, "weight_max": 4.187939703115262e-05, "weight_mean": 9.981152288673911e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8469134374754503e-05 }, { "epoch": 0.4157066363783872, "step": 1281, "weight_max": 3.782653220696375e-05, "weight_mean": 6.830262464063708e-06, "weight_min": 0.0, "weight_std": 1.3837129699822981e-05 }, { "epoch": 0.41603115365893234, "step": 1282, "weight_max": 4.252122016623616e-05, "weight_mean": 1.883091317722574e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0422041416168213e-05 }, { "epoch": 0.41603115365893234, "step": 1282, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4930524230294395e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0669371224357747e-05 }, { "epoch": 0.41603115365893234, "step": 1282, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5081061064847745e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0850888176937588e-05 }, { "epoch": 0.41603115365893234, "step": 1282, "weight_max": 4.187939703115262e-05, "weight_mean": 1.363141018373426e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7868474969873205e-05 }, { "epoch": 0.4163556709394775, "step": 1283, "weight_max": 2.440382377244532e-05, "weight_mean": 3.0511382647091523e-06, "weight_min": 5.958802340444436e-10, "weight_std": 8.627787792647723e-06 }, { "epoch": 0.4163556709394775, "step": 1283, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5036440345284063e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0756651792908087e-05 }, { "epoch": 0.4163556709394775, "step": 1283, "weight_max": 2.926079287135508e-05, "weight_mean": 7.315826678677695e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3544727153202984e-05 }, { "epoch": 0.4163556709394775, "step": 1283, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235651769908145e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806309081905056e-05 }, { "epoch": 0.41668018822002273, "step": 1284, "weight_max": 1.4847285456198733e-05, "weight_mean": 3.41605255016475e-06, "weight_min": 0.0, "weight_std": 6.355350251396885e-06 }, { "epoch": 0.41668018822002273, "step": 1284, "weight_max": 4.252122016623616e-05, "weight_mean": 1.766511377354618e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0858436982962303e-05 }, { "epoch": 0.41668018822002273, "step": 1284, "weight_max": 3.9535716496175155e-06, "weight_mean": 4.951482424075948e-07, "weight_min": 0.0, "weight_std": 1.3974145076645073e-06 }, { "epoch": 0.41668018822002273, "step": 1284, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4793086847930681e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8133781850337982e-05 }, { "epoch": 0.4170047055005679, "step": 1285, "weight_max": 4.187939703115262e-05, "weight_mean": 7.116306278476259e-06, "weight_min": 7.298507354036587e-10, "weight_std": 1.4999515769886784e-05 }, { "epoch": 0.4170047055005679, "step": 1285, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631365512381308e-05, "weight_min": 0.0, "weight_std": 1.9687362510012463e-05 }, { "epoch": 0.4170047055005679, "step": 1285, "weight_max": 5.5787999997392035e-08, "weight_mean": 8.253328687146677e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.9268000528427365e-08 }, { "epoch": 0.4170047055005679, "step": 1285, "weight_max": 2.952862632810138e-05, "weight_mean": 3.6992328205087688e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0436670891067479e-05 }, { "epoch": 0.4173292227811131, "step": 1286, "weight_max": 5.663789437448941e-09, "weight_mean": 1.5455727808699748e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.7155665776869e-09 }, { "epoch": 0.4173292227811131, "step": 1286, "weight_max": 3.782653220696375e-05, "weight_mean": 8.869095836416818e-06, "weight_min": 0.0, "weight_std": 1.64689845405519e-05 }, { "epoch": 0.4173292227811131, "step": 1286, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8932263628812507e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0320219846325926e-05 }, { "epoch": 0.4173292227811131, "step": 1286, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4042799193703104e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9578397768782452e-05 }, { "epoch": 0.4176537400616583, "step": 1287, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1188900316483341e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.6813961337902583e-05 }, { "epoch": 0.4176537400616583, "step": 1287, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504055596044054e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2736600183416158e-05 }, { "epoch": 0.4176537400616583, "step": 1287, "weight_max": 2.3677199351368472e-05, "weight_mean": 3.7154231904423796e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.338015504705254e-06 }, { "epoch": 0.4176537400616583, "step": 1287, "weight_max": 3.998322426923551e-05, "weight_mean": 9.427714758203365e-06, "weight_min": 0.0, "weight_std": 1.7497917724540457e-05 }, { "epoch": 0.41797825734220345, "step": 1288, "weight_max": 3.813097282545641e-05, "weight_mean": 7.265324711624999e-06, "weight_min": 0.0, "weight_std": 1.4298778296506498e-05 }, { "epoch": 0.41797825734220345, "step": 1288, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3565308108809404e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.896118737931829e-05 }, { "epoch": 0.41797825734220345, "step": 1288, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0883424440398812e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.779692138370592e-05 }, { "epoch": 0.41797825734220345, "step": 1288, "weight_max": 1.7212929970256141e-09, "weight_mean": 9.003229273218949e-10, "weight_min": 6.390288853630466e-10, "weight_std": 3.6108918632926645e-10 }, { "epoch": 0.4183027746227487, "step": 1289, "weight_max": 4.252122016623616e-05, "weight_mean": 1.047142632160103e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.938993955263868e-05 }, { "epoch": 0.4183027746227487, "step": 1289, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396560936787864e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.526193409517873e-05 }, { "epoch": 0.4183027746227487, "step": 1289, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4452284631261136e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.010392017837148e-05 }, { "epoch": 0.4183027746227487, "step": 1289, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2022132068523206e-05, "weight_min": 0.0, "weight_std": 1.7269783711526543e-05 }, { "epoch": 0.41862729190329384, "grad_norm": 0.8360490798950195, "learning_rate": 7.579591836734694e-06, "loss": 0.0712, "step": 1290 }, { "epoch": 0.41862729190329384, "step": 1290, "weight_max": 3.3513551933594954e-09, "weight_mean": 1.3095431405929503e-09, "weight_min": 6.053385570581327e-10, "weight_std": 9.129290656773037e-10 }, { "epoch": 0.41862729190329384, "step": 1290, "weight_max": 4.252122016623616e-05, "weight_mean": 1.270293887500884e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.915174834721256e-05 }, { "epoch": 0.41862729190329384, "step": 1290, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653397809306625e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3159428817743901e-05 }, { "epoch": 0.41862729190329384, "step": 1290, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767372956848703e-06, "weight_min": 0.0, "weight_std": 1.348092973785242e-05 }, { "epoch": 0.41895180918383906, "step": 1291, "weight_max": 9.534083744711097e-09, "weight_mean": 1.6626642285189064e-09, "weight_min": 0.0, "weight_std": 3.201564791055489e-09 }, { "epoch": 0.41895180918383906, "step": 1291, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2406331734382547e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7516349544166587e-05 }, { "epoch": 0.41895180918383906, "step": 1291, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9727829567273147e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1125008061062545e-05 }, { "epoch": 0.41895180918383906, "step": 1291, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578216703521321e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.294723460887326e-05 }, { "epoch": 0.41927632646438423, "step": 1292, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.5034334865049459e-05, "weight_min": 0.0, "weight_std": 2.0748284441651776e-05 }, { "epoch": 0.41927632646438423, "step": 1292, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4684372217743658e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.043391577899456e-05 }, { "epoch": 0.41927632646438423, "step": 1292, "weight_max": 4.187939703115262e-05, "weight_mean": 1.141266602644464e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.725540278130211e-05 }, { "epoch": 0.41927632646438423, "step": 1292, "weight_max": 3.813097282545641e-05, "weight_mean": 1.3601216778624803e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.883459117379971e-05 }, { "epoch": 0.4196008437449294, "step": 1293, "weight_max": 3.143761568935588e-05, "weight_mean": 5.715425686503295e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1532331882335711e-05 }, { "epoch": 0.4196008437449294, "step": 1293, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0374608791607898e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8585677025839686e-05 }, { "epoch": 0.4196008437449294, "step": 1293, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0394317541795317e-05, "weight_min": 0.0, "weight_std": 1.9264301954535767e-05 }, { "epoch": 0.4196008437449294, "step": 1293, "weight_max": 3.602553260861896e-05, "weight_mean": 1.2698366845143028e-05, "weight_min": 0.0, "weight_std": 1.7637741621001624e-05 }, { "epoch": 0.4199253610254746, "step": 1294, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077000878372928e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357664440467488e-05 }, { "epoch": 0.4199253610254746, "step": 1294, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9756011170102283e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1242949514999054e-05 }, { "epoch": 0.4199253610254746, "step": 1294, "weight_max": 3.782653220696375e-05, "weight_mean": 8.728125976631418e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.6234980648732744e-05 }, { "epoch": 0.4199253610254746, "step": 1294, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5117231669137254e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.088786095555406e-05 }, { "epoch": 0.4202498783060198, "step": 1295, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.1998674381175078e-05, "weight_min": 0.0, "weight_std": 1.7250889868591912e-05 }, { "epoch": 0.4202498783060198, "step": 1295, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4650267985416576e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0237417629687116e-05 }, { "epoch": 0.4202498783060198, "step": 1295, "weight_max": 7.07865721594203e-09, "weight_mean": 1.541548777517221e-09, "weight_min": 6.196655966128617e-10, "weight_std": 2.2412847133779223e-09 }, { "epoch": 0.4202498783060198, "step": 1295, "weight_max": 3.312147964606993e-05, "weight_mean": 1.4876132809149567e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.6438951206509955e-05 }, { "epoch": 0.420574395586565, "step": 1296, "weight_max": 3.998322426923551e-05, "weight_mean": 1.688952397671528e-05, "weight_min": 0.0, "weight_std": 1.8856735550798476e-05 }, { "epoch": 0.420574395586565, "step": 1296, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1899033779627644e-05, "weight_min": 0.0, "weight_std": 1.7523370843264274e-05 }, { "epoch": 0.420574395586565, "step": 1296, "weight_max": 3.632264997577295e-05, "weight_mean": 7.872274181863759e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.4802285477344412e-05 }, { "epoch": 0.420574395586565, "step": 1296, "weight_max": 1.03811572671475e-08, "weight_mean": 1.9035977238957003e-09, "weight_min": 0.0, "weight_std": 3.482250932052011e-09 }, { "epoch": 0.4208989128671102, "step": 1297, "weight_max": 1.4129907022208954e-08, "weight_mean": 2.4437671886090584e-09, "weight_min": 6.053385570581327e-10, "weight_std": 4.725710933684013e-09 }, { "epoch": 0.4208989128671102, "step": 1297, "weight_max": 4.252122016623616e-05, "weight_mean": 9.456608495383989e-06, "weight_min": 7.721521200210191e-10, "weight_std": 1.768642323440872e-05 }, { "epoch": 0.4208989128671102, "step": 1297, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882159373664763e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3806541574012954e-05 }, { "epoch": 0.4208989128671102, "step": 1297, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0240819392492995e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.8994487618329003e-05 }, { "epoch": 0.42122343014765534, "step": 1298, "weight_max": 8.267048712262692e-10, "weight_mean": 6.760905169045373e-10, "weight_min": 5.958802340444436e-10, "weight_std": 7.67611391117029e-11 }, { "epoch": 0.42122343014765534, "step": 1298, "weight_max": 3.782653220696375e-05, "weight_mean": 4.770921805175021e-06, "weight_min": 0.0, "weight_std": 1.3357001989788841e-05 }, { "epoch": 0.42122343014765534, "step": 1298, "weight_max": 2.8994179956498556e-05, "weight_mean": 6.495001798612066e-06, "weight_min": 0.0, "weight_std": 1.2132690244470723e-05 }, { "epoch": 0.42122343014765534, "step": 1298, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156203769729473e-06, "weight_min": 0.0, "weight_std": 1.4581049981643446e-05 }, { "epoch": 0.42154794742820056, "step": 1299, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0674484403571114e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.2153795725898817e-05 }, { "epoch": 0.42154794742820056, "step": 1299, "weight_max": 3.312147964606993e-05, "weight_mean": 4.14107717006118e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1709850696206558e-05 }, { "epoch": 0.42154794742820056, "step": 1299, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2345504728727974e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7959062461159192e-05 }, { "epoch": 0.42154794742820056, "step": 1299, "weight_max": 2.8994179956498556e-05, "weight_mean": 6.06838602834614e-06, "weight_min": 0.0, "weight_std": 1.1515126971062273e-05 }, { "epoch": 0.42187246470874573, "grad_norm": 0.6862631440162659, "learning_rate": 7.559183673469388e-06, "loss": 0.1322, "step": 1300 }, { "epoch": 0.42187246470874573, "eval_accuracy": 0.9785, "eval_auc": 0.9945982221439887, "eval_loss": 0.06843172013759613, "eval_runtime": 155.4499, "eval_samples_per_second": 12.866, "eval_steps_per_second": 1.608, "step": 1300 }, { "epoch": 0.42187246470874573, "step": 1300, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0076023500005249e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.866246384452097e-05 }, { "epoch": 0.42187246470874573, "step": 1300, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9729644918697886e-05, "weight_min": 0.0, "weight_std": 2.1133770133019425e-05 }, { "epoch": 0.42187246470874573, "step": 1300, "weight_max": 3.397988257347606e-05, "weight_mean": 1.2143073035986163e-05, "weight_min": 0.0, "weight_std": 1.6778232748038135e-05 }, { "epoch": 0.42187246470874573, "step": 1300, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793809355556732e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0726056643761694e-05 }, { "epoch": 0.42219698198929095, "step": 1301, "weight_max": 4.252122016623616e-05, "weight_mean": 2.423656769678928e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0324634533608332e-05 }, { "epoch": 0.42219698198929095, "step": 1301, "weight_max": 3.998322426923551e-05, "weight_mean": 9.318215234088711e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.731348856992554e-05 }, { "epoch": 0.42219698198929095, "step": 1301, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615785655914806e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3053092516202014e-05 }, { "epoch": 0.42219698198929095, "step": 1301, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0555227163422387e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9532526494003832e-05 }, { "epoch": 0.4225214992698361, "step": 1302, "weight_max": 2.1563215341302566e-05, "weight_mean": 2.8574743282661075e-06, "weight_min": 6.100955851628953e-10, "weight_std": 7.57172938392614e-06 }, { "epoch": 0.4225214992698361, "step": 1302, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396741926233517e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.526186133560259e-05 }, { "epoch": 0.4225214992698361, "step": 1302, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390842362539843e-05, "weight_min": 0.0, "weight_std": 1.9239349057897925e-05 }, { "epoch": 0.4225214992698361, "step": 1302, "weight_max": 9.437653147870151e-08, "weight_mean": 1.2806971483314555e-08, "weight_min": 5.864964625068581e-10, "weight_std": 3.297497741527877e-08 }, { "epoch": 0.4228460165503813, "step": 1303, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728651219920721e-06, "weight_min": 0.0, "weight_std": 1.3373562978813425e-05 }, { "epoch": 0.4228460165503813, "step": 1303, "weight_max": 1.7790385982152657e-06, "weight_mean": 2.2660817933228827e-07, "weight_min": 0.0, "weight_std": 6.27298334165971e-07 }, { "epoch": 0.4228460165503813, "step": 1303, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.918701673275791e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0842326193815097e-05 }, { "epoch": 0.4228460165503813, "step": 1303, "weight_max": 3.0339444492710754e-05, "weight_mean": 4.105592779524159e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0636054867063649e-05 }, { "epoch": 0.4231705338309265, "step": 1304, "weight_max": 4.252122016623616e-05, "weight_mean": 1.473464635637356e-05, "weight_min": 0.0, "weight_std": 2.0391082216519862e-05 }, { "epoch": 0.4231705338309265, "step": 1304, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4840090443613008e-05, "weight_min": 0.0, "weight_std": 2.0488947484409437e-05 }, { "epoch": 0.4231705338309265, "step": 1304, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.83313441084465e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0149947886238806e-05 }, { "epoch": 0.4231705338309265, "step": 1304, "weight_max": 3.998322426923551e-05, "weight_mean": 1.601684925844893e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.866979437181726e-05 }, { "epoch": 0.4234950511114717, "step": 1305, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8049623728729784e-06, "weight_min": 0.0, "weight_std": 1.3589246009360068e-05 }, { "epoch": 0.4234950511114717, "step": 1305, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.453488312108675e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0082998162251897e-05 }, { "epoch": 0.4234950511114717, "step": 1305, "weight_max": 4.252122016623616e-05, "weight_mean": 2.628284892125521e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1197705791564658e-05 }, { "epoch": 0.4234950511114717, "step": 1305, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393444669403834e-06, "weight_min": 0.0, "weight_std": 1.242409234691877e-05 }, { "epoch": 0.4238195683920169, "step": 1306, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0770431698765606e-06, "weight_min": 0.0, "weight_std": 1.435764534107875e-05 }, { "epoch": 0.4238195683920169, "step": 1306, "weight_max": 4.252122016623616e-05, "weight_mean": 1.793637238733936e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0528113964246586e-05 }, { "epoch": 0.4238195683920169, "step": 1306, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.865575647272635e-06, "weight_min": 0.0, "weight_std": 1.3688889339391608e-05 }, { "epoch": 0.4238195683920169, "step": 1306, "weight_max": 3.967146767536178e-05, "weight_mean": 9.50000321608968e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.761152088874951e-05 }, { "epoch": 0.42414408567256207, "step": 1307, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.93785818334436e-06, "weight_min": 0.0, "weight_std": 1.6750576833146624e-05 }, { "epoch": 0.42414408567256207, "step": 1307, "weight_max": 2.288811629114207e-08, "weight_mean": 3.6374445677722633e-09, "weight_min": 0.0, "weight_std": 7.825239123349093e-09 }, { "epoch": 0.42414408567256207, "step": 1307, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.151733709382825e-06, "weight_min": 0.0, "weight_std": 1.5153504136833362e-05 }, { "epoch": 0.42414408567256207, "step": 1307, "weight_max": 3.602553260861896e-05, "weight_mean": 8.643990440759808e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.602285192348063e-05 }, { "epoch": 0.42446860295310723, "step": 1308, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4379398635355756e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.914495078381151e-05 }, { "epoch": 0.42446860295310723, "step": 1308, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140954388276441e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1709900718415156e-05 }, { "epoch": 0.42446860295310723, "step": 1308, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.1425776392570697e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.753170909069013e-05 }, { "epoch": 0.42446860295310723, "step": 1308, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140900728089036e-06, "weight_min": 0.0, "weight_std": 1.1709922546287999e-05 }, { "epoch": 0.42479312023365245, "step": 1309, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2355658226588275e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806343642703723e-05 }, { "epoch": 0.42479312023365245, "step": 1309, "weight_max": 2.952862632810138e-05, "weight_mean": 3.692055770443403e-06, "weight_min": 0.0, "weight_std": 1.0439551260787994e-05 }, { "epoch": 0.42479312023365245, "step": 1309, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3868705536879133e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9238326785853133e-05 }, { "epoch": 0.42479312023365245, "step": 1309, "weight_max": 1.4159473593622351e-09, "weight_mean": 8.002430940123872e-10, "weight_min": 6.100955851628953e-10, "weight_std": 2.670696441775533e-10 }, { "epoch": 0.4251176375141976, "grad_norm": 4.1662678718566895, "learning_rate": 7.5387755102040825e-06, "loss": 0.1448, "step": 1310 }, { "epoch": 0.4251176375141976, "step": 1310, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6331077858922072e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9904578948626295e-05 }, { "epoch": 0.4251176375141976, "step": 1310, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0491307875781786e-05, "weight_min": 0.0, "weight_std": 1.9377746866666712e-05 }, { "epoch": 0.4251176375141976, "step": 1310, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4031726095709018e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9649443856906146e-05 }, { "epoch": 0.4251176375141976, "step": 1310, "weight_max": 4.187939703115262e-05, "weight_mean": 9.863964805845171e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8025148165179417e-05 }, { "epoch": 0.42544215479474284, "step": 1311, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3474158549797721e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8759041267912835e-05 }, { "epoch": 0.42544215479474284, "step": 1311, "weight_max": 3.998322426923551e-05, "weight_mean": 5.002107172913384e-06, "weight_min": 0.0, "weight_std": 1.4134510820440482e-05 }, { "epoch": 0.42544215479474284, "step": 1311, "weight_max": 4.252122016623616e-05, "weight_mean": 2.324345041415654e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.962378701136913e-05 }, { "epoch": 0.42544215479474284, "step": 1311, "weight_max": 2.926079287135508e-05, "weight_mean": 6.6177690314361826e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2342984518909361e-05 }, { "epoch": 0.425766672075288, "step": 1312, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.84460178995505e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.369722667732276e-05 }, { "epoch": 0.425766672075288, "step": 1312, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998752956453245e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135860510577913e-05 }, { "epoch": 0.425766672075288, "step": 1312, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.158196578733623e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.569487176311668e-05 }, { "epoch": 0.425766672075288, "step": 1312, "weight_max": 4.187939703115262e-05, "weight_mean": 9.66527295531705e-06, "weight_min": 0.0, "weight_std": 1.7976890376303345e-05 }, { "epoch": 0.4260911893558332, "step": 1313, "weight_max": 4.252122016623616e-05, "weight_mean": 1.425292794010602e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.97940844373079e-05 }, { "epoch": 0.4260911893558332, "step": 1313, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.2922242603963241e-05, "weight_min": 0.0, "weight_std": 1.5228079064399935e-05 }, { "epoch": 0.4260911893558332, "step": 1313, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631447366904467e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9687311578309163e-05 }, { "epoch": 0.4260911893558332, "step": 1313, "weight_max": 1.3620764320876333e-06, "weight_mean": 1.7912734051606094e-07, "weight_min": 0.0, "weight_std": 4.782334599440219e-07 }, { "epoch": 0.4264157066363784, "step": 1314, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959581929142587e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4025718883203808e-05 }, { "epoch": 0.4264157066363784, "step": 1314, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4128550901659764e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9736253307200968e-05 }, { "epoch": 0.4264157066363784, "step": 1314, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4660270608146675e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0067362129339017e-05 }, { "epoch": 0.4264157066363784, "step": 1314, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7932945815555286e-06, "weight_min": 0.0, "weight_std": 1.0726264008553699e-05 }, { "epoch": 0.42674022391692357, "step": 1315, "weight_max": 3.998322426923551e-05, "weight_mean": 1.5980358512024395e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.7962458514375612e-05 }, { "epoch": 0.42674022391692357, "step": 1315, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077105925010983e-06, "weight_min": 0.0, "weight_std": 1.4357619875227101e-05 }, { "epoch": 0.42674022391692357, "step": 1315, "weight_max": 3.340639523230493e-05, "weight_mean": 1.210779373650439e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.6727319234632887e-05 }, { "epoch": 0.42674022391692357, "step": 1315, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7438336726627313e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8854201698559336e-05 }, { "epoch": 0.4270647411974688, "step": 1316, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4920899957360234e-05, "weight_min": 0.0, "weight_std": 1.8717833881964907e-05 }, { "epoch": 0.4270647411974688, "step": 1316, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4827290215180255e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0739988030982204e-05 }, { "epoch": 0.4270647411974688, "step": 1316, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396782853495097e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.5261844964697957e-05 }, { "epoch": 0.4270647411974688, "step": 1316, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8927847122540697e-05, "weight_min": 0.0, "weight_std": 2.0724275600514375e-05 }, { "epoch": 0.42738925847801396, "step": 1317, "weight_max": 4.187939703115262e-05, "weight_mean": 8.994633390102535e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.694839920673985e-05 }, { "epoch": 0.42738925847801396, "step": 1317, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.003775469143875e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8593073036754504e-05 }, { "epoch": 0.42738925847801396, "step": 1317, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3680038793827407e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.993765908991918e-05 }, { "epoch": 0.42738925847801396, "step": 1317, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728959538624622e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3373439287533984e-05 }, { "epoch": 0.4277137757585591, "step": 1318, "weight_max": 4.381950566312298e-05, "weight_mean": 1.8685344912228175e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0292685803724453e-05 }, { "epoch": 0.4277137757585591, "step": 1318, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5433110092999414e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1319068764569238e-05 }, { "epoch": 0.4277137757585591, "step": 1318, "weight_max": 1.1931196031156333e-09, "weight_mean": 6.912634908928794e-10, "weight_min": 0.0, "weight_std": 4.546162613028315e-10 }, { "epoch": 0.4277137757585591, "step": 1318, "weight_max": 3.967146767536178e-05, "weight_mean": 2.2645310309599154e-05, "weight_min": 7.508524357824342e-10, "weight_std": 1.8985296264872886e-05 }, { "epoch": 0.42803829303910435, "step": 1319, "weight_max": 3.061215829802677e-05, "weight_mean": 3.8271200537565164e-06, "weight_min": 0.0, "weight_std": 1.082278868125286e-05 }, { "epoch": 0.42803829303910435, "step": 1319, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8243010345031507e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9686178347910754e-05 }, { "epoch": 0.42803829303910435, "step": 1319, "weight_max": 4.187939703115262e-05, "weight_mean": 9.963940101442859e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8479677237337455e-05 }, { "epoch": 0.42803829303910435, "step": 1319, "weight_max": 4.187939703115262e-05, "weight_mean": 1.048010108206654e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9380082449060865e-05 }, { "epoch": 0.4283628103196495, "grad_norm": 4.012815475463867, "learning_rate": 7.518367346938777e-06, "loss": 0.1749, "step": 1320 }, { "epoch": 0.4283628103196495, "step": 1320, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.407637218013406e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5257489394571166e-05 }, { "epoch": 0.4283628103196495, "step": 1320, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.311246685683727e-06, "weight_min": 0.0, "weight_std": 1.7267102521145716e-05 }, { "epoch": 0.4283628103196495, "step": 1320, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.998686437029392e-06, "weight_min": 0.0, "weight_std": 1.852492459875066e-05 }, { "epoch": 0.4283628103196495, "step": 1320, "weight_max": 2.440382377244532e-05, "weight_mean": 3.124468094028998e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.60059753904352e-06 }, { "epoch": 0.42868732760019473, "step": 1321, "weight_max": 3.998322426923551e-05, "weight_mean": 1.44555488077458e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9963383238064125e-05 }, { "epoch": 0.42868732760019473, "step": 1321, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960474926629104e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8457003534422256e-05 }, { "epoch": 0.42868732760019473, "step": 1321, "weight_max": 3.4268457966391e-05, "weight_mean": 1.2179172699688934e-05, "weight_min": 0.0, "weight_std": 1.6832087567308918e-05 }, { "epoch": 0.42868732760019473, "step": 1321, "weight_max": 4.187939703115262e-05, "weight_mean": 9.628938641981222e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7835407561506145e-05 }, { "epoch": 0.4290118448807399, "step": 1322, "weight_max": 3.967146767536178e-05, "weight_mean": 6.455845323216636e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4058005945116747e-05 }, { "epoch": 0.4290118448807399, "step": 1322, "weight_max": 4.772478412462533e-09, "weight_mean": 1.3527744480157367e-09, "weight_min": 0.0, "weight_std": 1.497315382792408e-09 }, { "epoch": 0.4290118448807399, "step": 1322, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.124124072608538e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.433922898286255e-05 }, { "epoch": 0.4290118448807399, "step": 1322, "weight_max": 3.998322426923551e-05, "weight_mean": 8.556316970498301e-06, "weight_min": 0.0, "weight_std": 1.61387215484865e-05 }, { "epoch": 0.42933636216128507, "step": 1323, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920671017316636e-06, "weight_min": 0.0, "weight_std": 1.391596561006736e-05 }, { "epoch": 0.42933636216128507, "step": 1323, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.0065677795792e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.685476854618173e-05 }, { "epoch": 0.42933636216128507, "step": 1323, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4773260772926733e-05, "weight_min": 0.0, "weight_std": 2.0605575627996586e-05 }, { "epoch": 0.42933636216128507, "step": 1323, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8687049305299297e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0291419787099585e-05 }, { "epoch": 0.4296608794418303, "step": 1324, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153811672353186e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8802733393386006e-05 }, { "epoch": 0.4296608794418303, "step": 1324, "weight_max": 3.718365615412722e-09, "weight_mean": 1.2204424137962633e-09, "weight_min": 6.29309993005478e-10, "weight_std": 1.060526666485373e-09 }, { "epoch": 0.4296608794418303, "step": 1324, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23556718690088e-06, "weight_min": 0.0, "weight_std": 1.4806343642703723e-05 }, { "epoch": 0.4296608794418303, "step": 1324, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.2951739083509892e-05, "weight_min": 0.0, "weight_std": 1.87700534297619e-05 }, { "epoch": 0.42998539672237546, "step": 1325, "weight_max": 4.252122016623616e-05, "weight_mean": 1.966160925803706e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1099891455378383e-05 }, { "epoch": 0.42998539672237546, "step": 1325, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5036906006571371e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0766221496160142e-05 }, { "epoch": 0.42998539672237546, "step": 1325, "weight_max": 3.8743506593164057e-05, "weight_mean": 5.42087946087122e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3560952538682614e-05 }, { "epoch": 0.42998539672237546, "step": 1325, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471039786352776e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9390177840250544e-05 }, { "epoch": 0.4303099140029207, "step": 1326, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2463679013308138e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.767275716701988e-05 }, { "epoch": 0.4303099140029207, "step": 1326, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.07746881339699e-06, "weight_min": 0.0, "weight_std": 1.4357474356074817e-05 }, { "epoch": 0.4303099140029207, "step": 1326, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1117326721432619e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7757192836143076e-05 }, { "epoch": 0.4303099140029207, "step": 1326, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882268058281625e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3806497918267269e-05 }, { "epoch": 0.43063443128346585, "step": 1327, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2913543287140783e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8802193153533153e-05 }, { "epoch": 0.43063443128346585, "step": 1327, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.46455604408402e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0211136870784685e-05 }, { "epoch": 0.43063443128346585, "step": 1327, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6723624867154285e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.828535823733546e-05 }, { "epoch": 0.43063443128346585, "step": 1327, "weight_max": 3.843663216684945e-05, "weight_mean": 9.125028554990422e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6926389434956945e-05 }, { "epoch": 0.430958948564011, "step": 1328, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5272422388079576e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1093541363370605e-05 }, { "epoch": 0.430958948564011, "step": 1328, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3490735000232235e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9013725250260904e-05 }, { "epoch": 0.430958948564011, "step": 1328, "weight_max": 3.572963396436535e-05, "weight_mean": 4.466746304387925e-06, "weight_min": 0.0, "weight_std": 1.2632114703592379e-05 }, { "epoch": 0.430958948564011, "step": 1328, "weight_max": 4.666427244615079e-09, "weight_mean": 1.5064063330072486e-09, "weight_min": 0.0, "weight_std": 1.4279633031577532e-09 }, { "epoch": 0.43128346584455624, "step": 1329, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.079353296518093e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9981907826149836e-05 }, { "epoch": 0.43128346584455624, "step": 1329, "weight_max": 3.22740561387036e-05, "weight_mean": 7.972064850036986e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4743094652658328e-05 }, { "epoch": 0.43128346584455624, "step": 1329, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.57314478000626e-06, "weight_min": 0.0, "weight_std": 1.7734342691255733e-05 }, { "epoch": 0.43128346584455624, "step": 1329, "weight_max": 3.578502383305704e-09, "weight_mean": 1.1712997238788603e-09, "weight_min": 0.0, "weight_std": 1.1300032021210882e-09 }, { "epoch": 0.4316079831251014, "grad_norm": 8.026043891906738, "learning_rate": 7.49795918367347e-06, "loss": 0.137, "step": 1330 }, { "epoch": 0.4316079831251014, "step": 1330, "weight_max": 4.252122016623616e-05, "weight_mean": 2.900455547205638e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8019638446276076e-05 }, { "epoch": 0.4316079831251014, "step": 1330, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9985274017672054e-06, "weight_min": 0.0, "weight_std": 1.4135952369542792e-05 }, { "epoch": 0.4316079831251014, "step": 1330, "weight_max": 4.252122016623616e-05, "weight_mean": 1.383414382871706e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9646216969704255e-05 }, { "epoch": 0.4316079831251014, "step": 1330, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.20004297059495e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9367682398296893e-05 }, { "epoch": 0.4319325004056466, "step": 1331, "weight_max": 4.187939703115262e-05, "weight_mean": 9.963767297449522e-06, "weight_min": 0.0, "weight_std": 1.8479784557712264e-05 }, { "epoch": 0.4319325004056466, "step": 1331, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5229928976623341e-05, "weight_min": 0.0, "weight_std": 2.1021071006543934e-05 }, { "epoch": 0.4319325004056466, "step": 1331, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390819625172298e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.923936179082375e-05 }, { "epoch": 0.4319325004056466, "step": 1331, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4561985153704882e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.038650018221233e-05 }, { "epoch": 0.4322570176861918, "step": 1332, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4308758181869052e-05, "weight_min": 0.0, "weight_std": 1.9948398403357714e-05 }, { "epoch": 0.4322570176861918, "step": 1332, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.071183396561537e-05, "weight_min": 0.0, "weight_std": 1.983350739465095e-05 }, { "epoch": 0.4322570176861918, "step": 1332, "weight_max": 4.252122016623616e-05, "weight_mean": 9.747999683895614e-06, "weight_min": 0.0, "weight_std": 1.8140339307137765e-05 }, { "epoch": 0.4322570176861918, "step": 1332, "weight_max": 3.722132169059478e-05, "weight_mean": 1.3335478797671385e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.844746839196887e-05 }, { "epoch": 0.43258153496673696, "step": 1333, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1704058124450967e-05, "weight_min": 0.0, "weight_std": 1.732514465402346e-05 }, { "epoch": 0.43258153496673696, "step": 1333, "weight_max": 3.998322426923551e-05, "weight_mean": 1.472397161705885e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.032623888226226e-05 }, { "epoch": 0.43258153496673696, "step": 1333, "weight_max": 3.752332122530788e-05, "weight_mean": 4.780482413480058e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3232469427748583e-05 }, { "epoch": 0.43258153496673696, "step": 1333, "weight_max": 3.967146767536178e-05, "weight_mean": 9.612534995540045e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.780868842615746e-05 }, { "epoch": 0.4329060522472822, "step": 1334, "weight_max": 1.1534618815645103e-09, "weight_mean": 6.052826573288428e-10, "weight_min": 0.0, "weight_std": 4.105664974662915e-10 }, { "epoch": 0.4329060522472822, "step": 1334, "weight_max": 3.312147964606993e-05, "weight_mean": 5.857929863850586e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.2017514563922305e-05 }, { "epoch": 0.4329060522472822, "step": 1334, "weight_max": 4.252122016623616e-05, "weight_mean": 5.37215601070784e-06, "weight_min": 0.0, "weight_std": 1.5011323739599902e-05 }, { "epoch": 0.4329060522472822, "step": 1334, "weight_max": 3.36925259034615e-05, "weight_mean": 9.395662345923483e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.418408464815002e-05 }, { "epoch": 0.43323056952782735, "step": 1335, "weight_max": 2.595289316786875e-09, "weight_mean": 9.63286450605949e-10, "weight_min": 6.686325937366178e-10, "weight_std": 6.636821647809654e-10 }, { "epoch": 0.43323056952782735, "step": 1335, "weight_max": 4.252122016623616e-05, "weight_mean": 1.997515937546268e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1363508494687267e-05 }, { "epoch": 0.43323056952782735, "step": 1335, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4968258255976252e-05, "weight_min": 0.0, "weight_std": 2.071414746751543e-05 }, { "epoch": 0.43323056952782735, "step": 1335, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.7296979422098957e-05, "weight_min": 0.0, "weight_std": 1.8713057215791196e-05 }, { "epoch": 0.43355508680837257, "step": 1336, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1343692676746286e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.703669477137737e-05 }, { "epoch": 0.43355508680837257, "step": 1336, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.4103979891806375e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5256391634466127e-05 }, { "epoch": 0.43355508680837257, "step": 1336, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3361167475522961e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8591061234474182e-05 }, { "epoch": 0.43355508680837257, "step": 1336, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844519025937188e-06, "weight_min": 0.0, "weight_std": 1.3697260328626726e-05 }, { "epoch": 0.43387960408891774, "step": 1337, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.30970418266952e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7255646525882185e-05 }, { "epoch": 0.43387960408891774, "step": 1337, "weight_max": 3.998322426923551e-05, "weight_mean": 8.86012458067853e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.6581376257818192e-05 }, { "epoch": 0.43387960408891774, "step": 1337, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.053832395176869e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8276918126503006e-05 }, { "epoch": 0.43387960408891774, "step": 1337, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.47809850104386e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4479091078101192e-05 }, { "epoch": 0.4342041213694629, "step": 1338, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.09108386823209e-05, "weight_min": 0.0, "weight_std": 1.797124423319474e-05 }, { "epoch": 0.4342041213694629, "step": 1338, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7684367776382715e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9494520529406145e-05 }, { "epoch": 0.4342041213694629, "step": 1338, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5279161743819714e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.8098560758517124e-05 }, { "epoch": 0.4342041213694629, "step": 1338, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235791832092218e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806252693233546e-05 }, { "epoch": 0.4345286386500081, "step": 1339, "weight_max": 3.722132169059478e-05, "weight_mean": 1.723300010780804e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8531074601924047e-05 }, { "epoch": 0.4345286386500081, "step": 1339, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.696422239358071e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8000440832111053e-05 }, { "epoch": 0.4345286386500081, "step": 1339, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.6156546886777505e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3053146176389419e-05 }, { "epoch": 0.4345286386500081, "step": 1339, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0481242245296016e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1969179215375334e-05 }, { "epoch": 0.4348531559305533, "grad_norm": 5.8708696365356445, "learning_rate": 7.477551020408164e-06, "loss": 0.2106, "step": 1340 }, { "epoch": 0.4348531559305533, "step": 1340, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.671869176789187e-05, "weight_min": 0.0, "weight_std": 1.9507469914969988e-05 }, { "epoch": 0.4348531559305533, "step": 1340, "weight_max": 4.252122016623616e-05, "weight_mean": 7.954182365210727e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5799647371750325e-05 }, { "epoch": 0.4348531559305533, "step": 1340, "weight_max": 3.722132169059478e-05, "weight_mean": 2.1812713384861127e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8198245015810244e-05 }, { "epoch": 0.4348531559305533, "step": 1340, "weight_max": 4.252122016623616e-05, "weight_mean": 6.17267187408288e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.488158250140259e-05 }, { "epoch": 0.4351776732110985, "step": 1341, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.204182788380422e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7095839211833663e-05 }, { "epoch": 0.4351776732110985, "step": 1341, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767839982378064e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3480741472449154e-05 }, { "epoch": 0.4351776732110985, "step": 1341, "weight_max": 3.782653220696375e-05, "weight_mean": 4.728856765723322e-06, "weight_min": 0.0, "weight_std": 1.3373480214795563e-05 }, { "epoch": 0.4351776732110985, "step": 1341, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2097063518012874e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8656079191714525e-05 }, { "epoch": 0.4355021904916437, "step": 1342, "weight_max": 4.3167921830900013e-05, "weight_mean": 3.4895710996352136e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.4494507922790945e-05 }, { "epoch": 0.4355021904916437, "step": 1342, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.8822689677763265e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3806497008772567e-05 }, { "epoch": 0.4355021904916437, "step": 1342, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284361693862593e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2115404388168827e-05 }, { "epoch": 0.4355021904916437, "step": 1342, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.78620812727604e-06, "weight_min": 0.0, "weight_std": 1.6267971659544855e-05 }, { "epoch": 0.43582670777218885, "step": 1343, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3866958397557028e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9670254914672114e-05 }, { "epoch": 0.43582670777218885, "step": 1343, "weight_max": 3.998322426923551e-05, "weight_mean": 4.999248631065711e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4135661331238225e-05 }, { "epoch": 0.43582670777218885, "step": 1343, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4217537682270631e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9909874026780017e-05 }, { "epoch": 0.43582670777218885, "step": 1343, "weight_max": 3.340639523230493e-05, "weight_mean": 4.70176473754691e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1691262443491723e-05 }, { "epoch": 0.43615122505273407, "step": 1344, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1699674359988421e-05, "weight_min": 7.508524357824342e-10, "weight_std": 1.8782146071316674e-05 }, { "epoch": 0.43615122505273407, "step": 1344, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.82327856036136e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8183220163336955e-05 }, { "epoch": 0.43615122505273407, "step": 1344, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805407570529496e-06, "weight_min": 0.0, "weight_std": 1.3589066838903818e-05 }, { "epoch": 0.43615122505273407, "step": 1344, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0133702744496986e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8837807147065178e-05 }, { "epoch": 0.43647574233327924, "step": 1345, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4240853235824034e-05, "weight_min": 0.0, "weight_std": 2.0184876120765693e-05 }, { "epoch": 0.43647574233327924, "step": 1345, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0080229913000949e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8650840502232313e-05 }, { "epoch": 0.43647574233327924, "step": 1345, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471125278854743e-05, "weight_min": 0.0, "weight_std": 1.9390126908547245e-05 }, { "epoch": 0.43647574233327924, "step": 1345, "weight_max": 3.813097282545641e-05, "weight_mean": 5.0239327720191795e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3396502254181542e-05 }, { "epoch": 0.43680025961382446, "step": 1346, "weight_max": 4.252122016623616e-05, "weight_mean": 1.578626142872963e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1789252059534192e-05 }, { "epoch": 0.43680025961382446, "step": 1346, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0631302757246885e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9682884158100933e-05 }, { "epoch": 0.43680025961382446, "step": 1346, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396633696364006e-06, "weight_min": 0.0, "weight_std": 1.5261904991348274e-05 }, { "epoch": 0.43680025961382446, "step": 1346, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140777946304297e-06, "weight_min": 0.0, "weight_std": 1.1709971659001894e-05 }, { "epoch": 0.4371247768943696, "step": 1347, "weight_max": 3.340639523230493e-05, "weight_mean": 1.119425542128738e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.5537427316303365e-05 }, { "epoch": 0.4371247768943696, "step": 1347, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5785439245519228e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1785912394989282e-05 }, { "epoch": 0.4371247768943696, "step": 1347, "weight_max": 4.252122016623616e-05, "weight_mean": 1.257396797882393e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7881075109471567e-05 }, { "epoch": 0.4371247768943696, "step": 1347, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5068511856952682e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0940440663252957e-05 }, { "epoch": 0.4374492941749148, "step": 1348, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4740839105797932e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0424591639311984e-05 }, { "epoch": 0.4374492941749148, "step": 1348, "weight_max": 2.3831858015910257e-06, "weight_mean": 2.9872623485971417e-07, "weight_min": 5.679522963042416e-10, "weight_std": 8.422490509474301e-07 }, { "epoch": 0.4374492941749148, "step": 1348, "weight_max": 3.813097282545641e-05, "weight_mean": 8.871928912412841e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.648636862228159e-05 }, { "epoch": 0.4374492941749148, "step": 1348, "weight_max": 3.8743506593164057e-05, "weight_mean": 5.227884230407653e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3585033229901455e-05 }, { "epoch": 0.43777381145546, "step": 1349, "weight_max": 3.9360926166409627e-05, "weight_mean": 6.009708613419207e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.381641959596891e-05 }, { "epoch": 0.43777381145546, "step": 1349, "weight_max": 4.252122016623616e-05, "weight_mean": 2.31684298341861e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9967374100815505e-05 }, { "epoch": 0.43777381145546, "step": 1349, "weight_max": 3.22740561387036e-05, "weight_mean": 5.728511041525053e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1726331649697386e-05 }, { "epoch": 0.43777381145546, "step": 1349, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8737584468908608e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0087130906176753e-05 }, { "epoch": 0.4380983287360052, "grad_norm": 2.487344980239868, "learning_rate": 7.457142857142857e-06, "loss": 0.0387, "step": 1350 }, { "epoch": 0.4380983287360052, "step": 1350, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0078234481625259e-05, "weight_min": 0.0, "weight_std": 1.8477152480045334e-05 }, { "epoch": 0.4380983287360052, "step": 1350, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.991049612115603e-05, "weight_min": 0.0, "weight_std": 2.143305573554244e-05 }, { "epoch": 0.4380983287360052, "step": 1350, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8055271690827794e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3589017726189923e-05 }, { "epoch": 0.4380983287360052, "step": 1350, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235625394561794e-06, "weight_min": 0.0, "weight_std": 1.4806319086346775e-05 }, { "epoch": 0.4384228460165504, "step": 1351, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.321010692365235e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2217692528793123e-05 }, { "epoch": 0.4384228460165504, "step": 1351, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.041582416102756e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6907553799683228e-05 }, { "epoch": 0.4384228460165504, "step": 1351, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6891459381440654e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.83986485353671e-05 }, { "epoch": 0.4384228460165504, "step": 1351, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.462184991396498e-05, "weight_min": 0.0, "weight_std": 2.0402627342264168e-05 }, { "epoch": 0.43874736329709557, "step": 1352, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3932641195424367e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9277218598290347e-05 }, { "epoch": 0.43874736329709557, "step": 1352, "weight_max": 3.722132169059478e-05, "weight_mean": 8.51427193993004e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5853951481403783e-05 }, { "epoch": 0.43874736329709557, "step": 1352, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.023218374029966e-05, "weight_min": 0.0, "weight_std": 1.8945653209812008e-05 }, { "epoch": 0.43874736329709557, "step": 1352, "weight_max": 3.340639523230493e-05, "weight_mean": 4.176601123617729e-06, "weight_min": 0.0, "weight_std": 1.1810619980678894e-05 }, { "epoch": 0.43907188057764074, "step": 1353, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.416970553691499e-05, "weight_min": 0.0, "weight_std": 1.965115370694548e-05 }, { "epoch": 0.43907188057764074, "step": 1353, "weight_max": 4.187939703115262e-05, "weight_mean": 5.237801815383136e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4805441423959564e-05 }, { "epoch": 0.43907188057764074, "step": 1353, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653435553336749e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3159413356333971e-05 }, { "epoch": 0.43907188057764074, "step": 1353, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315763246471761e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5033275303721894e-05 }, { "epoch": 0.43939639785818596, "step": 1354, "weight_max": 6.9493695242783815e-09, "weight_mean": 1.6473212793854941e-09, "weight_min": 6.100955851628953e-10, "weight_std": 2.1766803914857746e-09 }, { "epoch": 0.43939639785818596, "step": 1354, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8156189980800264e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9802197130047716e-05 }, { "epoch": 0.43939639785818596, "step": 1354, "weight_max": 4.252122016623616e-05, "weight_mean": 2.299753759871237e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9948487533838488e-05 }, { "epoch": 0.43939639785818596, "step": 1354, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430036824487615e-06, "weight_min": 0.0, "weight_std": 1.2527880244306289e-05 }, { "epoch": 0.4397209151387311, "step": 1355, "weight_max": 3.6920544516760856e-05, "weight_mean": 2.034102362813428e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.6974199752439745e-05 }, { "epoch": 0.4397209151387311, "step": 1355, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0314054634363856e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9107494154013693e-05 }, { "epoch": 0.4397209151387311, "step": 1355, "weight_max": 3.199401908204891e-05, "weight_mean": 4.024298050353536e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1301677659503184e-05 }, { "epoch": 0.4397209151387311, "step": 1355, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3965025573270395e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9301285647088662e-05 }, { "epoch": 0.44004543241927635, "step": 1356, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4914255189069081e-05, "weight_min": 0.0, "weight_std": 2.0739731553476304e-05 }, { "epoch": 0.44004543241927635, "step": 1356, "weight_max": 4.252122016623616e-05, "weight_mean": 9.174007573165e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.517681812401861e-05 }, { "epoch": 0.44004543241927635, "step": 1356, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4777780961594544e-05, "weight_min": 0.0, "weight_std": 2.046820918621961e-05 }, { "epoch": 0.44004543241927635, "step": 1356, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0261484931106679e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8981852917931974e-05 }, { "epoch": 0.4403699496998215, "step": 1357, "weight_max": 1.247039249641091e-09, "weight_mean": 6.460795787255336e-10, "weight_min": 0.0, "weight_std": 3.417215121981343e-10 }, { "epoch": 0.4403699496998215, "step": 1357, "weight_max": 3.199401908204891e-05, "weight_mean": 1.16869923658669e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.613330823602155e-05 }, { "epoch": 0.4403699496998215, "step": 1357, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.3322339731967077e-06, "weight_min": 0.0, "weight_std": 9.421632057637908e-06 }, { "epoch": 0.4403699496998215, "step": 1357, "weight_max": 1.9527599537383367e-09, "weight_mean": 8.501375159397639e-10, "weight_min": 0.0, "weight_std": 5.699626881572328e-10 }, { "epoch": 0.4406944669803667, "step": 1358, "weight_max": 4.252122016623616e-05, "weight_mean": 5.342625627235975e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5022606021375395e-05 }, { "epoch": 0.4406944669803667, "step": 1358, "weight_max": 4.187939703115262e-05, "weight_mean": 8.99413771549007e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6948699339991435e-05 }, { "epoch": 0.4406944669803667, "step": 1358, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311442565580364e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9091037756879814e-05 }, { "epoch": 0.4406944669803667, "step": 1358, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3020409824093804e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8385344446869567e-05 }, { "epoch": 0.4410189842609119, "step": 1359, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.530546776659321e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7682854377198964e-05 }, { "epoch": 0.4410189842609119, "step": 1359, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2017817425657995e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6823611076688394e-05 }, { "epoch": 0.4410189842609119, "step": 1359, "weight_max": 8.378387983398738e-10, "weight_mean": 6.003917918384616e-10, "weight_min": 0.0, "weight_std": 2.5649823931495064e-10 }, { "epoch": 0.4410189842609119, "step": 1359, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.61608033219818e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3052974281890783e-05 }, { "epoch": 0.4413435015414571, "grad_norm": 66.1968994140625, "learning_rate": 7.436734693877552e-06, "loss": 0.2715, "step": 1360 }, { "epoch": 0.4413435015414571, "step": 1360, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2120077371946536e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7285365174757317e-05 }, { "epoch": 0.4413435015414571, "step": 1360, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3561108062276617e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.892744330689311e-05 }, { "epoch": 0.4413435015414571, "step": 1360, "weight_max": 3.8743506593164057e-05, "weight_mean": 6.919907718838658e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4111117707216181e-05 }, { "epoch": 0.4413435015414571, "step": 1360, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.962403055396862e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8440012354403734e-05 }, { "epoch": 0.4416680188220023, "step": 1361, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2912590136693325e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8100499801221304e-05 }, { "epoch": 0.4416680188220023, "step": 1361, "weight_max": 4.187939703115262e-05, "weight_mean": 8.046632501645945e-06, "weight_min": 0.0, "weight_std": 1.5773952327435836e-05 }, { "epoch": 0.4416680188220023, "step": 1361, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959530087944586e-06, "weight_min": 0.0, "weight_std": 1.4025741620571353e-05 }, { "epoch": 0.4416680188220023, "step": 1361, "weight_max": 3.967146767536178e-05, "weight_mean": 4.95963467983529e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4025697964825667e-05 }, { "epoch": 0.44199253610254746, "step": 1362, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153960829484276e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.880264244391583e-05 }, { "epoch": 0.44199253610254746, "step": 1362, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3518168998416513e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.886001518869307e-05 }, { "epoch": 0.44199253610254746, "step": 1362, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1560637075454e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581106370314956e-05 }, { "epoch": 0.44199253610254746, "step": 1362, "weight_max": 4.252122016623616e-05, "weight_mean": 1.970900848391466e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.0202276573400013e-05 }, { "epoch": 0.44231705338309263, "step": 1363, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.962583337444812e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0751453121192753e-05 }, { "epoch": 0.44231705338309263, "step": 1363, "weight_max": 2.7549629244560947e-09, "weight_mean": 1.1690050039092625e-09, "weight_min": 5.958802340444436e-10, "weight_std": 7.425343118150352e-10 }, { "epoch": 0.44231705338309263, "step": 1363, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920810624753358e-06, "weight_min": 0.0, "weight_std": 1.3915908311901148e-05 }, { "epoch": 0.44231705338309263, "step": 1363, "weight_max": 3.813097282545641e-05, "weight_mean": 8.391041774302721e-06, "weight_min": 0.0, "weight_std": 1.5726940546301194e-05 }, { "epoch": 0.44264157066363785, "step": 1364, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235614480625372e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4806324543314986e-05 }, { "epoch": 0.44264157066363785, "step": 1364, "weight_max": 4.0610404539620504e-05, "weight_mean": 7.715106221439783e-06, "weight_min": 0.0, "weight_std": 1.5205833733489271e-05 }, { "epoch": 0.44264157066363785, "step": 1364, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4917010048520751e-05, "weight_min": 0.0, "weight_std": 2.060707629425451e-05 }, { "epoch": 0.44264157066363785, "step": 1364, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805764547199942e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3588922229246236e-05 }, { "epoch": 0.442966087944183, "step": 1365, "weight_max": 3.0339444492710754e-05, "weight_mean": 5.672996394423535e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.127049108617939e-05 }, { "epoch": 0.442966087944183, "step": 1365, "weight_max": 3.8743506593164057e-05, "weight_mean": 7.425528565363493e-06, "weight_min": 0.0, "weight_std": 1.4572954569302965e-05 }, { "epoch": 0.442966087944183, "step": 1365, "weight_max": 3.998322426923551e-05, "weight_mean": 5.019043328502448e-06, "weight_min": 0.0, "weight_std": 1.4127776921668556e-05 }, { "epoch": 0.442966087944183, "step": 1365, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4340465895656962e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9801125745289028e-05 }, { "epoch": 0.44329060522472824, "step": 1366, "weight_max": 3.28377827827353e-05, "weight_mean": 4.152021119807614e-06, "weight_min": 0.0, "weight_std": 1.1591532711463515e-05 }, { "epoch": 0.44329060522472824, "step": 1366, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.426241038250737e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.561275712447241e-05 }, { "epoch": 0.44329060522472824, "step": 1366, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.5304063456133008e-05, "weight_min": 0.0, "weight_std": 2.0956764274160378e-05 }, { "epoch": 0.44329060522472824, "step": 1366, "weight_max": 3.967146767536178e-05, "weight_mean": 9.24308187677525e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.717421764624305e-05 }, { "epoch": 0.4436151225052734, "step": 1367, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2989406968699768e-05, "weight_min": 0.0, "weight_std": 1.804871135391295e-05 }, { "epoch": 0.4436151225052734, "step": 1367, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.356945282779634e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2320758287387434e-05 }, { "epoch": 0.4436151225052734, "step": 1367, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.881634468911216e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.830459041229915e-05 }, { "epoch": 0.4436151225052734, "step": 1367, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156473434908548e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4580940842279233e-05 }, { "epoch": 0.4439396397858186, "step": 1368, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930321327061392e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1114625522168353e-05 }, { "epoch": 0.4439396397858186, "step": 1368, "weight_max": 4.252122016623616e-05, "weight_mean": 8.775800779403653e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6725034583942033e-05 }, { "epoch": 0.4439396397858186, "step": 1368, "weight_max": 3.782653220696375e-05, "weight_mean": 9.232147021975834e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7099811884691007e-05 }, { "epoch": 0.4439396397858186, "step": 1368, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.252600648324005e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7239106455235742e-05 }, { "epoch": 0.4442641570663638, "step": 1369, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4845763871562667e-05, "weight_min": 0.0, "weight_std": 2.0521738406387158e-05 }, { "epoch": 0.4442641570663638, "step": 1369, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316466740623582e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5032990631880239e-05 }, { "epoch": 0.4442641570663638, "step": 1369, "weight_max": 3.572963396436535e-05, "weight_mean": 4.466959580895491e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.263202830159571e-05 }, { "epoch": 0.4442641570663638, "step": 1369, "weight_max": 3.340639523230493e-05, "weight_mean": 6.689851943519898e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2255923138582148e-05 }, { "epoch": 0.44458867434690896, "grad_norm": 0.34589964151382446, "learning_rate": 7.4163265306122455e-06, "loss": 0.0589, "step": 1370 }, { "epoch": 0.44458867434690896, "step": 1370, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076730303699151e-06, "weight_min": 0.0, "weight_std": 1.4357772670336999e-05 }, { "epoch": 0.44458867434690896, "step": 1370, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396489996201126e-06, "weight_min": 0.0, "weight_std": 1.5261963199009188e-05 }, { "epoch": 0.44458867434690896, "step": 1370, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236441666056635e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4805989849264733e-05 }, { "epoch": 0.44458867434690896, "step": 1370, "weight_max": 3.9051607018336654e-05, "weight_mean": 5.290612989483634e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3673278772330377e-05 }, { "epoch": 0.4449131916274542, "step": 1371, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0114843462361023e-05, "weight_min": 0.0, "weight_std": 1.8732269381871447e-05 }, { "epoch": 0.4449131916274542, "step": 1371, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.341447659797268e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8740496670943685e-05 }, { "epoch": 0.4449131916274542, "step": 1371, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.857359347923193e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.001257053052541e-05 }, { "epoch": 0.4449131916274542, "step": 1371, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1010288289980963e-05, "weight_min": 0.0, "weight_std": 1.7836378901847638e-05 }, { "epoch": 0.44523770890799935, "step": 1372, "weight_max": 3.843663216684945e-05, "weight_mean": 8.80430616234662e-06, "weight_min": 0.0, "weight_std": 1.6391921235481277e-05 }, { "epoch": 0.44523770890799935, "step": 1372, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.436381333216559e-06, "weight_min": 0.0, "weight_std": 1.753084870870225e-05 }, { "epoch": 0.44523770890799935, "step": 1372, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5272242308128625e-05, "weight_min": 0.0, "weight_std": 2.1091371309012175e-05 }, { "epoch": 0.44523770890799935, "step": 1372, "weight_max": 4.1242459701607004e-05, "weight_mean": 7.0595078796031885e-06, "weight_min": 0.0, "weight_std": 1.4804502825427335e-05 }, { "epoch": 0.4455622261885445, "step": 1373, "weight_max": 4.252122016623616e-05, "weight_mean": 2.142689481843263e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8403596186544746e-05 }, { "epoch": 0.4455622261885445, "step": 1373, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.805487025005277e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.8169326722272672e-05 }, { "epoch": 0.4455622261885445, "step": 1373, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.998848327086307e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8524824554333463e-05 }, { "epoch": 0.4455622261885445, "step": 1373, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.9207437768927775e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.39159346872475e-05 }, { "epoch": 0.44588674346908974, "step": 1374, "weight_max": 4.252122016623616e-05, "weight_mean": 1.890970452222973e-05, "weight_min": 0.0, "weight_std": 2.0503675841609947e-05 }, { "epoch": 0.44588674346908974, "step": 1374, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5762167095090263e-05, "weight_min": 0.0, "weight_std": 2.1264069800963625e-05 }, { "epoch": 0.44588674346908974, "step": 1374, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2961251741216984e-05, "weight_min": 0.0, "weight_std": 1.8610720871947706e-05 }, { "epoch": 0.44588674346908974, "step": 1374, "weight_max": 4.252122016623616e-05, "weight_mean": 1.042355415847851e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9228395103709772e-05 }, { "epoch": 0.4462112607496349, "step": 1375, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0859095709747635e-05, "weight_min": 0.0, "weight_std": 1.974682163563557e-05 }, { "epoch": 0.4462112607496349, "step": 1375, "weight_max": 2.952862632810138e-05, "weight_mean": 3.691711981446133e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.0439690413477365e-05 }, { "epoch": 0.4462112607496349, "step": 1375, "weight_max": 4.252122016623616e-05, "weight_mean": 9.108141966862604e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.717502345854882e-05 }, { "epoch": 0.4462112607496349, "step": 1375, "weight_max": 3.143761568935588e-05, "weight_mean": 8.268827514257282e-06, "weight_min": 0.0, "weight_std": 1.1229558367631398e-05 }, { "epoch": 0.44653577803018013, "step": 1376, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.111358849622775e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7015474440995604e-05 }, { "epoch": 0.44653577803018013, "step": 1376, "weight_max": 4.252122016623616e-05, "weight_mean": 5.387936653278302e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.5005341992946342e-05 }, { "epoch": 0.44653577803018013, "step": 1376, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.27706241782289e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.721810804156121e-05 }, { "epoch": 0.44653577803018013, "step": 1376, "weight_max": 4.381950566312298e-05, "weight_mean": 9.54773167904932e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7931311958818696e-05 }, { "epoch": 0.4468602953107253, "step": 1377, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077078185422579e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4357631698658224e-05 }, { "epoch": 0.4468602953107253, "step": 1377, "weight_max": 3.22740561387036e-05, "weight_mean": 4.2106330511160195e-06, "weight_min": 0.0, "weight_std": 1.1349985470587853e-05 }, { "epoch": 0.4468602953107253, "step": 1377, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.69381562754279e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.7970498447539285e-05 }, { "epoch": 0.4468602953107253, "step": 1377, "weight_max": 3.572963396436535e-05, "weight_mean": 1.1424435797380283e-05, "weight_min": 0.0, "weight_std": 1.6229696484515443e-05 }, { "epoch": 0.44718481259127046, "step": 1378, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5200837879092433e-05, "weight_min": 0.0, "weight_std": 1.8505423213355243e-05 }, { "epoch": 0.44718481259127046, "step": 1378, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9760125724133104e-05, "weight_min": 0.0, "weight_std": 2.0791996576008387e-05 }, { "epoch": 0.44718481259127046, "step": 1378, "weight_max": 1.1019851697824379e-08, "weight_mean": 2.185075675953385e-09, "weight_min": 6.053385570581327e-10, "weight_std": 3.5851652757656893e-09 }, { "epoch": 0.44718481259127046, "step": 1378, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4472588190983515e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0060671886312775e-05 }, { "epoch": 0.4475093298718157, "step": 1379, "weight_max": 4.252122016623616e-05, "weight_mean": 7.031740096863359e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5122952390811406e-05 }, { "epoch": 0.4475093298718157, "step": 1379, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6675287042744458e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8160386389354244e-05 }, { "epoch": 0.4475093298718157, "step": 1379, "weight_max": 3.9360926166409627e-05, "weight_mean": 5.471621989272535e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3779919754597358e-05 }, { "epoch": 0.4475093298718157, "step": 1379, "weight_max": 3.36925259034615e-05, "weight_mean": 7.73761985328747e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4400168765860144e-05 }, { "epoch": 0.44783384715236085, "grad_norm": 0.09684896469116211, "learning_rate": 7.39591836734694e-06, "loss": 0.0656, "step": 1380 }, { "epoch": 0.44783384715236085, "step": 1380, "weight_max": 4.187939703115262e-05, "weight_mean": 2.297978790011257e-05, "weight_min": 0.0, "weight_std": 1.9265302398707718e-05 }, { "epoch": 0.44783384715236085, "step": 1380, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.15614601681591e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.458107271901099e-05 }, { "epoch": 0.44783384715236085, "step": 1380, "weight_max": 3.813097282545641e-05, "weight_mean": 9.548543857818004e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5211962818284519e-05 }, { "epoch": 0.44783384715236085, "step": 1380, "weight_max": 3.116124207735993e-05, "weight_mean": 3.898184786521597e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1015940799552482e-05 }, { "epoch": 0.4481583644329061, "step": 1381, "weight_max": 2.1106867279740982e-05, "weight_mean": 2.6390111997898202e-06, "weight_min": 0.0, "weight_std": 7.462140729330713e-06 }, { "epoch": 0.4481583644329061, "step": 1381, "weight_max": 1.3870513626557113e-09, "weight_mean": 6.933650875673436e-10, "weight_min": 0.0, "weight_std": 3.916619528698817e-10 }, { "epoch": 0.4481583644329061, "step": 1381, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.078503363620257e-06, "weight_min": 0.0, "weight_std": 1.4357056898006704e-05 }, { "epoch": 0.4481583644329061, "step": 1381, "weight_max": 4.252122016623616e-05, "weight_mean": 7.171975084929727e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5198676010186318e-05 }, { "epoch": 0.44848288171345124, "step": 1382, "weight_max": 3.602553260861896e-05, "weight_mean": 8.933136996347457e-06, "weight_min": 0.0, "weight_std": 1.65402889251709e-05 }, { "epoch": 0.44848288171345124, "step": 1382, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2157084711361676e-05, "weight_min": 0.0, "weight_std": 1.7343798390356824e-05 }, { "epoch": 0.44848288171345124, "step": 1382, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5480327419936657e-05, "weight_min": 7.091466858177853e-10, "weight_std": 2.136156035703607e-05 }, { "epoch": 0.44848288171345124, "step": 1382, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076838533568662e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357729014591314e-05 }, { "epoch": 0.4488073989939964, "step": 1383, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3584022781287786e-05, "weight_min": 0.0, "weight_std": 1.9067118046223186e-05 }, { "epoch": 0.4488073989939964, "step": 1383, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5037251614558045e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0770228729816154e-05 }, { "epoch": 0.4488073989939964, "step": 1383, "weight_max": 7.298507354036587e-10, "weight_mean": 6.435835753215713e-10, "weight_min": 6.053385570581327e-10, "weight_std": 5.2053871629764714e-11 }, { "epoch": 0.4488073989939964, "step": 1383, "weight_max": 1.753466261078529e-09, "weight_mean": 8.914615712285467e-10, "weight_min": 5.958802340444436e-10, "weight_std": 4.395019626013408e-10 }, { "epoch": 0.44913191627454163, "step": 1384, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.565842831041664e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.7041940736817196e-05 }, { "epoch": 0.44913191627454163, "step": 1384, "weight_max": 2.563926318543963e-05, "weight_mean": 6.256027063500369e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1587033441173844e-05 }, { "epoch": 0.44913191627454163, "step": 1384, "weight_max": 4.381950566312298e-05, "weight_mean": 5.478807906911243e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5491981685045175e-05 }, { "epoch": 0.44913191627454163, "step": 1384, "weight_max": 3.312147964606993e-05, "weight_mean": 1.3160573871573433e-05, "weight_min": 0.0, "weight_std": 1.5315477867261507e-05 }, { "epoch": 0.4494564335550868, "step": 1385, "weight_max": 3.843663216684945e-05, "weight_mean": 4.9095569920609705e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3550032235798426e-05 }, { "epoch": 0.4494564335550868, "step": 1385, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0158806617255323e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.88357662409544e-05 }, { "epoch": 0.4494564335550868, "step": 1385, "weight_max": 3.813097282545641e-05, "weight_mean": 1.283493111259304e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.765596061886754e-05 }, { "epoch": 0.4494564335550868, "step": 1385, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5414763008011505e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.284153313463321e-05 }, { "epoch": 0.449780950835632, "step": 1386, "weight_max": 3.843663216684945e-05, "weight_mean": 8.330276614287868e-06, "weight_min": 0.0, "weight_std": 1.5664156308048405e-05 }, { "epoch": 0.449780950835632, "step": 1386, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3943868907517754e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9469547623884864e-05 }, { "epoch": 0.449780950835632, "step": 1386, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.1349011401762255e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.5835707017686218e-05 }, { "epoch": 0.449780950835632, "step": 1386, "weight_max": 1.6181260207304149e-06, "weight_mean": 2.0299854952554597e-07, "weight_min": 0.0, "weight_std": 5.717982958231005e-07 }, { "epoch": 0.4501054681161772, "step": 1387, "weight_max": 4.252122016623616e-05, "weight_mean": 2.641276478243526e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1031011783634312e-05 }, { "epoch": 0.4501054681161772, "step": 1387, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.07696586282691e-06, "weight_min": 0.0, "weight_std": 1.4357678082888015e-05 }, { "epoch": 0.4501054681161772, "step": 1387, "weight_max": 2.952862632810138e-05, "weight_mean": 3.71196074411273e-06, "weight_min": 0.0, "weight_std": 1.0431664122734219e-05 }, { "epoch": 0.4501054681161772, "step": 1387, "weight_max": 4.187939703115262e-05, "weight_mean": 9.340861652162857e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.7460935850976966e-05 }, { "epoch": 0.45042998539672235, "step": 1388, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.188968761009164e-06, "weight_min": 0.0, "weight_std": 1.735533805913292e-05 }, { "epoch": 0.45042998539672235, "step": 1388, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4193081369739957e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9613948097685352e-05 }, { "epoch": 0.45042998539672235, "step": 1388, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5540603271801956e-05, "weight_min": 0.0, "weight_std": 1.9491262719384395e-05 }, { "epoch": 0.45042998539672235, "step": 1388, "weight_max": 3.813097282545641e-05, "weight_mean": 1.201626582769677e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.708145282464102e-05 }, { "epoch": 0.4507545026772676, "step": 1389, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0393086085969117e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.8860551790567115e-05 }, { "epoch": 0.4507545026772676, "step": 1389, "weight_max": 3.0886084800840763e-09, "weight_mean": 1.3203872439859765e-09, "weight_min": 6.586902134841921e-10, "weight_std": 8.863210720910786e-10 }, { "epoch": 0.4507545026772676, "step": 1389, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4882685718475841e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0558247342705727e-05 }, { "epoch": 0.4507545026772676, "step": 1389, "weight_max": 6.345375709315704e-07, "weight_mean": 7.981464023032458e-08, "weight_min": 0.0, "weight_std": 2.2414205602672155e-07 }, { "epoch": 0.45107901995781274, "grad_norm": 13.871344566345215, "learning_rate": 7.375510204081633e-06, "loss": 0.1861, "step": 1390 }, { "epoch": 0.45107901995781274, "step": 1390, "weight_max": 3.312147964606993e-05, "weight_mean": 7.96772201283602e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4766080312256236e-05 }, { "epoch": 0.45107901995781274, "step": 1390, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9144959878758527e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0632911400753073e-05 }, { "epoch": 0.45107901995781274, "step": 1390, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960562238120474e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.845694896474015e-05 }, { "epoch": 0.45107901995781274, "step": 1390, "weight_max": 2.202445102739148e-05, "weight_mean": 2.757625679805642e-06, "weight_min": 6.196655966128617e-10, "weight_std": 7.784979970892891e-06 }, { "epoch": 0.45140353723835797, "step": 1391, "weight_max": 3.4268457966391e-05, "weight_mean": 8.11099471320631e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5048139175632969e-05 }, { "epoch": 0.45140353723835797, "step": 1391, "weight_max": 4.252122016623616e-05, "weight_mean": 2.003967165364884e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1459432900883257e-05 }, { "epoch": 0.45140353723835797, "step": 1391, "weight_max": 3.0339444492710754e-05, "weight_mean": 7.586009814986028e-06, "weight_min": 0.0, "weight_std": 1.4043725059309509e-05 }, { "epoch": 0.45140353723835797, "step": 1391, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0281824870617129e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7087637388613075e-05 }, { "epoch": 0.45172805451890313, "step": 1392, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3582191968453117e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8896154870162718e-05 }, { "epoch": 0.45172805451890313, "step": 1392, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.610283996153157e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.779652484401595e-05 }, { "epoch": 0.45172805451890313, "step": 1392, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316131137078628e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5033126146590803e-05 }, { "epoch": 0.45172805451890313, "step": 1392, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3902817045163829e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9574494217522442e-05 }, { "epoch": 0.4520525717994483, "step": 1393, "weight_max": 3.782653220696375e-05, "weight_mean": 1.1250616807956249e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.6125897673191503e-05 }, { "epoch": 0.4520525717994483, "step": 1393, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.4168432244332507e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.5987267033779062e-05 }, { "epoch": 0.4520525717994483, "step": 1393, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315605449141003e-06, "weight_min": 0.0, "weight_std": 1.5033338058856316e-05 }, { "epoch": 0.4520525717994483, "step": 1393, "weight_max": 2.4160392058547586e-05, "weight_mean": 3.2998350434354506e-06, "weight_min": 6.053385570581327e-10, "weight_std": 8.465054634143598e-06 }, { "epoch": 0.4523770890799935, "step": 1394, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793158384723938e-06, "weight_min": 0.0, "weight_std": 1.0726319487730507e-05 }, { "epoch": 0.4523770890799935, "step": 1394, "weight_max": 2.1057827481740787e-08, "weight_mean": 3.59510066161306e-09, "weight_min": 5.679522963042416e-10, "weight_std": 7.11751724224996e-09 }, { "epoch": 0.4523770890799935, "step": 1394, "weight_max": 3.0886087188264355e-05, "weight_mean": 5.887589395570103e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1584427738853265e-05 }, { "epoch": 0.4523770890799935, "step": 1394, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4196544725564308e-05, "weight_min": 0.0, "weight_std": 1.9641451217466965e-05 }, { "epoch": 0.4527016063605387, "step": 1395, "weight_max": 3.006794759130571e-05, "weight_mean": 4.514152806223137e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.0539339200477116e-05 }, { "epoch": 0.4527016063605387, "step": 1395, "weight_max": 1.0068959888087647e-08, "weight_mean": 1.944507221907088e-09, "weight_min": 0.0, "weight_std": 3.3069060823009977e-09 }, { "epoch": 0.4527016063605387, "step": 1395, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3769837096333504e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9164599507348612e-05 }, { "epoch": 0.4527016063605387, "step": 1395, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3473648323270027e-05, "weight_min": 0.0, "weight_std": 1.8763828848022968e-05 }, { "epoch": 0.4530261236410839, "step": 1396, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.804124258342199e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0180392311885953e-05 }, { "epoch": 0.4530261236410839, "step": 1396, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0037372703664005e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.859330768638756e-05 }, { "epoch": 0.4530261236410839, "step": 1396, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4961695342208259e-05, "weight_min": 0.0, "weight_std": 2.0676316125900485e-05 }, { "epoch": 0.4530261236410839, "step": 1396, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.752877536404412e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.498286746937083e-05 }, { "epoch": 0.4533506409216291, "step": 1397, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9776940462179482e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1372170522226952e-05 }, { "epoch": 0.4533506409216291, "step": 1397, "weight_max": 3.813097282545641e-05, "weight_mean": 8.979759513749741e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6665057046338916e-05 }, { "epoch": 0.4533506409216291, "step": 1397, "weight_max": 3.998322426923551e-05, "weight_mean": 1.83185184141621e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9770181097555906e-05 }, { "epoch": 0.4533506409216291, "step": 1397, "weight_max": 3.782653220696375e-05, "weight_mean": 9.41928919928614e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.743994653224945e-05 }, { "epoch": 0.45367515820217424, "step": 1398, "weight_max": 3.998322426923551e-05, "weight_mean": 9.614378541300539e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.781770697562024e-05 }, { "epoch": 0.45367515820217424, "step": 1398, "weight_max": 1.063030504155904e-05, "weight_mean": 1.3295619964992511e-06, "weight_min": 5.771870759119224e-10, "weight_std": 3.758067805392784e-06 }, { "epoch": 0.45367515820217424, "step": 1398, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4530491171171889e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0057383153471164e-05 }, { "epoch": 0.45367515820217424, "step": 1398, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1351365174050443e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6685120499460027e-05 }, { "epoch": 0.45399967548271947, "step": 1399, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0472859685251024e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9402608813834377e-05 }, { "epoch": 0.45399967548271947, "step": 1399, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1683020602504257e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.7769307305570692e-05 }, { "epoch": 0.45399967548271947, "step": 1399, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.111086001212243e-06, "weight_min": 0.0, "weight_std": 1.701564178802073e-05 }, { "epoch": 0.45399967548271947, "step": 1399, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4700301107950509e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0346853489172645e-05 }, { "epoch": 0.45432419276326463, "grad_norm": 11.233059883117676, "learning_rate": 7.355102040816328e-06, "loss": 0.0874, "step": 1400 }, { "epoch": 0.45432419276326463, "eval_accuracy": 0.9875, "eval_auc": 0.9944942071658319, "eval_loss": 0.05588250607252121, "eval_runtime": 154.9896, "eval_samples_per_second": 12.904, "eval_steps_per_second": 1.613, "step": 1400 }, { "epoch": 0.45432419276326463, "step": 1400, "weight_max": 4.187939703115262e-05, "weight_mean": 9.591592970537022e-06, "weight_min": 0.0, "weight_std": 1.785803397069685e-05 }, { "epoch": 0.45432419276326463, "step": 1400, "weight_max": 3.9360926166409627e-05, "weight_mean": 7.971138984430581e-06, "weight_min": 0.0, "weight_std": 1.5290130249923095e-05 }, { "epoch": 0.45432419276326463, "step": 1400, "weight_max": 3.255531191825867e-05, "weight_mean": 7.828686648281291e-06, "weight_min": 0.0, "weight_std": 1.4509224456560332e-05 }, { "epoch": 0.45432419276326463, "step": 1400, "weight_max": 1.407247367524178e-07, "weight_mean": 1.9191674027752015e-08, "weight_min": 6.29309993005478e-10, "weight_std": 4.912564577352896e-08 }, { "epoch": 0.45464871004380986, "step": 1401, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0623303498723544e-05, "weight_min": 0.0, "weight_std": 2.205460623372346e-05 }, { "epoch": 0.45464871004380986, "step": 1401, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5262123674328905e-06, "weight_min": 5.958802340444436e-10, "weight_std": 9.970388418878429e-06 }, { "epoch": 0.45464871004380986, "step": 1401, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232520253339317e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8945447664009407e-05 }, { "epoch": 0.45464871004380986, "step": 1401, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5214762242976576e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1103320250404067e-05 }, { "epoch": 0.454973227324355, "step": 1402, "weight_max": 3.22740561387036e-05, "weight_mean": 4.067766894877423e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.1397254638723098e-05 }, { "epoch": 0.454973227324355, "step": 1402, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.937171853380278e-06, "weight_min": 7.298507354036587e-10, "weight_std": 1.8488704881747253e-05 }, { "epoch": 0.454973227324355, "step": 1402, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6847936421982013e-05, "weight_min": 8.716874999592505e-10, "weight_std": 1.862631870608311e-05 }, { "epoch": 0.454973227324355, "step": 1402, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.944143878063187e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0908761143800803e-05 }, { "epoch": 0.4552977446049002, "step": 1403, "weight_max": 1.4159473593622351e-09, "weight_mean": 8.150926045225049e-10, "weight_min": 0.0, "weight_std": 4.5398057535450675e-10 }, { "epoch": 0.4552977446049002, "step": 1403, "weight_max": 3.602553260861896e-05, "weight_mean": 1.1885851563420147e-05, "weight_min": 0.0, "weight_std": 1.652594073675573e-05 }, { "epoch": 0.4552977446049002, "step": 1403, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.892538981948746e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3802377907268237e-05 }, { "epoch": 0.4552977446049002, "step": 1403, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0258709153276868e-05, "weight_min": 0.0, "weight_std": 1.6875685105333105e-05 }, { "epoch": 0.4556222618854454, "step": 1404, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5081337551237084e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0674058760050684e-05 }, { "epoch": 0.4556222618854454, "step": 1404, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0328452844987623e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9085113308392465e-05 }, { "epoch": 0.4556222618854454, "step": 1404, "weight_max": 4.187939703115262e-05, "weight_mean": 2.030970608757343e-05, "weight_min": 0.0, "weight_std": 2.172955464629922e-05 }, { "epoch": 0.4556222618854454, "step": 1404, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076969500805717e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357676263898611e-05 }, { "epoch": 0.4559467791659906, "step": 1405, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.736469130264595e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8004791854764335e-05 }, { "epoch": 0.4559467791659906, "step": 1405, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.4427135940641165e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0474279153859243e-05 }, { "epoch": 0.4559467791659906, "step": 1405, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006297088693827e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.857435199781321e-05 }, { "epoch": 0.4559467791659906, "step": 1405, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9367596905794926e-05, "weight_min": 9.178617865757133e-10, "weight_std": 2.0842864614678547e-05 }, { "epoch": 0.4562712964465358, "step": 1406, "weight_max": 4.252122016623616e-05, "weight_mean": 1.1631144843704533e-05, "weight_min": 1.0888551171817085e-09, "weight_std": 1.9102400983683765e-05 }, { "epoch": 0.4562712964465358, "step": 1406, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9990280786005314e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4135749552224297e-05 }, { "epoch": 0.4562712964465358, "step": 1406, "weight_max": 4.187939703115262e-05, "weight_mean": 2.323784792679362e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9883284039678983e-05 }, { "epoch": 0.4562712964465358, "step": 1406, "weight_max": 2.563926318543963e-05, "weight_mean": 3.205639359293855e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.064552614290733e-06 }, { "epoch": 0.45659581372708097, "step": 1407, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.407904377032537e-06, "weight_min": 0.0, "weight_std": 1.4611186088586692e-05 }, { "epoch": 0.45659581372708097, "step": 1407, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4036802895134315e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9693130525411107e-05 }, { "epoch": 0.45659581372708097, "step": 1407, "weight_max": 3.0886087188264355e-05, "weight_mean": 7.688607183808926e-06, "weight_min": 0.0, "weight_std": 1.4233447473088745e-05 }, { "epoch": 0.45659581372708097, "step": 1407, "weight_max": 1.3363533071242273e-05, "weight_mean": 1.6709119563529384e-06, "weight_min": 0.0, "weight_std": 4.724532118416391e-06 }, { "epoch": 0.45692033100762613, "step": 1408, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471198038430884e-05, "weight_min": 0.0, "weight_std": 1.9390081433812156e-05 }, { "epoch": 0.45692033100762613, "step": 1408, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212407020531828e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1911766705452465e-05 }, { "epoch": 0.45692033100762613, "step": 1408, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3161206778895576e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5033130694064312e-05 }, { "epoch": 0.45692033100762613, "step": 1408, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.573253580834717e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6075548046501353e-05 }, { "epoch": 0.45724484828817136, "step": 1409, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.616189016815042e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.30529315356398e-05 }, { "epoch": 0.45724484828817136, "step": 1409, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.1795324098784477e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.645085467316676e-05 }, { "epoch": 0.45724484828817136, "step": 1409, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7193440726259723e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.863889701780863e-05 }, { "epoch": 0.45724484828817136, "step": 1409, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284201622795081e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.2115468962292653e-05 }, { "epoch": 0.4575693655687165, "grad_norm": 5.186723232269287, "learning_rate": 7.334693877551021e-06, "loss": 0.0827, "step": 1410 }, { "epoch": 0.4575693655687165, "step": 1410, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3704942830372602e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9023273125640117e-05 }, { "epoch": 0.4575693655687165, "step": 1410, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5410156417347025e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2841719581047073e-05 }, { "epoch": 0.4575693655687165, "step": 1410, "weight_max": 3.813097282545641e-05, "weight_mean": 4.7669514060544316e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3481100722856354e-05 }, { "epoch": 0.4575693655687165, "step": 1410, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.66149218584178e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7936492440639995e-05 }, { "epoch": 0.45789388284926175, "step": 1411, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9690338376676664e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1087083950988017e-05 }, { "epoch": 0.45789388284926175, "step": 1411, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3157805268710945e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.503326802776428e-05 }, { "epoch": 0.45789388284926175, "step": 1411, "weight_max": 3.998322426923551e-05, "weight_mean": 9.727228643896524e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.80180595634738e-05 }, { "epoch": 0.45789388284926175, "step": 1411, "weight_max": 1.179825903641074e-09, "weight_mean": 8.1311901656278e-10, "weight_min": 5.958802340444436e-10, "weight_std": 2.1603070721187834e-10 }, { "epoch": 0.4582184001298069, "step": 1412, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844140676141251e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3697413123736624e-05 }, { "epoch": 0.4582184001298069, "step": 1412, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5259743071947014e-06, "weight_min": 5.864964625068581e-10, "weight_std": 9.970484825316817e-06 }, { "epoch": 0.4582184001298069, "step": 1412, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.53088567458326e-05, "weight_min": 0.0, "weight_std": 2.1131056200829335e-05 }, { "epoch": 0.4582184001298069, "step": 1412, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882283974438906e-06, "weight_min": 0.0, "weight_std": 1.3806491551804356e-05 }, { "epoch": 0.4585429174103521, "step": 1413, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.847481123870239e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8256727344123647e-05 }, { "epoch": 0.4585429174103521, "step": 1413, "weight_max": 4.1044416754232316e-09, "weight_mean": 1.1957274059781753e-09, "weight_min": 0.0, "weight_std": 1.2484572264881422e-09 }, { "epoch": 0.4585429174103521, "step": 1413, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232297427137382e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8945584088214673e-05 }, { "epoch": 0.4585429174103521, "step": 1413, "weight_max": 3.397988257347606e-05, "weight_mean": 8.040952707233373e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.491815328336088e-05 }, { "epoch": 0.4588674346908973, "step": 1414, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4274558452598285e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9753482774831355e-05 }, { "epoch": 0.4588674346908973, "step": 1414, "weight_max": 3.782653220696375e-05, "weight_mean": 8.221422831411473e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5449235434061848e-05 }, { "epoch": 0.4588674346908973, "step": 1414, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.789386078598909e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.938283458002843e-05 }, { "epoch": 0.4588674346908973, "step": 1414, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0055182428914122e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8253120288136415e-05 }, { "epoch": 0.45919195197144247, "step": 1415, "weight_max": 1.2650940334424376e-05, "weight_mean": 1.5821179886188474e-06, "weight_min": 5.679522963042416e-10, "weight_std": 4.472479758987902e-06 }, { "epoch": 0.45919195197144247, "step": 1415, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5430406316882e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2840902854804881e-05 }, { "epoch": 0.45919195197144247, "step": 1415, "weight_max": 4.187939703115262e-05, "weight_mean": 5.243187388259685e-06, "weight_min": 0.0, "weight_std": 1.480327955505345e-05 }, { "epoch": 0.45919195197144247, "step": 1415, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.696126653579995e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8000620912062004e-05 }, { "epoch": 0.4595164692519877, "step": 1416, "weight_max": 3.752332122530788e-05, "weight_mean": 9.381910786032677e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.736925696604885e-05 }, { "epoch": 0.4595164692519877, "step": 1416, "weight_max": 3.843663216684945e-05, "weight_mean": 9.161230082099792e-06, "weight_min": 0.0, "weight_std": 1.698916094028391e-05 }, { "epoch": 0.4595164692519877, "step": 1416, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9091108697466552e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0657029381254688e-05 }, { "epoch": 0.4595164692519877, "step": 1416, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882247139903484e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3806506103719585e-05 }, { "epoch": 0.45984098653253286, "step": 1417, "weight_max": 3.36925259034615e-05, "weight_mean": 4.908102710032836e-06, "weight_min": 7.614652242082798e-10, "weight_std": 1.179243736260105e-05 }, { "epoch": 0.45984098653253286, "step": 1417, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.730006240715738e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8036660549114458e-05 }, { "epoch": 0.45984098653253286, "step": 1417, "weight_max": 3.1715208024252206e-05, "weight_mean": 7.68957852415042e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4246398677641992e-05 }, { "epoch": 0.45984098653253286, "step": 1417, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3855456018063705e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9325845642015338e-05 }, { "epoch": 0.460165503813078, "step": 1418, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0631554687279277e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.968272954400163e-05 }, { "epoch": 0.460165503813078, "step": 1418, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.881558980850969e-06, "weight_min": 0.0, "weight_std": 1.830463770602364e-05 }, { "epoch": 0.460165503813078, "step": 1418, "weight_max": 4.187939703115262e-05, "weight_mean": 9.964082892111037e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.847959174483549e-05 }, { "epoch": 0.460165503813078, "step": 1418, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235607204667758e-06, "weight_min": 0.0, "weight_std": 1.4806327271799091e-05 }, { "epoch": 0.46049002109362325, "step": 1419, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.97246900422033e-06, "weight_min": 0.0, "weight_std": 1.6802152458694763e-05 }, { "epoch": 0.46049002109362325, "step": 1419, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.548832854809007e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4756630662304815e-05 }, { "epoch": 0.46049002109362325, "step": 1419, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4574780834664125e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0145698726992123e-05 }, { "epoch": 0.46049002109362325, "step": 1419, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.8818787541240454e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8104577975464053e-05 }, { "epoch": 0.4608145383741684, "grad_norm": 3.546126365661621, "learning_rate": 7.314285714285715e-06, "loss": 0.3132, "step": 1420 }, { "epoch": 0.4608145383741684, "step": 1420, "weight_max": 3.998322426923551e-05, "weight_mean": 9.318237061961554e-06, "weight_min": 0.0, "weight_std": 1.7313475836999714e-05 }, { "epoch": 0.4608145383741684, "step": 1420, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0078643754241057e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8678905689739622e-05 }, { "epoch": 0.4608145383741684, "step": 1420, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0121488230652176e-05, "weight_min": 0.0, "weight_std": 1.876879105111584e-05 }, { "epoch": 0.4608145383741684, "step": 1420, "weight_max": 3.602553260861896e-05, "weight_mean": 4.516043645708123e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2731798960885499e-05 }, { "epoch": 0.46113905565471364, "step": 1421, "weight_max": 4.252122016623616e-05, "weight_mean": 1.32433533508447e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.909724778670352e-05 }, { "epoch": 0.46113905565471364, "step": 1421, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.377796070300974e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9352384697413072e-05 }, { "epoch": 0.46113905565471364, "step": 1421, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2450999747670721e-05, "weight_min": 0.0, "weight_std": 1.7906231732922606e-05 }, { "epoch": 0.46113905565471364, "step": 1421, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.356895715318387e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2320777386776172e-05 }, { "epoch": 0.4614635729352588, "step": 1422, "weight_max": 3.4849264920921996e-05, "weight_mean": 7.4072813731618226e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3995281733514275e-05 }, { "epoch": 0.4614635729352588, "step": 1422, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0132305508013815e-05, "weight_min": 0.0, "weight_std": 2.1618023311020806e-05 }, { "epoch": 0.4614635729352588, "step": 1422, "weight_max": 3.5434961318969727e-05, "weight_mean": 6.846091309853364e-06, "weight_min": 0.0, "weight_std": 1.3386261343839578e-05 }, { "epoch": 0.4614635729352588, "step": 1422, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235366188571788e-06, "weight_min": 0.0, "weight_std": 1.4806424587732181e-05 }, { "epoch": 0.46178809021580397, "step": 1423, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.608559594198596e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.7968466636375524e-05 }, { "epoch": 0.46178809021580397, "step": 1423, "weight_max": 4.187939703115262e-05, "weight_mean": 5.240210612100782e-06, "weight_min": 0.0, "weight_std": 1.4804471902607474e-05 }, { "epoch": 0.46178809021580397, "step": 1423, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.634997357148677e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6313595551764593e-05 }, { "epoch": 0.46178809021580397, "step": 1423, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.286485323566012e-05, "weight_min": 0.0, "weight_std": 1.9303861336084083e-05 }, { "epoch": 0.4621126074963492, "step": 1424, "weight_max": 3.088609048518265e-07, "weight_mean": 3.9248483574283455e-08, "weight_min": 0.0, "weight_std": 1.089404051413112e-07 }, { "epoch": 0.4621126074963492, "step": 1424, "weight_max": 7.403143098549947e-10, "weight_mean": 6.537539398721037e-10, "weight_min": 5.958802340444436e-10, "weight_std": 6.147566483649669e-11 }, { "epoch": 0.4621126074963492, "step": 1424, "weight_max": 3.998322426923551e-05, "weight_mean": 9.765202776179649e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8086126146954484e-05 }, { "epoch": 0.4621126074963492, "step": 1424, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.326290299009997e-06, "weight_min": 0.0, "weight_std": 1.6292706277454272e-05 }, { "epoch": 0.46243712477689436, "step": 1425, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4089485375734512e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9664972569444217e-05 }, { "epoch": 0.46243712477689436, "step": 1425, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8741855456028134e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0079933165106922e-05 }, { "epoch": 0.46243712477689436, "step": 1425, "weight_max": 3.632264997577295e-05, "weight_mean": 8.992067705548834e-06, "weight_min": 0.0, "weight_std": 1.5625886589987203e-05 }, { "epoch": 0.46243712477689436, "step": 1425, "weight_max": 2.6347608539367684e-09, "weight_mean": 1.181308384445856e-09, "weight_min": 0.0, "weight_std": 9.083764296313745e-10 }, { "epoch": 0.4627616420574396, "step": 1426, "weight_max": 3.813097282545641e-05, "weight_mean": 1.7119120457209647e-05, "weight_min": 0.0, "weight_std": 1.8388565877103247e-05 }, { "epoch": 0.4627616420574396, "step": 1426, "weight_max": 6.694368170911957e-09, "weight_mean": 1.4189844854683997e-09, "weight_min": 0.0, "weight_std": 2.159592504824559e-09 }, { "epoch": 0.4627616420574396, "step": 1426, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3908591427025385e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.946255906659644e-05 }, { "epoch": 0.4627616420574396, "step": 1426, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.0202885277976748e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.4958536667109001e-05 }, { "epoch": 0.46308615933798475, "step": 1427, "weight_max": 4.252122016623616e-05, "weight_mean": 9.893380592984613e-06, "weight_min": 0.0, "weight_std": 1.838523894548416e-05 }, { "epoch": 0.46308615933798475, "step": 1427, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.1219086193013936e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.944517862284556e-05 }, { "epoch": 0.46308615933798475, "step": 1427, "weight_max": 4.252122016623616e-05, "weight_mean": 8.775802598393057e-06, "weight_min": 0.0, "weight_std": 1.672503276495263e-05 }, { "epoch": 0.46308615933798475, "step": 1427, "weight_max": 4.252122016623616e-05, "weight_mean": 1.966059789992869e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.109545675921254e-05 }, { "epoch": 0.4634106766185299, "step": 1428, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.712971975910477e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6131669326568954e-05 }, { "epoch": 0.4634106766185299, "step": 1428, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5626837921445258e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1572332116193138e-05 }, { "epoch": 0.4634106766185299, "step": 1428, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0183313861489296e-05, "weight_min": 0.0, "weight_std": 1.8852313587558456e-05 }, { "epoch": 0.4634106766185299, "step": 1428, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236385277385125e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806012586632278e-05 }, { "epoch": 0.46373519389907514, "step": 1429, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.409746442339383e-06, "weight_min": 0.0, "weight_std": 1.6007206795620732e-05 }, { "epoch": 0.46373519389907514, "step": 1429, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.742229505907744e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0156250684522092e-05 }, { "epoch": 0.46373519389907514, "step": 1429, "weight_max": 9.652274535199012e-10, "weight_mean": 6.878319025460655e-10, "weight_min": 5.679522963042416e-10, "weight_std": 1.2836616491984643e-10 }, { "epoch": 0.46373519389907514, "step": 1429, "weight_max": 3.998322426923551e-05, "weight_mean": 9.838780897553079e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.813324161048513e-05 }, { "epoch": 0.4640597111796203, "grad_norm": 5.577298641204834, "learning_rate": 7.2938775510204086e-06, "loss": 0.1564, "step": 1430 }, { "epoch": 0.4640597111796203, "step": 1430, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320732841733843e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2217803487146739e-05 }, { "epoch": 0.4640597111796203, "step": 1430, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3003488675167318e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.829320353863295e-05 }, { "epoch": 0.4640597111796203, "step": 1430, "weight_max": 4.252122016623616e-05, "weight_mean": 9.420255082659423e-06, "weight_min": 0.0, "weight_std": 1.7632941307965666e-05 }, { "epoch": 0.4640597111796203, "step": 1430, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578248081088532e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2947221875947434e-05 }, { "epoch": 0.4643842284601655, "step": 1431, "weight_max": 4.252122016623616e-05, "weight_mean": 8.885868737706915e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6783696992206387e-05 }, { "epoch": 0.4643842284601655, "step": 1431, "weight_max": 8.716874999592505e-10, "weight_mean": 6.232054317045765e-10, "weight_min": 0.0, "weight_std": 2.7668828339599827e-10 }, { "epoch": 0.4643842284601655, "step": 1431, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2219892596476711e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.758715370669961e-05 }, { "epoch": 0.4643842284601655, "step": 1431, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3940155440650415e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7515316358185373e-05 }, { "epoch": 0.4647087457407107, "step": 1432, "weight_max": 4.187939703115262e-05, "weight_mean": 9.92585682979552e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8414677469991148e-05 }, { "epoch": 0.4647087457407107, "step": 1432, "weight_max": 4.187939703115262e-05, "weight_mean": 1.482892366766464e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.9815912310150452e-05 }, { "epoch": 0.4647087457407107, "step": 1432, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2413995136739686e-05, "weight_min": 0.0, "weight_std": 1.7547226889291778e-05 }, { "epoch": 0.4647087457407107, "step": 1432, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6615891581750475e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.883613549580332e-05 }, { "epoch": 0.46503326302125586, "step": 1433, "weight_max": 2.716206836339552e-05, "weight_mean": 3.3958981475734618e-06, "weight_min": 0.0, "weight_std": 9.602982572687324e-06 }, { "epoch": 0.46503326302125586, "step": 1433, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212056865071645e-06, "weight_min": 0.0, "weight_std": 1.191190767713124e-05 }, { "epoch": 0.46503326302125586, "step": 1433, "weight_max": 3.752332122530788e-05, "weight_mean": 9.343502824776806e-06, "weight_min": 0.0, "weight_std": 1.729995165078435e-05 }, { "epoch": 0.46503326302125586, "step": 1433, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0551861123531125e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9543633243301883e-05 }, { "epoch": 0.4653577803018011, "step": 1434, "weight_max": 3.752332122530788e-05, "weight_mean": 1.3186545402277261e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8233327864436433e-05 }, { "epoch": 0.4653577803018011, "step": 1434, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430155058798846e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2527832950581796e-05 }, { "epoch": 0.4653577803018011, "step": 1434, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4236366041586734e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9860462998622097e-05 }, { "epoch": 0.4653577803018011, "step": 1434, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.568948942411225e-06, "weight_min": 0.0, "weight_std": 1.5872246876824647e-05 }, { "epoch": 0.46568229758234625, "step": 1435, "weight_max": 4.252122016623616e-05, "weight_mean": 2.4276640033349395e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0202522136969492e-05 }, { "epoch": 0.46568229758234625, "step": 1435, "weight_max": 4.187939703115262e-05, "weight_mean": 9.807368769543245e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8141536202165298e-05 }, { "epoch": 0.46568229758234625, "step": 1435, "weight_max": 3.143761568935588e-05, "weight_mean": 6.654393473581877e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2586957382154651e-05 }, { "epoch": 0.46568229758234625, "step": 1435, "weight_max": 4.187939703115262e-05, "weight_mean": 9.963825505110435e-06, "weight_min": 0.0, "weight_std": 1.8479748177924193e-05 }, { "epoch": 0.46600681486289147, "step": 1436, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998505573894363e-06, "weight_min": 0.0, "weight_std": 1.4135960554995108e-05 }, { "epoch": 0.46600681486289147, "step": 1436, "weight_max": 3.998322426923551e-05, "weight_mean": 1.9582832464948297e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8148433809983544e-05 }, { "epoch": 0.46600681486289147, "step": 1436, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236154265730875e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4806105355091859e-05 }, { "epoch": 0.46600681486289147, "step": 1436, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827736236416968e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0822541298693977e-05 }, { "epoch": 0.46633133214343664, "step": 1437, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0049120646726806e-05, "weight_min": 0.0, "weight_std": 1.867395985755138e-05 }, { "epoch": 0.46633133214343664, "step": 1437, "weight_max": 3.2068148811958963e-06, "weight_mean": 4.0154054659069516e-07, "weight_min": 6.196655966128617e-10, "weight_std": 1.1335019962643855e-06 }, { "epoch": 0.46633133214343664, "step": 1437, "weight_max": 4.381950566312298e-05, "weight_mean": 1.7400278011336923e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.99158675968647e-05 }, { "epoch": 0.46633133214343664, "step": 1437, "weight_max": 3.752332122530788e-05, "weight_mean": 9.343846613774076e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.7299738829024136e-05 }, { "epoch": 0.4666558494239818, "step": 1438, "weight_max": 9.413956281179026e-10, "weight_mean": 6.94401924850041e-10, "weight_min": 6.053385570581327e-10, "weight_std": 1.1424183127273224e-10 }, { "epoch": 0.4666558494239818, "step": 1438, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0552170351729728e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9543444068403915e-05 }, { "epoch": 0.4666558494239818, "step": 1438, "weight_max": 6.10095594311133e-06, "weight_mean": 7.634646976839576e-07, "weight_min": 0.0, "weight_std": 2.1566722807619954e-06 }, { "epoch": 0.4666558494239818, "step": 1438, "weight_max": 3.199401908204891e-05, "weight_mean": 4.000125954917166e-06, "weight_min": 0.0, "weight_std": 1.1311240996292327e-05 }, { "epoch": 0.46698036670452703, "step": 1439, "weight_max": 3.967146767536178e-05, "weight_mean": 8.889189302863088e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.6604679331067018e-05 }, { "epoch": 0.46698036670452703, "step": 1439, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2486331797845196e-05, "weight_min": 0.0, "weight_std": 1.7850505173555575e-05 }, { "epoch": 0.46698036670452703, "step": 1439, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236191100266296e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.480609080317663e-05 }, { "epoch": 0.46698036670452703, "step": 1439, "weight_max": 3.312147964606993e-05, "weight_mean": 1.1286611879768316e-05, "weight_min": 0.0, "weight_std": 1.5762860130053014e-05 }, { "epoch": 0.4673048839850722, "grad_norm": 2.0270166397094727, "learning_rate": 7.273469387755102e-06, "loss": 0.2654, "step": 1440 }, { "epoch": 0.4673048839850722, "step": 1440, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5278828868758865e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1129681044840254e-05 }, { "epoch": 0.4673048839850722, "step": 1440, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.770875294634607e-06, "weight_min": 0.0, "weight_std": 1.8127751900465228e-05 }, { "epoch": 0.4673048839850722, "step": 1440, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3347189451451413e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.821010300773196e-05 }, { "epoch": 0.4673048839850722, "step": 1440, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9599900042521767e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1099360310472548e-05 }, { "epoch": 0.4676294012656174, "step": 1441, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920728770230198e-06, "weight_min": 0.0, "weight_std": 1.3915941053710412e-05 }, { "epoch": 0.4676294012656174, "step": 1441, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1211521268705837e-05, "weight_min": 0.0, "weight_std": 1.68379774549976e-05 }, { "epoch": 0.4676294012656174, "step": 1441, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.1309839212335646e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9000444808625616e-05 }, { "epoch": 0.4676294012656174, "step": 1441, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5608246030751616e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0601850337698124e-05 }, { "epoch": 0.4679539185461626, "step": 1442, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4701392501592636e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0522302293102257e-05 }, { "epoch": 0.4679539185461626, "step": 1442, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.003724719339516e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8593385902931914e-05 }, { "epoch": 0.4679539185461626, "step": 1442, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.899300155462697e-06, "weight_min": 0.0, "weight_std": 1.6491874703206122e-05 }, { "epoch": 0.4679539185461626, "step": 1442, "weight_max": 2.7939937353949063e-05, "weight_mean": 3.493071517368662e-06, "weight_min": 5.771870759119224e-10, "weight_std": 9.87802559393458e-06 }, { "epoch": 0.46827843582670775, "step": 1443, "weight_max": 4.187939703115262e-05, "weight_mean": 9.555040378472768e-06, "weight_min": 0.0, "weight_std": 1.7799546185415238e-05 }, { "epoch": 0.46827843582670775, "step": 1443, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.3573440962063614e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2320596397330519e-05 }, { "epoch": 0.46827843582670775, "step": 1443, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4762264981982298e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0541696358122863e-05 }, { "epoch": 0.46827843582670775, "step": 1443, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3027058230363764e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8284847101313062e-05 }, { "epoch": 0.468602953107253, "step": 1444, "weight_max": 4.187939703115262e-05, "weight_mean": 1.7970312910620123e-05, "weight_min": 0.0, "weight_std": 1.7212267266586423e-05 }, { "epoch": 0.468602953107253, "step": 1444, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.404499835189199e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9428493033046834e-05 }, { "epoch": 0.468602953107253, "step": 1444, "weight_max": 3.5141503758495674e-05, "weight_mean": 6.394901902240235e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2895903637399897e-05 }, { "epoch": 0.468602953107253, "step": 1444, "weight_max": 4.252122016623616e-05, "weight_mean": 1.473516476835357e-05, "weight_min": 0.0, "weight_std": 2.039451828750316e-05 }, { "epoch": 0.46892747038779814, "step": 1445, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4310271581052803e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0107640011701733e-05 }, { "epoch": 0.46892747038779814, "step": 1445, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3288545233081095e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8784494386636652e-05 }, { "epoch": 0.46892747038779814, "step": 1445, "weight_max": 3.998322426923551e-05, "weight_mean": 9.625890015740879e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7722666598274373e-05 }, { "epoch": 0.46892747038779814, "step": 1445, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.405990042258054e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.790895294107031e-05 }, { "epoch": 0.46925198766834336, "step": 1446, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653663836506894e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.315932058787439e-05 }, { "epoch": 0.46925198766834336, "step": 1446, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550602382863872e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.953537685039919e-05 }, { "epoch": 0.46925198766834336, "step": 1446, "weight_max": 3.782653220696375e-05, "weight_mean": 1.517025702924002e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7518994354759343e-05 }, { "epoch": 0.46925198766834336, "step": 1446, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3385124222841114e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9248212993261404e-05 }, { "epoch": 0.46957650494888853, "step": 1447, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9193350453861058e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.057420897472184e-05 }, { "epoch": 0.46957650494888853, "step": 1447, "weight_max": 2.952862632810138e-05, "weight_mean": 3.6952317259419942e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.04382716017426e-05 }, { "epoch": 0.46957650494888853, "step": 1447, "weight_max": 3.722132169059478e-05, "weight_mean": 9.532293915981427e-06, "weight_min": 0.0, "weight_std": 1.4781843674427364e-05 }, { "epoch": 0.46957650494888853, "step": 1447, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8464092136127874e-05, "weight_min": 0.0, "weight_std": 2.006779504881706e-05 }, { "epoch": 0.4699010222294337, "step": 1448, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3537714039557613e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8976759747602046e-05 }, { "epoch": 0.4699010222294337, "step": 1448, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3380437849264126e-05, "weight_min": 0.0, "weight_std": 1.925251490320079e-05 }, { "epoch": 0.4699010222294337, "step": 1448, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4020828530192375e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9611019524745643e-05 }, { "epoch": 0.4699010222294337, "step": 1448, "weight_max": 4.187939703115262e-05, "weight_mean": 2.4440621928079054e-05, "weight_min": 0.0, "weight_std": 2.0379869965836406e-05 }, { "epoch": 0.4702255395099789, "step": 1449, "weight_max": 4.252122016623616e-05, "weight_mean": 1.414200414728839e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.9954632080043666e-05 }, { "epoch": 0.4702255395099789, "step": 1449, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691024059866322e-06, "weight_min": 0.0, "weight_std": 1.3266251698951237e-05 }, { "epoch": 0.4702255395099789, "step": 1449, "weight_max": 4.887017894361634e-06, "weight_mean": 1.0551231071076472e-06, "weight_min": 0.0, "weight_std": 1.9782883100560866e-06 }, { "epoch": 0.4702255395099789, "step": 1449, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.512208634987473e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7694046618998982e-05 }, { "epoch": 0.4705500567905241, "grad_norm": 0.2681106925010681, "learning_rate": 7.253061224489797e-06, "loss": 0.0652, "step": 1450 }, { "epoch": 0.4705500567905241, "step": 1450, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2891363439848647e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8244343664264306e-05 }, { "epoch": 0.4705500567905241, "step": 1450, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.120198228629306e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7077014490496367e-05 }, { "epoch": 0.4705500567905241, "step": 1450, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311800906492863e-05, "weight_min": 0.0, "weight_std": 1.9090817659161985e-05 }, { "epoch": 0.4705500567905241, "step": 1450, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315907401381992e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5033216186566278e-05 }, { "epoch": 0.4708745740710693, "step": 1451, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959614670951851e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4025706150277983e-05 }, { "epoch": 0.4708745740710693, "step": 1451, "weight_max": 3.998322426923551e-05, "weight_mean": 1.5839264960959554e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8881728465203196e-05 }, { "epoch": 0.4708745740710693, "step": 1451, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8303078832104802e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9690594854182564e-05 }, { "epoch": 0.4708745740710693, "step": 1451, "weight_max": 4.187939703115262e-05, "weight_mean": 9.73878923105076e-06, "weight_min": 0.0, "weight_std": 1.8098819055012427e-05 }, { "epoch": 0.4711990913516145, "step": 1452, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5707637430750765e-05, "weight_min": 0.0, "weight_std": 2.1688441847800277e-05 }, { "epoch": 0.4711990913516145, "step": 1452, "weight_max": 1.4159473593622351e-09, "weight_mean": 7.034494098334676e-10, "weight_min": 0.0, "weight_std": 3.827171912718086e-10 }, { "epoch": 0.4711990913516145, "step": 1452, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3618098819279112e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9099108612863347e-05 }, { "epoch": 0.4711990913516145, "step": 1452, "weight_max": 4.187939703115262e-05, "weight_mean": 1.391133355355123e-05, "weight_min": 0.0, "weight_std": 1.9325876564835198e-05 }, { "epoch": 0.47152360863215964, "step": 1453, "weight_max": 4.252122016623616e-05, "weight_mean": 3.3733937016222626e-05, "weight_min": 0.0, "weight_std": 1.4595340871892404e-05 }, { "epoch": 0.47152360863215964, "step": 1453, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236349807091756e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4806027138547506e-05 }, { "epoch": 0.47152360863215964, "step": 1453, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4283437849371694e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9809342120424844e-05 }, { "epoch": 0.47152360863215964, "step": 1453, "weight_max": 3.967146767536178e-05, "weight_mean": 4.9600880629441235e-06, "weight_min": 0.0, "weight_std": 1.4025515156390611e-05 }, { "epoch": 0.47184812591270486, "step": 1454, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.807895028439816e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.641550443309825e-05 }, { "epoch": 0.47184812591270486, "step": 1454, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.739550139580388e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6305320968967862e-05 }, { "epoch": 0.47184812591270486, "step": 1454, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0296134860254824e-05, "weight_min": 0.0, "weight_std": 1.7863167158793658e-05 }, { "epoch": 0.47184812591270486, "step": 1454, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0120364095200785e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.876948408607859e-05 }, { "epoch": 0.47217264319325003, "step": 1455, "weight_max": 3.967146767536178e-05, "weight_mean": 9.80261029326357e-06, "weight_min": 0.0, "weight_std": 1.8150778487324715e-05 }, { "epoch": 0.47217264319325003, "step": 1455, "weight_max": 2.4160392058547586e-05, "weight_mean": 3.0206201699911617e-06, "weight_min": 5.771870759119224e-10, "weight_std": 8.54175777931232e-06 }, { "epoch": 0.47217264319325003, "step": 1455, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4655237464467064e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.025943285843823e-05 }, { "epoch": 0.47217264319325003, "step": 1455, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311361620551907e-05, "weight_min": 0.0, "weight_std": 1.9091088688583113e-05 }, { "epoch": 0.47249716047379525, "step": 1456, "weight_max": 4.187939703115262e-05, "weight_mean": 1.538838114356622e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.123970443790313e-05 }, { "epoch": 0.47249716047379525, "step": 1456, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2130234608775936e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.769896516634617e-05 }, { "epoch": 0.47249716047379525, "step": 1456, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.881489520717878e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.655121195653919e-05 }, { "epoch": 0.47249716047379525, "step": 1456, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3232387320604175e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.957330823643133e-05 }, { "epoch": 0.4728216777543404, "step": 1457, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6433190467068925e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9058206817135215e-05 }, { "epoch": 0.4728216777543404, "step": 1457, "weight_max": 4.252122016623616e-05, "weight_mean": 9.49158493313007e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.774147676769644e-05 }, { "epoch": 0.4728216777543404, "step": 1457, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4628614735556766e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.028472772508394e-05 }, { "epoch": 0.4728216777543404, "step": 1457, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8561426259111613e-05, "weight_min": 7.614652242082798e-10, "weight_std": 2.036380465142429e-05 }, { "epoch": 0.4731461950348856, "step": 1458, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861466211674269e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0919596206804272e-05 }, { "epoch": 0.4731461950348856, "step": 1458, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8746210116660222e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0683648472186178e-05 }, { "epoch": 0.4731461950348856, "step": 1458, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236396191321546e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4806008039158769e-05 }, { "epoch": 0.4731461950348856, "step": 1458, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3454298823489808e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.896340290841181e-05 }, { "epoch": 0.4734707123154308, "step": 1459, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.964923962485045e-06, "weight_min": 0.0, "weight_std": 1.1212807294214144e-05 }, { "epoch": 0.4734707123154308, "step": 1459, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.065969238174148e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8168922906625085e-05 }, { "epoch": 0.4734707123154308, "step": 1459, "weight_max": 3.998322426923551e-05, "weight_mean": 8.859137778927106e-06, "weight_min": 0.0, "weight_std": 1.6581978343310766e-05 }, { "epoch": 0.4734707123154308, "step": 1459, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.687032616056968e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7938566088560037e-05 }, { "epoch": 0.473795229595976, "grad_norm": 10.361928939819336, "learning_rate": 7.23265306122449e-06, "loss": 0.0254, "step": 1460 }, { "epoch": 0.473795229595976, "step": 1460, "weight_max": 3.752332122530788e-05, "weight_mean": 1.7689195374259725e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8936105334432796e-05 }, { "epoch": 0.473795229595976, "step": 1460, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4685996575281024e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.027949994953815e-05 }, { "epoch": 0.473795229595976, "step": 1460, "weight_max": 3.255531191825867e-05, "weight_mean": 7.562427072116407e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4055836800253019e-05 }, { "epoch": 0.473795229595976, "step": 1460, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.152083834516816e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8768127119983546e-05 }, { "epoch": 0.4741197468765212, "step": 1461, "weight_max": 6.806617136589921e-08, "weight_mean": 1.2556759187987154e-08, "weight_min": 6.29309993005478e-10, "weight_std": 2.3363229573192257e-08 }, { "epoch": 0.4741197468765212, "step": 1461, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235527169134002e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806360013608355e-05 }, { "epoch": 0.4741197468765212, "step": 1461, "weight_max": 3.4268457966391e-05, "weight_mean": 8.21389039629139e-06, "weight_min": 0.0, "weight_std": 1.5226420146063901e-05 }, { "epoch": 0.4741197468765212, "step": 1461, "weight_max": 4.187939703115262e-05, "weight_mean": 7.996203748916741e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5709187209722586e-05 }, { "epoch": 0.47444426415706636, "step": 1462, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2593857718457002e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7704078345559537e-05 }, { "epoch": 0.47444426415706636, "step": 1462, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.862617844191846e-06, "weight_min": 0.0, "weight_std": 1.836886440287344e-05 }, { "epoch": 0.47444426415706636, "step": 1462, "weight_max": 1.8471255316399038e-05, "weight_mean": 2.3114982923289062e-06, "weight_min": 6.29309993005478e-10, "weight_std": 6.52952849122812e-06 }, { "epoch": 0.47444426415706636, "step": 1462, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3188887250144035e-06, "weight_min": 0.0, "weight_std": 1.5032013834570535e-05 }, { "epoch": 0.47476878143761153, "step": 1463, "weight_max": 3.722132169059478e-05, "weight_mean": 8.990015885501634e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5974721463862807e-05 }, { "epoch": 0.47476878143761153, "step": 1463, "weight_max": 2.2902415253156505e-07, "weight_mean": 2.9335762263826837e-08, "weight_min": 5.864964625068581e-10, "weight_std": 8.068685986017954e-08 }, { "epoch": 0.47476878143761153, "step": 1463, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5259133710496826e-06, "weight_min": 5.864964625068581e-10, "weight_std": 9.970509381673764e-06 }, { "epoch": 0.47476878143761153, "step": 1463, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0121097147930413e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8769032976706512e-05 }, { "epoch": 0.47509329871815675, "step": 1464, "weight_max": 3.632264997577295e-05, "weight_mean": 7.622572411492001e-06, "weight_min": 0.0, "weight_std": 1.4452153664024081e-05 }, { "epoch": 0.47509329871815675, "step": 1464, "weight_max": 3.36925259034615e-05, "weight_mean": 4.221842573315371e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.1907980479008984e-05 }, { "epoch": 0.47509329871815675, "step": 1464, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.787753252661787e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4210147128324024e-05 }, { "epoch": 0.47509329871815675, "step": 1464, "weight_max": 6.2595857386327225e-09, "weight_mean": 1.5697130262282144e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.996073306642643e-09 }, { "epoch": 0.4754178159987019, "step": 1465, "weight_max": 4.187939703115262e-05, "weight_mean": 1.7121890778071247e-05, "weight_min": 0.0, "weight_std": 1.7980491975322366e-05 }, { "epoch": 0.4754178159987019, "step": 1465, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.793616542679956e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0726134860306047e-05 }, { "epoch": 0.4754178159987019, "step": 1465, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4255056157708168e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.011082870012615e-05 }, { "epoch": 0.4754178159987019, "step": 1465, "weight_max": 8.716874999592505e-10, "weight_mean": 7.422099601583909e-10, "weight_min": 6.053385570581327e-10, "weight_std": 1.137398925044053e-10 }, { "epoch": 0.47574233327924714, "step": 1466, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.20899976536748e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7367454347549938e-05 }, { "epoch": 0.47574233327924714, "step": 1466, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0631471013766713e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.968278047570493e-05 }, { "epoch": 0.47574233327924714, "step": 1466, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653290943679167e-06, "weight_min": 0.0, "weight_std": 1.3159471563994884e-05 }, { "epoch": 0.47574233327924714, "step": 1466, "weight_max": 4.0610404539620504e-05, "weight_mean": 6.425764240702847e-06, "weight_min": 0.0, "weight_std": 1.4319557521957904e-05 }, { "epoch": 0.4760668505597923, "step": 1467, "weight_max": 4.252122016623616e-05, "weight_mean": 2.854095509974286e-05, "weight_min": 0.0, "weight_std": 1.8055206965073012e-05 }, { "epoch": 0.4760668505597923, "step": 1467, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.611400855646934e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7801485228119418e-05 }, { "epoch": 0.4760668505597923, "step": 1467, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7253260011784732e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9313491065986454e-05 }, { "epoch": 0.4760668505597923, "step": 1467, "weight_max": 3.782653220696375e-05, "weight_mean": 4.7684998207842e-06, "weight_min": 0.0, "weight_std": 1.3357768693822436e-05 }, { "epoch": 0.4763913678403375, "step": 1468, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077025889477227e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357653526531067e-05 }, { "epoch": 0.4763913678403375, "step": 1468, "weight_max": 3.998322426923551e-05, "weight_mean": 9.246960871678311e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7193189705722034e-05 }, { "epoch": 0.4763913678403375, "step": 1468, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4849696526653133e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0544031940517016e-05 }, { "epoch": 0.4763913678403375, "step": 1468, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9453866116236895e-05, "weight_min": 0.0, "weight_std": 2.097017022606451e-05 }, { "epoch": 0.4767158851208827, "step": 1469, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.142145240213722e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1400390323833562e-05 }, { "epoch": 0.4767158851208827, "step": 1469, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5707533748354763e-05, "weight_min": 0.0, "weight_std": 2.1688529159291647e-05 }, { "epoch": 0.4767158851208827, "step": 1469, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3429151294985786e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.866387901827693e-05 }, { "epoch": 0.4767158851208827, "step": 1469, "weight_max": 1.7212929970256141e-09, "weight_mean": 8.033651521799356e-10, "weight_min": 0.0, "weight_std": 5.384473422687108e-10 }, { "epoch": 0.47704040240142787, "grad_norm": 3.2379260063171387, "learning_rate": 7.212244897959184e-06, "loss": 0.1586, "step": 1470 }, { "epoch": 0.47704040240142787, "step": 1470, "weight_max": 1.901931279135738e-09, "weight_mean": 9.524085964329743e-10, "weight_min": 6.29309993005478e-10, "weight_std": 4.793878627395998e-10 }, { "epoch": 0.47704040240142787, "step": 1470, "weight_max": 2.846461757144425e-05, "weight_mean": 3.5586606372817187e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0063526133308187e-05 }, { "epoch": 0.47704040240142787, "step": 1470, "weight_max": 3.36925259034615e-05, "weight_mean": 4.21281856688438e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1911601177416742e-05 }, { "epoch": 0.47704040240142787, "step": 1470, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5702818927820772e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.9843851987388916e-05 }, { "epoch": 0.4773649196819731, "step": 1471, "weight_max": 3.9535716496175155e-06, "weight_mean": 4.969683686795179e-07, "weight_min": 5.864964625068581e-10, "weight_std": 1.3966855476610363e-06 }, { "epoch": 0.4773649196819731, "step": 1471, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2602901733771432e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7562764696776867e-05 }, { "epoch": 0.4773649196819731, "step": 1471, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.250920286111068e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.750084811646957e-05 }, { "epoch": 0.4773649196819731, "step": 1471, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3925971870776266e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9233710190746933e-05 }, { "epoch": 0.47768943696251825, "step": 1472, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.894931978720706e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0906477655225899e-05 }, { "epoch": 0.47768943696251825, "step": 1472, "weight_max": 3.813097282545641e-05, "weight_mean": 4.7685766730865e-06, "weight_min": 0.0, "weight_std": 1.3480444067681674e-05 }, { "epoch": 0.47768943696251825, "step": 1472, "weight_max": 4.252122016623616e-05, "weight_mean": 1.848318424890749e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0165132809779607e-05 }, { "epoch": 0.47768943696251825, "step": 1472, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4892804756527767e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0610135834431276e-05 }, { "epoch": 0.4780139542430634, "step": 1473, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4104346519161481e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.6686790331732482e-05 }, { "epoch": 0.4780139542430634, "step": 1473, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3168422810849734e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.843463360273745e-05 }, { "epoch": 0.4780139542430634, "step": 1473, "weight_max": 3.813097282545641e-05, "weight_mean": 4.767194695887156e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.3481000678439159e-05 }, { "epoch": 0.4780139542430634, "step": 1473, "weight_max": 4.187939703115262e-05, "weight_mean": 5.318278454069514e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.4774685951124411e-05 }, { "epoch": 0.47833847152360864, "step": 1474, "weight_max": 3.662098970380612e-05, "weight_mean": 1.2618464097613469e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7494514395366423e-05 }, { "epoch": 0.47833847152360864, "step": 1474, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5471734513994306e-05, "weight_min": 0.0, "weight_std": 2.13686562346993e-05 }, { "epoch": 0.47833847152360864, "step": 1474, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.48795352288289e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0542753190966323e-05 }, { "epoch": 0.47833847152360864, "step": 1474, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4338134860736318e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0239740479155444e-05 }, { "epoch": 0.4786629888041538, "step": 1475, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.809376024350058e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.819174394768197e-05 }, { "epoch": 0.4786629888041538, "step": 1475, "weight_max": 3.782653220696375e-05, "weight_mean": 8.763245205045678e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.6292440705001354e-05 }, { "epoch": 0.4786629888041538, "step": 1475, "weight_max": 6.984984338487266e-06, "weight_mean": 8.743194257476716e-07, "weight_min": 0.0, "weight_std": 2.4690816644579172e-06 }, { "epoch": 0.4786629888041538, "step": 1475, "weight_max": 3.998322426923551e-05, "weight_mean": 9.539433449390344e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.768771471688524e-05 }, { "epoch": 0.47898750608469903, "step": 1476, "weight_max": 2.7549629244560947e-09, "weight_mean": 9.129254574524737e-10, "weight_min": 6.053385570581327e-10, "weight_std": 7.460385642588108e-10 }, { "epoch": 0.47898750608469903, "step": 1476, "weight_max": 4.187939703115262e-05, "weight_mean": 9.200108252116479e-06, "weight_min": 0.0, "weight_std": 1.7248612493858673e-05 }, { "epoch": 0.47898750608469903, "step": 1476, "weight_max": 3.782653220696375e-05, "weight_mean": 8.976423487183638e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6651340047246777e-05 }, { "epoch": 0.47898750608469903, "step": 1476, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.534690434520599e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7660175217315555e-05 }, { "epoch": 0.4793120233652442, "step": 1477, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5044575775391422e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.081198908854276e-05 }, { "epoch": 0.4793120233652442, "step": 1477, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.873428664111998e-05, "weight_min": 0.0, "weight_std": 2.022662192757707e-05 }, { "epoch": 0.4793120233652442, "step": 1477, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.3966559789841995e-06, "weight_min": 0.0, "weight_std": 1.5261895896401256e-05 }, { "epoch": 0.4793120233652442, "step": 1477, "weight_max": 4.252122016623616e-05, "weight_mean": 2.151890475943219e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.7986683815252036e-05 }, { "epoch": 0.47963654064578937, "step": 1478, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3776079867966473e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9034900105907582e-05 }, { "epoch": 0.47963654064578937, "step": 1478, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4493849448626861e-05, "weight_min": 0.0, "weight_std": 2.0016319467686117e-05 }, { "epoch": 0.47963654064578937, "step": 1478, "weight_max": 4.252122016623616e-05, "weight_mean": 6.726462743245065e-06, "weight_min": 0.0, "weight_std": 1.4992511751188431e-05 }, { "epoch": 0.47963654064578937, "step": 1478, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.30500584418769e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8379771063337103e-05 }, { "epoch": 0.4799610579263346, "step": 1479, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3326455700735096e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.872168832051102e-05 }, { "epoch": 0.4799610579263346, "step": 1479, "weight_max": 1.5423825971083716e-05, "weight_mean": 1.9285516827949323e-06, "weight_min": 0.0, "weight_std": 5.452914138004417e-06 }, { "epoch": 0.4799610579263346, "step": 1479, "weight_max": 3.813097282545641e-05, "weight_mean": 1.2149169378972147e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.6975815015030093e-05 }, { "epoch": 0.4799610579263346, "step": 1479, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320452262618346e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.221791717398446e-05 }, { "epoch": 0.48028557520687976, "grad_norm": 15.979942321777344, "learning_rate": 7.1918367346938775e-06, "loss": 0.3045, "step": 1480 }, { "epoch": 0.48028557520687976, "step": 1480, "weight_max": 4.187939703115262e-05, "weight_mean": 1.94450458366191e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0929102902300656e-05 }, { "epoch": 0.48028557520687976, "step": 1480, "weight_max": 3.572963396436535e-05, "weight_mean": 4.466839527594857e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2632076504814904e-05 }, { "epoch": 0.48028557520687976, "step": 1480, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5161525880103e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.096943171636667e-05 }, { "epoch": 0.48028557520687976, "step": 1480, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311116056982428e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9091239664703608e-05 }, { "epoch": 0.480610092487425, "step": 1481, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.276323018828407e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8636390450410545e-05 }, { "epoch": 0.480610092487425, "step": 1481, "weight_max": 4.988156998564364e-09, "weight_mean": 1.2061130982843338e-09, "weight_min": 0.0, "weight_std": 1.564696594513748e-09 }, { "epoch": 0.480610092487425, "step": 1481, "weight_max": 4.187939703115262e-05, "weight_mean": 1.948997851286549e-05, "weight_min": 9.178617865757133e-10, "weight_std": 1.934146166604478e-05 }, { "epoch": 0.480610092487425, "step": 1481, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4162253137328662e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.991546196222771e-05 }, { "epoch": 0.48093460976797014, "step": 1482, "weight_max": 3.752332122530788e-05, "weight_mean": 1.2790764230885543e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7817947082221508e-05 }, { "epoch": 0.48093460976797014, "step": 1482, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0194567039434332e-05, "weight_min": 0.0, "weight_std": 1.88841422641417e-05 }, { "epoch": 0.48093460976797014, "step": 1482, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0008929166360758e-05, "weight_min": 0.0, "weight_std": 1.8572734916233458e-05 }, { "epoch": 0.48093460976797014, "step": 1482, "weight_max": 2.464847420924343e-05, "weight_mean": 3.0815895115665626e-06, "weight_min": 0.0, "weight_std": 8.714337127457839e-06 }, { "epoch": 0.4812591270485153, "step": 1483, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.648635230201762e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6402045730501413e-05 }, { "epoch": 0.4812591270485153, "step": 1483, "weight_max": 1.4159473593622351e-09, "weight_mean": 8.052304933947596e-10, "weight_min": 5.864964625068581e-10, "weight_std": 2.798097031853075e-10 }, { "epoch": 0.4812591270485153, "step": 1483, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4131074749457184e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9751532818190753e-05 }, { "epoch": 0.4812591270485153, "step": 1483, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.808010534266941e-06, "weight_min": 0.0, "weight_std": 1.641543349251151e-05 }, { "epoch": 0.48158364432906053, "step": 1484, "weight_max": 3.782653220696375e-05, "weight_mean": 1.649473051656969e-05, "weight_min": 0.0, "weight_std": 1.852521199907642e-05 }, { "epoch": 0.48158364432906053, "step": 1484, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9740695279324427e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1161255062907003e-05 }, { "epoch": 0.48158364432906053, "step": 1484, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.61629211815307e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7632457456784323e-05 }, { "epoch": 0.48158364432906053, "step": 1484, "weight_max": 3.4558255720185116e-05, "weight_mean": 7.312304660445079e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3828514056513086e-05 }, { "epoch": 0.4819081616096057, "step": 1485, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.9905291739851236e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8625678421813063e-05 }, { "epoch": 0.4819081616096057, "step": 1485, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.090660579298856e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4352193829836324e-05 }, { "epoch": 0.4819081616096057, "step": 1485, "weight_max": 3.116124207735993e-05, "weight_mean": 3.895940153597621e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1016845746780746e-05 }, { "epoch": 0.4819081616096057, "step": 1485, "weight_max": 4.252122016623616e-05, "weight_mean": 2.5245000870199874e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0992152713006362e-05 }, { "epoch": 0.4822326788901509, "step": 1486, "weight_max": 3.0339444492710754e-05, "weight_mean": 1.2987957234145142e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.4596032087865751e-05 }, { "epoch": 0.4822326788901509, "step": 1486, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7540072803967632e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7445910998503678e-05 }, { "epoch": 0.4822326788901509, "step": 1486, "weight_max": 3.312147964606993e-05, "weight_mean": 1.492152841819916e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6044748917920515e-05 }, { "epoch": 0.4822326788901509, "step": 1486, "weight_max": 3.782653220696375e-05, "weight_mean": 1.2341000001470093e-05, "weight_min": 0.0, "weight_std": 1.7560023479745723e-05 }, { "epoch": 0.4825571961706961, "step": 1487, "weight_max": 3.967146767536178e-05, "weight_mean": 1.1013178664143197e-05, "weight_min": 0.0, "weight_std": 1.772157702362165e-05 }, { "epoch": 0.4825571961706961, "step": 1487, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.22559956961777e-06, "weight_min": 0.0, "weight_std": 1.7237398424185812e-05 }, { "epoch": 0.4825571961706961, "step": 1487, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.0397588741616346e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.1612775526591577e-05 }, { "epoch": 0.4825571961706961, "step": 1487, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248234290571418e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2013400009891484e-05 }, { "epoch": 0.48288171345124126, "step": 1488, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3644892533193342e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9508857803884894e-05 }, { "epoch": 0.48288171345124126, "step": 1488, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959680609317729e-06, "weight_min": 0.0, "weight_std": 1.4025679774931632e-05 }, { "epoch": 0.48288171345124126, "step": 1488, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3343717000680044e-05, "weight_min": 0.0, "weight_std": 1.849649561336264e-05 }, { "epoch": 0.48288171345124126, "step": 1488, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729172815132188e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3373351976042613e-05 }, { "epoch": 0.4832062307317865, "step": 1489, "weight_max": 9.062063877074422e-10, "weight_mean": 5.854794427051502e-10, "weight_min": 0.0, "weight_std": 2.5683868920545194e-10 }, { "epoch": 0.4832062307317865, "step": 1489, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550594197411556e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.95353823073674e-05 }, { "epoch": 0.4832062307317865, "step": 1489, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4387082046596333e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9900520783266984e-05 }, { "epoch": 0.4832062307317865, "step": 1489, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9048449757974595e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0598543414962478e-05 }, { "epoch": 0.48353074801233165, "grad_norm": 7.346413612365723, "learning_rate": 7.1714285714285725e-06, "loss": 0.1376, "step": 1490 }, { "epoch": 0.48353074801233165, "step": 1490, "weight_max": 4.447596802492626e-05, "weight_mean": 1.0175188435823657e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8947153876069933e-05 }, { "epoch": 0.48353074801233165, "step": 1490, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882047051069094e-06, "weight_min": 0.0, "weight_std": 1.3806587048748042e-05 }, { "epoch": 0.48353074801233165, "step": 1490, "weight_max": 1.1618659300438594e-05, "weight_mean": 1.463412331759173e-06, "weight_min": 0.0, "weight_std": 4.103443643543869e-06 }, { "epoch": 0.48353074801233165, "step": 1490, "weight_max": 2.919402941614635e-09, "weight_mean": 1.164094376449043e-09, "weight_min": 6.053385570581327e-10, "weight_std": 8.584759569885136e-10 }, { "epoch": 0.48385526529287687, "step": 1491, "weight_max": 4.187939703115262e-05, "weight_mean": 1.80947517947061e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9834284103126265e-05 }, { "epoch": 0.48385526529287687, "step": 1491, "weight_max": 4.252122016623616e-05, "weight_mean": 5.530082489713095e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4958688552724198e-05 }, { "epoch": 0.48385526529287687, "step": 1491, "weight_max": 3.998322426923551e-05, "weight_mean": 9.539852726447862e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7687456420389935e-05 }, { "epoch": 0.48385526529287687, "step": 1491, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.894285924208816e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.037234116869513e-05 }, { "epoch": 0.48417978257342204, "step": 1492, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.3320975489914417e-06, "weight_min": 5.958802340444436e-10, "weight_std": 9.421687536814716e-06 }, { "epoch": 0.48417978257342204, "step": 1492, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.805187801248394e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.816951044020243e-05 }, { "epoch": 0.48417978257342204, "step": 1492, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3602320652571507e-05, "weight_min": 1.4305072681963793e-09, "weight_std": 2.0127910829614848e-05 }, { "epoch": 0.48417978257342204, "step": 1492, "weight_max": 3.998322426923551e-05, "weight_mean": 9.747249350766651e-06, "weight_min": 0.0, "weight_std": 1.80057995748939e-05 }, { "epoch": 0.4845042998539672, "step": 1493, "weight_max": 2.926079287135508e-05, "weight_mean": 6.7085738919558935e-06, "weight_min": 0.0, "weight_std": 1.2488285392464604e-05 }, { "epoch": 0.4845042998539672, "step": 1493, "weight_max": 3.572963396436535e-05, "weight_mean": 8.678439371578861e-06, "weight_min": 0.0, "weight_std": 1.607691592653282e-05 }, { "epoch": 0.4845042998539672, "step": 1493, "weight_max": 5.3908038353256416e-06, "weight_mean": 6.74657940180623e-07, "weight_min": 0.0, "weight_std": 1.9056108158110874e-06 }, { "epoch": 0.4845042998539672, "step": 1493, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077128662378527e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357611689774785e-05 }, { "epoch": 0.4848288171345124, "step": 1494, "weight_max": 3.572963396436535e-05, "weight_mean": 8.86013003764674e-06, "weight_min": 0.0, "weight_std": 1.640346999920439e-05 }, { "epoch": 0.4848288171345124, "step": 1494, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.87150235939771e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.664785850152839e-05 }, { "epoch": 0.4848288171345124, "step": 1494, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235424850980053e-06, "weight_min": 0.0, "weight_std": 1.4806400940869935e-05 }, { "epoch": 0.4848288171345124, "step": 1494, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.544058588682674e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1148482119315304e-05 }, { "epoch": 0.4851533344150576, "step": 1495, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2971142496098764e-05, "weight_min": 8.603301404619401e-10, "weight_std": 1.8649272533366457e-05 }, { "epoch": 0.4851533344150576, "step": 1495, "weight_max": 5.564223556575598e-06, "weight_mean": 1.3843214219377842e-06, "weight_min": 0.0, "weight_std": 2.5620054202590836e-06 }, { "epoch": 0.4851533344150576, "step": 1495, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.592114105937071e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1571149773080833e-05 }, { "epoch": 0.4851533344150576, "step": 1495, "weight_max": 3.813097282545641e-05, "weight_mean": 8.67852577357553e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.606037767487578e-05 }, { "epoch": 0.4854778516956028, "step": 1496, "weight_max": 1.2064879095774472e-09, "weight_mean": 6.424932252890869e-10, "weight_min": 0.0, "weight_std": 3.2685137929533425e-10 }, { "epoch": 0.4854778516956028, "step": 1496, "weight_max": 3.843663216684945e-05, "weight_mean": 9.054285328602418e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.68020687851822e-05 }, { "epoch": 0.4854778516956028, "step": 1496, "weight_max": 1.4598501296703148e-09, "weight_mean": 7.51108508723064e-10, "weight_min": 5.679522963042416e-10, "weight_std": 3.0435215458979314e-10 }, { "epoch": 0.4854778516956028, "step": 1496, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3966732694825623e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9284716472611763e-05 }, { "epoch": 0.485802368976148, "step": 1497, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.79339780920418e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.0726222171797417e-05 }, { "epoch": 0.485802368976148, "step": 1497, "weight_max": 3.722132169059478e-05, "weight_mean": 7.235128578031436e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4108231880527455e-05 }, { "epoch": 0.485802368976148, "step": 1497, "weight_max": 4.187939703115262e-05, "weight_mean": 5.250110916676931e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4800507415202446e-05 }, { "epoch": 0.485802368976148, "step": 1497, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4839371942798607e-05, "weight_min": 0.0, "weight_std": 1.8980175809701905e-05 }, { "epoch": 0.48612688625669315, "step": 1498, "weight_max": 3.199401908204891e-05, "weight_mean": 4.000121862191008e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.131124281528173e-05 }, { "epoch": 0.48612688625669315, "step": 1498, "weight_max": 3.813097282545641e-05, "weight_mean": 7.909540727268904e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5050275578687433e-05 }, { "epoch": 0.48612688625669315, "step": 1498, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.782297300058417e-06, "weight_min": 0.0, "weight_std": 1.6415162463090383e-05 }, { "epoch": 0.48612688625669315, "step": 1498, "weight_max": 3.752332122530788e-05, "weight_mean": 1.004272598947864e-05, "weight_min": 0.0, "weight_std": 1.54793742694892e-05 }, { "epoch": 0.48645140353723837, "step": 1499, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0078351806441788e-05, "weight_min": 0.0, "weight_std": 1.8679085769690573e-05 }, { "epoch": 0.48645140353723837, "step": 1499, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3631337424158119e-05, "weight_min": 0.0, "weight_std": 1.930435792019125e-05 }, { "epoch": 0.48645140353723837, "step": 1499, "weight_max": 3.662098970380612e-05, "weight_mean": 8.843320756568573e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.450019499316113e-05 }, { "epoch": 0.48645140353723837, "step": 1499, "weight_max": 4.613847526258041e-09, "weight_mean": 1.3734282600097458e-09, "weight_min": 0.0, "weight_std": 1.466442856035144e-09 }, { "epoch": 0.48677592081778354, "grad_norm": 2.2140605449676514, "learning_rate": 7.151020408163266e-06, "loss": 0.202, "step": 1500 }, { "epoch": 0.48677592081778354, "eval_accuracy": 0.9875, "eval_auc": 0.9962899657550687, "eval_loss": 0.075770802795887, "eval_runtime": 155.6499, "eval_samples_per_second": 12.849, "eval_steps_per_second": 1.606, "step": 1500 }, { "epoch": 0.48677592081778354, "step": 1500, "weight_max": 4.381950566312298e-05, "weight_mean": 1.5290483133867383e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1195481167524122e-05 }, { "epoch": 0.48677592081778354, "step": 1500, "weight_max": 4.252122016623616e-05, "weight_mean": 9.491550372331403e-06, "weight_min": 0.0, "weight_std": 1.774149677657988e-05 }, { "epoch": 0.48677592081778354, "step": 1500, "weight_max": 3.632264997577295e-05, "weight_mean": 6.903681423864327e-06, "weight_min": 0.0, "weight_std": 1.360249279969139e-05 }, { "epoch": 0.48677592081778354, "step": 1500, "weight_max": 3.157625769745209e-07, "weight_mean": 4.067769765470075e-08, "weight_min": 0.0, "weight_std": 1.1116082276885209e-07 }, { "epoch": 0.48710043809832876, "step": 1501, "weight_max": 2.846461757144425e-05, "weight_mean": 3.5599512102635344e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0063004992844071e-05 }, { "epoch": 0.48710043809832876, "step": 1501, "weight_max": 3.602553260861896e-05, "weight_mean": 1.074225474440027e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.5707888451288454e-05 }, { "epoch": 0.48710043809832876, "step": 1501, "weight_max": 3.572963396436535e-05, "weight_mean": 9.676455192675348e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5840096239116974e-05 }, { "epoch": 0.48710043809832876, "step": 1501, "weight_max": 8.603301404619401e-10, "weight_mean": 7.201771401454948e-10, "weight_min": 5.864964625068581e-10, "weight_std": 1.1286877682481489e-10 }, { "epoch": 0.4874249553788739, "step": 1502, "weight_max": 3.843663216684945e-05, "weight_mean": 1.4000754163134843e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9342083760420792e-05 }, { "epoch": 0.4874249553788739, "step": 1502, "weight_max": 4.381950566312298e-05, "weight_mean": 1.682676884229295e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8548089428804815e-05 }, { "epoch": 0.4874249553788739, "step": 1502, "weight_max": 4.252122016623616e-05, "weight_mean": 2.05904489121167e-05, "weight_min": 0.0, "weight_std": 2.204764496127609e-05 }, { "epoch": 0.4874249553788739, "step": 1502, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.162134584679734e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4578663467545994e-05 }, { "epoch": 0.4877494726594191, "step": 1503, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077187324786792e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357588042912539e-05 }, { "epoch": 0.4877494726594191, "step": 1503, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.604751201346517e-06, "weight_min": 0.0, "weight_std": 1.5936380805214867e-05 }, { "epoch": 0.4877494726594191, "step": 1503, "weight_max": 3.782653220696375e-05, "weight_mean": 1.325266384810675e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.664563751546666e-05 }, { "epoch": 0.4877494726594191, "step": 1503, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.536046304157935e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.107906766468659e-05 }, { "epoch": 0.4880739899399643, "step": 1504, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8432419892633334e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.9756151232286356e-05 }, { "epoch": 0.4880739899399643, "step": 1504, "weight_max": 1.763176260283217e-05, "weight_mean": 2.204862994403811e-06, "weight_min": 6.100955851628953e-10, "weight_std": 6.233409294509329e-06 }, { "epoch": 0.4880739899399643, "step": 1504, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.535747287562117e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.702926238067448e-05 }, { "epoch": 0.4880739899399643, "step": 1504, "weight_max": 4.187939703115262e-05, "weight_mean": 8.927158887672704e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.685344432189595e-05 }, { "epoch": 0.4883985072205095, "step": 1505, "weight_max": 3.5141503758495674e-05, "weight_mean": 1.2744774721795693e-05, "weight_min": 0.0, "weight_std": 1.759752194629982e-05 }, { "epoch": 0.4883985072205095, "step": 1505, "weight_max": 4.187939703115262e-05, "weight_mean": 2.253546699648723e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9325345419929363e-05 }, { "epoch": 0.4883985072205095, "step": 1505, "weight_max": 9.013804813662318e-09, "weight_mean": 1.727782916738363e-09, "weight_min": 5.864964625068581e-10, "weight_std": 2.9448319338598594e-09 }, { "epoch": 0.4883985072205095, "step": 1505, "weight_max": 1.2909623592349817e-06, "weight_mean": 1.6235840405443014e-07, "weight_min": 5.771870759119224e-10, "weight_std": 4.560255604246777e-07 }, { "epoch": 0.4887230245010547, "step": 1506, "weight_max": 4.252122016623616e-05, "weight_mean": 9.45605461311061e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.768676156643778e-05 }, { "epoch": 0.4887230245010547, "step": 1506, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.724332474230323e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5776593500049785e-05 }, { "epoch": 0.4887230245010547, "step": 1506, "weight_max": 3.843663216684945e-05, "weight_mean": 4.8056590458145365e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3588965884991921e-05 }, { "epoch": 0.4887230245010547, "step": 1506, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235728167463094e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4806278159085196e-05 }, { "epoch": 0.48904754178159987, "step": 1507, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998677468392998e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135891433397774e-05 }, { "epoch": 0.48904754178159987, "step": 1507, "weight_max": 3.782653220696375e-05, "weight_mean": 7.68877816881286e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4728581845702138e-05 }, { "epoch": 0.48904754178159987, "step": 1507, "weight_max": 1.0509849657225345e-09, "weight_mean": 7.172260563237387e-10, "weight_min": 5.864964625068581e-10, "weight_std": 1.6367897315294755e-10 }, { "epoch": 0.48904754178159987, "step": 1507, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.642939064884558e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6308804333675653e-05 }, { "epoch": 0.48937205906214504, "step": 1508, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470557754160836e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9385955965844914e-05 }, { "epoch": 0.48937205906214504, "step": 1508, "weight_max": 3.116124207735993e-05, "weight_mean": 3.89603701478336e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.1016805729013868e-05 }, { "epoch": 0.48937205906214504, "step": 1508, "weight_max": 3.572963396436535e-05, "weight_mean": 8.259342394012492e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.5359291865024716e-05 }, { "epoch": 0.48937205906214504, "step": 1508, "weight_max": 4.252122016623616e-05, "weight_mean": 8.907014489523135e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6897667592274956e-05 }, { "epoch": 0.48969657634269026, "step": 1509, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0472860594745725e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9402606994844973e-05 }, { "epoch": 0.48969657634269026, "step": 1509, "weight_max": 3.632264997577295e-05, "weight_mean": 4.615465968527133e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2813327884941828e-05 }, { "epoch": 0.48969657634269026, "step": 1509, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5057883501867764e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.046629015239887e-05 }, { "epoch": 0.48969657634269026, "step": 1509, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077142759546405e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357606232806575e-05 }, { "epoch": 0.4900210936232354, "grad_norm": 0.43940141797065735, "learning_rate": 7.13061224489796e-06, "loss": 0.1692, "step": 1510 }, { "epoch": 0.4900210936232354, "step": 1510, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3660762306244578e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.931496080942452e-05 }, { "epoch": 0.4900210936232354, "step": 1510, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155729796679225e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.881509888335131e-05 }, { "epoch": 0.4900210936232354, "step": 1510, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5499190340051427e-05, "weight_min": 0.0, "weight_std": 1.8283741155755706e-05 }, { "epoch": 0.4900210936232354, "step": 1510, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5548615920124575e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1469315470312722e-05 }, { "epoch": 0.49034561090378065, "step": 1511, "weight_max": 3.998322426923551e-05, "weight_mean": 1.6407298971898854e-05, "weight_min": 0.0, "weight_std": 1.8175418517785147e-05 }, { "epoch": 0.49034561090378065, "step": 1511, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.408312346029561e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.9487020836095326e-05 }, { "epoch": 0.49034561090378065, "step": 1511, "weight_max": 4.187939703115262e-05, "weight_mean": 2.6814990633283742e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7480642782174982e-05 }, { "epoch": 0.49034561090378065, "step": 1511, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.957926764996955e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0669720722944476e-05 }, { "epoch": 0.4906701281843258, "step": 1512, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3262711036077235e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.85162243724335e-05 }, { "epoch": 0.4906701281843258, "step": 1512, "weight_max": 3.998322426923551e-05, "weight_mean": 1.472376061428804e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0326413505245e-05 }, { "epoch": 0.4906701281843258, "step": 1512, "weight_max": 2.3087483036476897e-09, "weight_mean": 7.747937291746609e-10, "weight_min": 0.0, "weight_std": 6.618239845046503e-10 }, { "epoch": 0.4906701281843258, "step": 1512, "weight_max": 4.187939703115262e-05, "weight_mean": 1.93452688108664e-05, "weight_min": 0.0, "weight_std": 2.0731169570353813e-05 }, { "epoch": 0.490994645464871, "step": 1513, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0125025912420824e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1577643565251492e-05 }, { "epoch": 0.490994645464871, "step": 1513, "weight_max": 3.602553260861896e-05, "weight_mean": 8.61884473124519e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.5872365111135878e-05 }, { "epoch": 0.490994645464871, "step": 1513, "weight_max": 3.397988257347606e-05, "weight_mean": 1.2250241525180172e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.693338344921358e-05 }, { "epoch": 0.490994645464871, "step": 1513, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4150728020467795e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.954332401510328e-05 }, { "epoch": 0.4913191627454162, "step": 1514, "weight_max": 4.447596802492626e-05, "weight_mean": 5.5623841035412624e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5723464457551017e-05 }, { "epoch": 0.4913191627454162, "step": 1514, "weight_max": 3.998322426923551e-05, "weight_mean": 5.582533503911691e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.399571010551881e-05 }, { "epoch": 0.4913191627454162, "step": 1514, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2356858759594616e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.480629543948453e-05 }, { "epoch": 0.4913191627454162, "step": 1514, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.019700696109794e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1401841877377592e-05 }, { "epoch": 0.49164368002596137, "step": 1515, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0117086276295595e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8746688510873355e-05 }, { "epoch": 0.49164368002596137, "step": 1515, "weight_max": 3.662098970380612e-05, "weight_mean": 4.578304015012691e-06, "weight_min": 0.0, "weight_std": 1.2947200048074592e-05 }, { "epoch": 0.49164368002596137, "step": 1515, "weight_max": 3.722132169059478e-05, "weight_mean": 8.548355253878981e-06, "weight_min": 0.0, "weight_std": 1.5909810826997273e-05 }, { "epoch": 0.49164368002596137, "step": 1515, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.090919775189832e-06, "weight_min": 0.0, "weight_std": 1.6879968825378455e-05 }, { "epoch": 0.4919681973065066, "step": 1516, "weight_max": 3.967146767536178e-05, "weight_mean": 9.687811143521685e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.7943732018466108e-05 }, { "epoch": 0.4919681973065066, "step": 1516, "weight_max": 3.813097282545641e-05, "weight_mean": 1.724997127894312e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8750259187072515e-05 }, { "epoch": 0.4919681973065066, "step": 1516, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235724984231638e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4806279978074599e-05 }, { "epoch": 0.4919681973065066, "step": 1516, "weight_max": 4.187939703115262e-05, "weight_mean": 9.738911103340797e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8098744476446882e-05 }, { "epoch": 0.49229271458705176, "step": 1517, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3448683603201061e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.8613778593135066e-05 }, { "epoch": 0.49229271458705176, "step": 1517, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5067596905282699e-05, "weight_min": 0.0, "weight_std": 2.0941193724866025e-05 }, { "epoch": 0.49229271458705176, "step": 1517, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4846371414023452e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.052123636531178e-05 }, { "epoch": 0.49229271458705176, "step": 1517, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2682850865530781e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8745733541436493e-05 }, { "epoch": 0.4926172318675969, "step": 1518, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2491876987041906e-05, "weight_min": 0.0, "weight_std": 1.8080863810610026e-05 }, { "epoch": 0.4926172318675969, "step": 1518, "weight_max": 3.572963396436535e-05, "weight_mean": 1.2722589417535346e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.762479223543778e-05 }, { "epoch": 0.4926172318675969, "step": 1518, "weight_max": 4.187939703115262e-05, "weight_mean": 7.902168363216333e-06, "weight_min": 0.0, "weight_std": 1.5627301763743162e-05 }, { "epoch": 0.4926172318675969, "step": 1518, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316939223121153e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5032799637992866e-05 }, { "epoch": 0.49294174914814215, "step": 1519, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3936261893832125e-05, "weight_min": 0.0, "weight_std": 1.9300388885312714e-05 }, { "epoch": 0.49294174914814215, "step": 1519, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.394324150780449e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2423736734490376e-05 }, { "epoch": 0.49294174914814215, "step": 1519, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.543142368784174e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7716978618409485e-05 }, { "epoch": 0.49294174914814215, "step": 1519, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8144335626857355e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.948234421433881e-05 }, { "epoch": 0.4932662664286873, "grad_norm": 0.44432333111763, "learning_rate": 7.110204081632653e-06, "loss": 0.096, "step": 1520 }, { "epoch": 0.4932662664286873, "step": 1520, "weight_max": 4.187939703115262e-05, "weight_mean": 9.963778211385943e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.847977728175465e-05 }, { "epoch": 0.4932662664286873, "step": 1520, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235688604443567e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.4806294529989827e-05 }, { "epoch": 0.4932662664286873, "step": 1520, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3841559848515317e-05, "weight_min": 0.0, "weight_std": 1.9398466974962503e-05 }, { "epoch": 0.4932662664286873, "step": 1520, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0001813279814087e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8545561033533886e-05 }, { "epoch": 0.49359078370923254, "step": 1521, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843508577323519e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.369766778225312e-05 }, { "epoch": 0.49359078370923254, "step": 1521, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4767849279451184e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0408253476489335e-05 }, { "epoch": 0.49359078370923254, "step": 1521, "weight_max": 1.6735920427279893e-09, "weight_mean": 7.358342268837248e-10, "weight_min": 0.0, "weight_std": 4.6025180888698003e-10 }, { "epoch": 0.49359078370923254, "step": 1521, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921505933452863e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.39156272780383e-05 }, { "epoch": 0.4939153009897777, "step": 1522, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.2123564374633133e-05, "weight_min": 0.0, "weight_std": 1.6445013898191974e-05 }, { "epoch": 0.4939153009897777, "step": 1522, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5097555660759099e-05, "weight_min": 0.0, "weight_std": 2.0943702111253515e-05 }, { "epoch": 0.4939153009897777, "step": 1522, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2660931133723352e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8047814592137e-05 }, { "epoch": 0.4939153009897777, "step": 1522, "weight_max": 4.187939703115262e-05, "weight_mean": 1.6195906937355176e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.902106851048302e-05 }, { "epoch": 0.4942398182703229, "step": 1523, "weight_max": 4.381950566312298e-05, "weight_mean": 1.5185693882813212e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1106146959937178e-05 }, { "epoch": 0.4942398182703229, "step": 1523, "weight_max": 1.8684177121031098e-05, "weight_mean": 2.3360112209047657e-06, "weight_min": 0.0, "weight_std": 6.6056572904926725e-06 }, { "epoch": 0.4942398182703229, "step": 1523, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3389512787398417e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.875857742561493e-05 }, { "epoch": 0.4942398182703229, "step": 1523, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235989192442503e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.480617265769979e-05 }, { "epoch": 0.4945643355508681, "step": 1524, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4739791367901489e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0421806766535155e-05 }, { "epoch": 0.4945643355508681, "step": 1524, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0235717127216049e-05, "weight_min": 0.0, "weight_std": 1.897056972666178e-05 }, { "epoch": 0.4945643355508681, "step": 1524, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4773476323171053e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0440893422346562e-05 }, { "epoch": 0.4945643355508681, "step": 1524, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998537860956276e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135947822069284e-05 }, { "epoch": 0.49488885283141326, "step": 1525, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.3972644309396856e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5261650332831778e-05 }, { "epoch": 0.49488885283141326, "step": 1525, "weight_max": 3.718365615412722e-09, "weight_mean": 1.0752740919883763e-09, "weight_min": 0.0, "weight_std": 1.2007274063918771e-09 }, { "epoch": 0.49488885283141326, "step": 1525, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0083645975100808e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.870288724603597e-05 }, { "epoch": 0.49488885283141326, "step": 1525, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.92109893457382e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.391579098708462e-05 }, { "epoch": 0.4952133701119585, "step": 1526, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827808086498408e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0822513104358222e-05 }, { "epoch": 0.4952133701119585, "step": 1526, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.535735443932936e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.76674311660463e-05 }, { "epoch": 0.4952133701119585, "step": 1526, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153956282010768e-05, "weight_min": 0.0, "weight_std": 1.8802644262905233e-05 }, { "epoch": 0.4952133701119585, "step": 1526, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2357222557475325e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.48062808875693e-05 }, { "epoch": 0.49553788739250365, "step": 1527, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.982138751889579e-06, "weight_min": 0.0, "weight_std": 1.687184521870222e-05 }, { "epoch": 0.49553788739250365, "step": 1527, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5161114788497798e-05, "weight_min": 0.0, "weight_std": 2.0969771867385134e-05 }, { "epoch": 0.49553788739250365, "step": 1527, "weight_max": 3.752332122530788e-05, "weight_mean": 5.343666998669505e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3130179468134884e-05 }, { "epoch": 0.49553788739250365, "step": 1527, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.33229922944156e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.421606591786258e-06 }, { "epoch": 0.4958624046730488, "step": 1528, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844435807171976e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3697292160941288e-05 }, { "epoch": 0.4958624046730488, "step": 1528, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430428361956729e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2527723811217584e-05 }, { "epoch": 0.4958624046730488, "step": 1528, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.367377970193047e-06, "weight_min": 0.0, "weight_std": 1.74607484950684e-05 }, { "epoch": 0.4958624046730488, "step": 1528, "weight_max": 4.252122016623616e-05, "weight_mean": 2.570794094935991e-05, "weight_min": 7.937498991417158e-10, "weight_std": 1.94203184946673e-05 }, { "epoch": 0.49618692195359404, "step": 1529, "weight_max": 4.381950566312298e-05, "weight_mean": 1.9908153262804262e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.142030825780239e-05 }, { "epoch": 0.49618692195359404, "step": 1529, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8491424270905554e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0048042642883956e-05 }, { "epoch": 0.49618692195359404, "step": 1529, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5013103620731272e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0801315258722752e-05 }, { "epoch": 0.49618692195359404, "step": 1529, "weight_max": 2.714597657771378e-09, "weight_mean": 1.030071583585368e-09, "weight_min": 6.196655966128617e-10, "weight_std": 7.035646509834237e-10 }, { "epoch": 0.4965114392341392, "grad_norm": 15.963890075683594, "learning_rate": 7.089795918367347e-06, "loss": 0.1228, "step": 1530 }, { "epoch": 0.4965114392341392, "step": 1530, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155850431037834e-06, "weight_min": 0.0, "weight_std": 1.4581191862816922e-05 }, { "epoch": 0.4965114392341392, "step": 1530, "weight_max": 2.2888118564878823e-06, "weight_mean": 2.867956823138229e-07, "weight_min": 5.679522963042416e-10, "weight_std": 8.089367042884987e-07 }, { "epoch": 0.4965114392341392, "step": 1530, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.3937831429066136e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.995001730392687e-05 }, { "epoch": 0.4965114392341392, "step": 1530, "weight_max": 4.613847526258041e-09, "weight_mean": 1.7708978727881686e-09, "weight_min": 6.29309993005478e-10, "weight_std": 1.571153651624968e-09 }, { "epoch": 0.49683595651468443, "step": 1531, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0090290490770712e-05, "weight_min": 0.0, "weight_std": 2.1562269466812722e-05 }, { "epoch": 0.49683595651468443, "step": 1531, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4426223060581833e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.996138053073082e-05 }, { "epoch": 0.49683595651468443, "step": 1531, "weight_max": 3.632264997577295e-05, "weight_mean": 8.098967555270065e-06, "weight_min": 0.0, "weight_std": 1.5141347830649465e-05 }, { "epoch": 0.49683595651468443, "step": 1531, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5272158634616062e-05, "weight_min": 0.0, "weight_std": 2.109144043060951e-05 }, { "epoch": 0.4971604737952296, "step": 1532, "weight_max": 4.381950566312298e-05, "weight_mean": 5.4853894653206225e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5489335055463016e-05 }, { "epoch": 0.4971604737952296, "step": 1532, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9643806808744557e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9287237591925077e-05 }, { "epoch": 0.4971604737952296, "step": 1532, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3852182746632025e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9133602108922787e-05 }, { "epoch": 0.4971604737952296, "step": 1532, "weight_max": 2.716206836339552e-05, "weight_mean": 3.39628195433761e-06, "weight_min": 6.196655966128617e-10, "weight_std": 9.602827958588023e-06 }, { "epoch": 0.49748499107577476, "step": 1533, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.1747374628612306e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.6044157746364363e-05 }, { "epoch": 0.49748499107577476, "step": 1533, "weight_max": 3.5434961318969727e-05, "weight_mean": 8.677446203364525e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.607075500942301e-05 }, { "epoch": 0.49748499107577476, "step": 1533, "weight_max": 3.632264997577295e-05, "weight_mean": 1.5871126379352063e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.716543920338154e-05 }, { "epoch": 0.49748499107577476, "step": 1533, "weight_max": 2.697995114431251e-06, "weight_mean": 3.37908460323888e-07, "weight_min": 5.864964625068581e-10, "weight_std": 9.536190077596984e-07 }, { "epoch": 0.49780950835632, "step": 1534, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959618308930658e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4025705240783282e-05 }, { "epoch": 0.49780950835632, "step": 1534, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930522325390484e-06, "weight_min": 0.0, "weight_std": 1.1114544577139895e-05 }, { "epoch": 0.49780950835632, "step": 1534, "weight_max": 3.7750612591480603e-07, "weight_mean": 4.768506656205318e-08, "weight_min": 0.0, "weight_std": 1.3326807390967588e-07 }, { "epoch": 0.49780950835632, "step": 1534, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8592931155581027e-05, "weight_min": 0.0, "weight_std": 2.0378971385071054e-05 }, { "epoch": 0.49813402563686515, "step": 1535, "weight_max": 3.722132169059478e-05, "weight_mean": 1.3660346667165868e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8857812392525375e-05 }, { "epoch": 0.49813402563686515, "step": 1535, "weight_max": 3.998322426923551e-05, "weight_mean": 1.398794847773388e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9429313397267833e-05 }, { "epoch": 0.49813402563686515, "step": 1535, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5707792044850066e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1688312699552625e-05 }, { "epoch": 0.49813402563686515, "step": 1535, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3133294487488456e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.8406288290861994e-05 }, { "epoch": 0.4984585429174104, "step": 1536, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9647945009637624e-05, "weight_min": 0.0, "weight_std": 2.1028859919169918e-05 }, { "epoch": 0.4984585429174104, "step": 1536, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4999751329014543e-05, "weight_min": 0.0, "weight_std": 2.0716370272566564e-05 }, { "epoch": 0.4984585429174104, "step": 1536, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.481614526710473e-05, "weight_min": 0.0, "weight_std": 2.052115996775683e-05 }, { "epoch": 0.4984585429174104, "step": 1536, "weight_max": 3.722132169059478e-05, "weight_mean": 4.715648174169473e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3135437257005833e-05 }, { "epoch": 0.49878306019795554, "step": 1537, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882005214312812e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3806603419652674e-05 }, { "epoch": 0.49878306019795554, "step": 1537, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5432202568918e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.2840829185734037e-05 }, { "epoch": 0.49878306019795554, "step": 1537, "weight_max": 4.252122016623616e-05, "weight_mean": 1.586600956215989e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1896346879657358e-05 }, { "epoch": 0.49878306019795554, "step": 1537, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9371935195522383e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9329521819599904e-05 }, { "epoch": 0.4991075774785007, "step": 1538, "weight_max": 3.813097282545641e-05, "weight_mean": 9.394641892868094e-06, "weight_min": 0.0, "weight_std": 1.637493187445216e-05 }, { "epoch": 0.4991075774785007, "step": 1538, "weight_max": 3.813097282545641e-05, "weight_mean": 4.766984147863695e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3481087080435827e-05 }, { "epoch": 0.4991075774785007, "step": 1538, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0768830988090485e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4357710824697278e-05 }, { "epoch": 0.4991075774785007, "step": 1538, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.95849768514745e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8442418877384625e-05 }, { "epoch": 0.49943209475904593, "step": 1539, "weight_max": 4.187939703115262e-05, "weight_mean": 1.897382753668353e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0664887415478006e-05 }, { "epoch": 0.49943209475904593, "step": 1539, "weight_max": 3.662098970380612e-05, "weight_mean": 6.4820928855624516e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3292976291268133e-05 }, { "epoch": 0.49943209475904593, "step": 1539, "weight_max": 4.447596802492626e-05, "weight_mean": 1.013836299534887e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.888656879600603e-05 }, { "epoch": 0.49943209475904593, "step": 1539, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3944643797003664e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.9509319827193394e-05 }, { "epoch": 0.4997566120395911, "grad_norm": 0.989924430847168, "learning_rate": 7.069387755102041e-06, "loss": 0.0477, "step": 1540 }, { "epoch": 0.4997566120395911, "step": 1540, "weight_max": 3.397988257347606e-05, "weight_mean": 4.249274752510246e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2012979823339265e-05 }, { "epoch": 0.4997566120395911, "step": 1540, "weight_max": 3.967146767536178e-05, "weight_mean": 9.72577527136309e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.801218240871094e-05 }, { "epoch": 0.4997566120395911, "step": 1540, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076866727904417e-06, "weight_min": 0.0, "weight_std": 1.4357718100654893e-05 }, { "epoch": 0.4997566120395911, "step": 1540, "weight_max": 4.252122016623616e-05, "weight_mean": 8.646935384604149e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6561998563702218e-05 }, { "epoch": 0.5000811293201363, "step": 1541, "weight_max": 1.6811793102533557e-05, "weight_mean": 2.1020944132033037e-06, "weight_min": 5.958802340444436e-10, "weight_std": 5.943615633441368e-06 }, { "epoch": 0.5000811293201363, "step": 1541, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653190444514621e-06, "weight_min": 0.0, "weight_std": 1.3159512491256464e-05 }, { "epoch": 0.5000811293201363, "step": 1541, "weight_max": 3.967146767536178e-05, "weight_mean": 1.2821698874176946e-05, "weight_min": 0.0, "weight_std": 1.8019998606177978e-05 }, { "epoch": 0.5000811293201363, "step": 1541, "weight_max": 3.843663216684945e-05, "weight_mean": 9.175762897939421e-06, "weight_min": 0.0, "weight_std": 1.6125590263982303e-05 }, { "epoch": 0.5004056466006815, "step": 1542, "weight_max": 3.572963396436535e-05, "weight_mean": 8.138581506500486e-06, "weight_min": 0.0, "weight_std": 1.3589511581812985e-05 }, { "epoch": 0.5004056466006815, "step": 1542, "weight_max": 1.6578410866330273e-09, "weight_mean": 8.483548308291233e-10, "weight_min": 0.0, "weight_std": 4.866165803640854e-10 }, { "epoch": 0.5004056466006815, "step": 1542, "weight_max": 3.782653220696375e-05, "weight_mean": 1.0377158105256967e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.5160389921220485e-05 }, { "epoch": 0.5004056466006815, "step": 1542, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.322354925534455e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2217149560456164e-05 }, { "epoch": 0.5007301638812267, "step": 1543, "weight_max": 4.252122016623616e-05, "weight_mean": 8.775750757195055e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.672506368777249e-05 }, { "epoch": 0.5007301638812267, "step": 1543, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.733670594869182e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.806370710255578e-05 }, { "epoch": 0.5007301638812267, "step": 1543, "weight_max": 3.843663216684945e-05, "weight_mean": 9.4204478955362e-06, "weight_min": 0.0, "weight_std": 1.7446007404942065e-05 }, { "epoch": 0.5007301638812267, "step": 1543, "weight_max": 4.187939703115262e-05, "weight_mean": 1.518840144854039e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9258441170677543e-05 }, { "epoch": 0.5010546811617719, "step": 1544, "weight_max": 3.998322426923551e-05, "weight_mean": 9.841559403866995e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.822426747821737e-05 }, { "epoch": 0.5010546811617719, "step": 1544, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0257401299895719e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.875447742349934e-05 }, { "epoch": 0.5010546811617719, "step": 1544, "weight_max": 2.7030688798390656e-08, "weight_mean": 4.269709918958142e-09, "weight_min": 0.0, "weight_std": 9.227182573567916e-09 }, { "epoch": 0.5010546811617719, "step": 1544, "weight_max": 1.2882609423670033e-09, "weight_mean": 7.412837010889461e-10, "weight_min": 0.0, "weight_std": 3.86550375042205e-10 }, { "epoch": 0.501379198442317, "step": 1545, "weight_max": 3.36925259034615e-05, "weight_mean": 6.3901952671585605e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.2604637049662415e-05 }, { "epoch": 0.501379198442317, "step": 1545, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4917812222847715e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0593319277395494e-05 }, { "epoch": 0.501379198442317, "step": 1545, "weight_max": 7.937498991417158e-10, "weight_mean": 6.656826756490375e-10, "weight_min": 5.771870759119224e-10, "weight_std": 6.641888428138287e-11 }, { "epoch": 0.501379198442317, "step": 1545, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077137302578194e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4357608051795978e-05 }, { "epoch": 0.5017037157228622, "step": 1546, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3600514648715034e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8836102753994055e-05 }, { "epoch": 0.5017037157228622, "step": 1546, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077936293673702e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4357285181176849e-05 }, { "epoch": 0.5017037157228622, "step": 1546, "weight_max": 3.5141503758495674e-05, "weight_mean": 9.845803106145468e-06, "weight_min": 0.0, "weight_std": 1.4196239135344513e-05 }, { "epoch": 0.5017037157228622, "step": 1546, "weight_max": 4.252122016623616e-05, "weight_mean": 8.873668775777332e-06, "weight_min": 0.0, "weight_std": 1.6853795386850834e-05 }, { "epoch": 0.5020282330034075, "step": 1547, "weight_max": 3.967146767536178e-05, "weight_mean": 1.759703991410788e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9073468138230965e-05 }, { "epoch": 0.5020282330034075, "step": 1547, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4973511497373693e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0743185814353637e-05 }, { "epoch": 0.5020282330034075, "step": 1547, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.958986993297003e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8442116925143637e-05 }, { "epoch": 0.5020282330034075, "step": 1547, "weight_max": 2.7956259529560157e-09, "weight_mean": 9.67464108825311e-10, "weight_min": 5.958802340444436e-10, "weight_std": 7.443543559304544e-10 }, { "epoch": 0.5023527502839527, "step": 1548, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.047088673862163e-05, "weight_min": 0.0, "weight_std": 2.1916126570431516e-05 }, { "epoch": 0.5023527502839527, "step": 1548, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0078495506604668e-05, "weight_min": 0.0, "weight_std": 1.8678998458199203e-05 }, { "epoch": 0.5023527502839527, "step": 1548, "weight_max": 2.3831859152778634e-08, "weight_mean": 3.658619629476334e-09, "weight_min": 0.0, "weight_std": 8.173578258663383e-09 }, { "epoch": 0.5023527502839527, "step": 1548, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.729946214065421e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.803669692890253e-05 }, { "epoch": 0.5026772675644978, "step": 1549, "weight_max": 3.340639523230493e-05, "weight_mean": 8.281075679406058e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5332960174418986e-05 }, { "epoch": 0.5026772675644978, "step": 1549, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.201718967233319e-06, "weight_min": 0.0, "weight_std": 1.707934461592231e-05 }, { "epoch": 0.5026772675644978, "step": 1549, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.7932186387479305e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.0726294931373559e-05 }, { "epoch": 0.5026772675644978, "step": 1549, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7113878129748628e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.896198227768764e-05 }, { "epoch": 0.503001784845043, "grad_norm": 8.739167213439941, "learning_rate": 7.0489795918367355e-06, "loss": 0.1917, "step": 1550 }, { "epoch": 0.503001784845043, "step": 1550, "weight_max": 3.1715208024252206e-05, "weight_mean": 7.723756425548345e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4306264347396791e-05 }, { "epoch": 0.503001784845043, "step": 1550, "weight_max": 2.872879122151062e-05, "weight_mean": 3.591687118387199e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0156923963222653e-05 }, { "epoch": 0.503001784845043, "step": 1550, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4936684237909503e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.060326369246468e-05 }, { "epoch": 0.503001784845043, "step": 1550, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.054441761691123e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.717057421046775e-05 }, { "epoch": 0.5033263021255882, "step": 1551, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998891199647915e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4135805031401105e-05 }, { "epoch": 0.5033263021255882, "step": 1551, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9899704057024792e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1373951312853023e-05 }, { "epoch": 0.5033263021255882, "step": 1551, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3089254935039207e-05, "weight_min": 9.413956281179026e-10, "weight_std": 1.8426346287014894e-05 }, { "epoch": 0.5033263021255882, "step": 1551, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8683327652979642e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0269617380108684e-05 }, { "epoch": 0.5036508194061334, "step": 1552, "weight_max": 3.662098970380612e-05, "weight_mean": 7.941308467707131e-06, "weight_min": 0.0, "weight_std": 1.4930660654499661e-05 }, { "epoch": 0.5036508194061334, "step": 1552, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.326562576461583e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.754958248056937e-05 }, { "epoch": 0.5036508194061334, "step": 1552, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1943887329834979e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.717295526759699e-05 }, { "epoch": 0.5036508194061334, "step": 1552, "weight_max": 3.006794759130571e-05, "weight_mean": 6.779138402634999e-06, "weight_min": 0.0, "weight_std": 1.2649969903577585e-05 }, { "epoch": 0.5039753366866786, "step": 1553, "weight_max": 3.782653220696375e-05, "weight_mean": 4.72942519991193e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.3373251022130717e-05 }, { "epoch": 0.5039753366866786, "step": 1553, "weight_max": 2.639517333591357e-05, "weight_mean": 3.3001226711348863e-06, "weight_min": 0.0, "weight_std": 9.331810360890813e-06 }, { "epoch": 0.5039753366866786, "step": 1553, "weight_max": 2.407077559496429e-08, "weight_mean": 5.863191709920557e-09, "weight_min": 6.100955851628953e-10, "weight_std": 9.46400025014782e-09 }, { "epoch": 0.5039753366866786, "step": 1553, "weight_max": 3.813097282545641e-05, "weight_mean": 9.495355698163621e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.758052167133428e-05 }, { "epoch": 0.5042998539672238, "step": 1554, "weight_max": 4.252122016623616e-05, "weight_mean": 9.280082849727478e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7423091776436195e-05 }, { "epoch": 0.5042998539672238, "step": 1554, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4058799933991395e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.966160380106885e-05 }, { "epoch": 0.5042998539672238, "step": 1554, "weight_max": 3.4849264920921996e-05, "weight_mean": 4.356657427706523e-06, "weight_min": 0.0, "weight_std": 1.232087379321456e-05 }, { "epoch": 0.5042998539672238, "step": 1554, "weight_max": 3.967146767536178e-05, "weight_mean": 1.408804837410571e-05, "weight_min": 0.0, "weight_std": 1.9514465748216026e-05 }, { "epoch": 0.5046243712477689, "step": 1555, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3601236787508242e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8999624444404617e-05 }, { "epoch": 0.5046243712477689, "step": 1555, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5432073269039392e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.131992732756771e-05 }, { "epoch": 0.5046243712477689, "step": 1555, "weight_max": 3.662098970380612e-05, "weight_mean": 2.079630758089479e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.746879206621088e-05 }, { "epoch": 0.5046243712477689, "step": 1555, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0623403795761988e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8465509128873236e-05 }, { "epoch": 0.5049488885283141, "step": 1556, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.440269084530883e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6049880287027918e-05 }, { "epoch": 0.5049488885283141, "step": 1556, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805495791515568e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3589030459115747e-05 }, { "epoch": 0.5049488885283141, "step": 1556, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.430354238138534e-06, "weight_min": 0.0, "weight_std": 1.252775291504804e-05 }, { "epoch": 0.5049488885283141, "step": 1556, "weight_max": 3.843663216684945e-05, "weight_mean": 4.805246135219932e-06, "weight_min": 0.0, "weight_std": 1.3589132322522346e-05 }, { "epoch": 0.5052734058088594, "step": 1557, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998592885385733e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4135925994196441e-05 }, { "epoch": 0.5052734058088594, "step": 1557, "weight_max": 1.0068959888087647e-08, "weight_mean": 1.8724624073485074e-09, "weight_min": 0.0, "weight_std": 3.3442131286420818e-09 }, { "epoch": 0.5052734058088594, "step": 1557, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0055471395608038e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.1535914129344746e-05 }, { "epoch": 0.5052734058088594, "step": 1557, "weight_max": 4.187939703115262e-05, "weight_mean": 1.2997781595913693e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.827120649977587e-05 }, { "epoch": 0.5055979230894045, "step": 1558, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882023404206848e-06, "weight_min": 0.0, "weight_std": 1.380659614369506e-05 }, { "epoch": 0.5055979230894045, "step": 1558, "weight_max": 1.7014956029015593e-05, "weight_mean": 2.127703055521124e-06, "weight_min": 5.864964625068581e-10, "weight_std": 6.01535839450662e-06 }, { "epoch": 0.5055979230894045, "step": 1558, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5125030586204957e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.0887307982775383e-05 }, { "epoch": 0.5055979230894045, "step": 1558, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4251574611989781e-05, "weight_min": 0.0, "weight_std": 1.9942473954870366e-05 }, { "epoch": 0.5059224403699497, "step": 1559, "weight_max": 2.5890016331686638e-05, "weight_mean": 3.2369043765356764e-06, "weight_min": 0.0, "weight_std": 9.153240171144716e-06 }, { "epoch": 0.5059224403699497, "step": 1559, "weight_max": 3.4849264920921996e-05, "weight_mean": 7.881906640250236e-06, "weight_min": 0.0, "weight_std": 1.4700873180117924e-05 }, { "epoch": 0.5059224403699497, "step": 1559, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3917468095314689e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9650613467092626e-05 }, { "epoch": 0.5059224403699497, "step": 1559, "weight_max": 3.722132169059478e-05, "weight_mean": 9.00951727089705e-06, "weight_min": 0.0, "weight_std": 1.6692712961230427e-05 }, { "epoch": 0.5062469576504949, "grad_norm": 0.5925986170768738, "learning_rate": 7.028571428571429e-06, "loss": 0.0431, "step": 1560 }, { "epoch": 0.5062469576504949, "step": 1560, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4439858205150813e-05, "weight_min": 0.0, "weight_std": 2.004147791012656e-05 }, { "epoch": 0.5062469576504949, "step": 1560, "weight_max": 4.187939703115262e-05, "weight_mean": 1.943730421771761e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.0270315872039646e-05 }, { "epoch": 0.5062469576504949, "step": 1560, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.9654555621382315e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1212593562959228e-05 }, { "epoch": 0.5062469576504949, "step": 1560, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.991865792660974e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1311885575414635e-05 }, { "epoch": 0.50657147493104, "step": 1561, "weight_max": 3.967146767536178e-05, "weight_mean": 1.87724126590183e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.010694151977077e-05 }, { "epoch": 0.50657147493104, "step": 1561, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3738166671828367e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.957138374564238e-05 }, { "epoch": 0.50657147493104, "step": 1561, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4268951417761855e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.6812253306852654e-05 }, { "epoch": 0.50657147493104, "step": 1561, "weight_max": 4.1242459701607004e-05, "weight_mean": 6.328506970021408e-06, "weight_min": 0.0, "weight_std": 1.4484181519947015e-05 }, { "epoch": 0.5068959922115853, "step": 1562, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.263804006157443e-05, "weight_min": 0.0, "weight_std": 1.761590647220146e-05 }, { "epoch": 0.5068959922115853, "step": 1562, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156191491550999e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.4581055438611656e-05 }, { "epoch": 0.5068959922115853, "step": 1562, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1455211279098876e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.7571257558302023e-05 }, { "epoch": 0.5068959922115853, "step": 1562, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.74357283464633e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4593096238968428e-05 }, { "epoch": 0.5072205094921305, "step": 1563, "weight_max": 5.107872766529908e-06, "weight_mean": 6.392409090949513e-07, "weight_min": 6.196655966128617e-10, "weight_std": 1.8055999362331931e-06 }, { "epoch": 0.5072205094921305, "step": 1563, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.8471873520174995e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.821802834456321e-05 }, { "epoch": 0.5072205094921305, "step": 1563, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.4150713468552567e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.954333856701851e-05 }, { "epoch": 0.5072205094921305, "step": 1563, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3528803719964344e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.904671808006242e-05 }, { "epoch": 0.5075450267726757, "step": 1564, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5278794307960197e-05, "weight_min": 0.0, "weight_std": 2.1129708329681307e-05 }, { "epoch": 0.5075450267726757, "step": 1564, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.8821248128660955e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3806556125928182e-05 }, { "epoch": 0.5075450267726757, "step": 1564, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.240722576767439e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7220054360223003e-05 }, { "epoch": 0.5075450267726757, "step": 1564, "weight_max": 1.2938077986746066e-07, "weight_mean": 1.722751719057669e-08, "weight_min": 0.0, "weight_std": 4.5333738540875856e-08 }, { "epoch": 0.5078695440532208, "step": 1565, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4297448615252506e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0038401999045163e-05 }, { "epoch": 0.5078695440532208, "step": 1565, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2538595910882577e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.7667372958385386e-05 }, { "epoch": 0.5078695440532208, "step": 1565, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8459128114045598e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.016120561165735e-05 }, { "epoch": 0.5078695440532208, "step": 1565, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8299295334145427e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.981620880542323e-05 }, { "epoch": 0.508194061333766, "step": 1566, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1910436771577224e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8952485334011726e-05 }, { "epoch": 0.508194061333766, "step": 1566, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.157516615232453e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.695195351203438e-05 }, { "epoch": 0.508194061333766, "step": 1566, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.424082711309893e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9884650100721046e-05 }, { "epoch": 0.508194061333766, "step": 1566, "weight_max": 3.4268457966391e-05, "weight_mean": 7.3355586209800094e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.38326249725651e-05 }, { "epoch": 0.5085185786143113, "step": 1567, "weight_max": 2.5141427613561973e-05, "weight_mean": 3.2036709853855427e-06, "weight_min": 0.0, "weight_std": 8.865794370649382e-06 }, { "epoch": 0.5085185786143113, "step": 1567, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4252006621973123e-05, "weight_min": 7.508524357824342e-10, "weight_std": 1.978055661311373e-05 }, { "epoch": 0.5085185786143113, "step": 1567, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5355155483121052e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1223526346147992e-05 }, { "epoch": 0.5085185786143113, "step": 1567, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.6158829718478955e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.305305431742454e-05 }, { "epoch": 0.5088430958948564, "step": 1568, "weight_max": 3.998322426923551e-05, "weight_mean": 4.9985351324721705e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135948731563985e-05 }, { "epoch": 0.5088430958948564, "step": 1568, "weight_max": 3.998322426923551e-05, "weight_mean": 1.5860277926549315e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.755287667037919e-05 }, { "epoch": 0.5088430958948564, "step": 1568, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4412367818295024e-05, "weight_min": 0.0, "weight_std": 2.020123793045059e-05 }, { "epoch": 0.5088430958948564, "step": 1568, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5037437151477206e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0770075934706256e-05 }, { "epoch": 0.5091676131754016, "step": 1569, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.281755071948282e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.5846933820284903e-05 }, { "epoch": 0.5091676131754016, "step": 1569, "weight_max": 4.187939703115262e-05, "weight_mean": 6.021120043442352e-06, "weight_min": 0.0, "weight_std": 1.4654880942543969e-05 }, { "epoch": 0.5091676131754016, "step": 1569, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4173252566251904e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.982058893190697e-05 }, { "epoch": 0.5091676131754016, "step": 1569, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.2997470548725687e-05, "weight_min": 0.0, "weight_std": 1.7943642887985334e-05 }, { "epoch": 0.5094921304559468, "grad_norm": 10.636889457702637, "learning_rate": 7.008163265306123e-06, "loss": 0.154, "step": 1570 }, { "epoch": 0.5094921304559468, "step": 1570, "weight_max": 1.3004798347537871e-05, "weight_mean": 1.6260410120594315e-06, "weight_min": 0.0, "weight_std": 4.597712177201174e-06 }, { "epoch": 0.5094921304559468, "step": 1570, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3846785805071704e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9407225408940576e-05 }, { "epoch": 0.5094921304559468, "step": 1570, "weight_max": 2.6649584469851106e-05, "weight_mean": 3.3317974157398567e-06, "weight_min": 0.0, "weight_std": 9.421808499610052e-06 }, { "epoch": 0.5094921304559468, "step": 1570, "weight_max": 4.304629541707072e-09, "weight_mean": 1.053955811514129e-09, "weight_min": 0.0, "weight_std": 1.3368159912374722e-09 }, { "epoch": 0.5098166477364919, "step": 1571, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156796760275029e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.4580809875042178e-05 }, { "epoch": 0.5098166477364919, "step": 1571, "weight_max": 3.36925259034615e-05, "weight_mean": 7.741465196886566e-06, "weight_min": 0.0, "weight_std": 1.4325159099826124e-05 }, { "epoch": 0.5098166477364919, "step": 1571, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4270251085690688e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.973166217794642e-05 }, { "epoch": 0.5098166477364919, "step": 1571, "weight_max": 4.252122016623616e-05, "weight_mean": 1.784942469384987e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9760871509788558e-05 }, { "epoch": 0.5101411650170372, "step": 1572, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1951899068662897e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6919506379053928e-05 }, { "epoch": 0.5101411650170372, "step": 1572, "weight_max": 4.187939703115262e-05, "weight_mean": 9.411557584826369e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.757129211910069e-05 }, { "epoch": 0.5101411650170372, "step": 1572, "weight_max": 3.1715208024252206e-05, "weight_mean": 5.204875833442202e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.1259869097557385e-05 }, { "epoch": 0.5101411650170372, "step": 1572, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998660642741015e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4135897799860686e-05 }, { "epoch": 0.5104656822975824, "step": 1573, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.399731425219215e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7389618733432144e-05 }, { "epoch": 0.5104656822975824, "step": 1573, "weight_max": 2.872879122151062e-05, "weight_mean": 3.59853856934933e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0154174560739193e-05 }, { "epoch": 0.5104656822975824, "step": 1573, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1349104170221835e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9346050976309925e-05 }, { "epoch": 0.5104656822975824, "step": 1573, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4689740964968223e-05, "weight_min": 0.0, "weight_std": 2.030018367804587e-05 }, { "epoch": 0.5107901995781275, "step": 1574, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4039202142157592e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9498627807479352e-05 }, { "epoch": 0.5107901995781275, "step": 1574, "weight_max": 2.5890016331686638e-05, "weight_mean": 6.131682312116027e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0995535376423504e-05 }, { "epoch": 0.5107901995781275, "step": 1574, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.50367159221787e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0766377929248847e-05 }, { "epoch": 0.5107901995781275, "step": 1574, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.07783306602505e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357326108438428e-05 }, { "epoch": 0.5111147168586727, "step": 1575, "weight_max": 3.6920544516760856e-05, "weight_mean": 6.110590675234562e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3133397260389756e-05 }, { "epoch": 0.5111147168586727, "step": 1575, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0842244339291938e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.209165359090548e-05 }, { "epoch": 0.5111147168586727, "step": 1575, "weight_max": 3.998322426923551e-05, "weight_mean": 1.2845983292208984e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8283570170751773e-05 }, { "epoch": 0.5111147168586727, "step": 1575, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.461934152757749e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7833192032412626e-05 }, { "epoch": 0.5114392341392179, "step": 1576, "weight_max": 1.3612994109735155e-07, "weight_mean": 1.7674988583848972e-08, "weight_min": 0.0, "weight_std": 4.786562612935086e-08 }, { "epoch": 0.5114392341392179, "step": 1576, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8574748537503183e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0162848159088753e-05 }, { "epoch": 0.5114392341392179, "step": 1576, "weight_max": 1.2081417253284599e-06, "weight_mean": 1.5194933666862198e-07, "weight_min": 0.0, "weight_std": 4.267667748081294e-07 }, { "epoch": 0.5114392341392179, "step": 1576, "weight_max": 3.9051607018336654e-05, "weight_mean": 8.74965189723298e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.5654351955163293e-05 }, { "epoch": 0.5117637514197632, "step": 1577, "weight_max": 3.4849264920921996e-05, "weight_mean": 7.256560820678715e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3791291166853625e-05 }, { "epoch": 0.5117637514197632, "step": 1577, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959589659847552e-06, "weight_min": 0.0, "weight_std": 1.4025716154719703e-05 }, { "epoch": 0.5117637514197632, "step": 1577, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3031248272454832e-05, "weight_min": 0.0, "weight_std": 1.7577871403773315e-05 }, { "epoch": 0.5117637514197632, "step": 1577, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320400876167696e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2217939001857303e-05 }, { "epoch": 0.5120882687003083, "step": 1578, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.509220055595506e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0913250409648754e-05 }, { "epoch": 0.5120882687003083, "step": 1578, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315785074344603e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5033265299280174e-05 }, { "epoch": 0.5120882687003083, "step": 1578, "weight_max": 3.752332122530788e-05, "weight_mean": 6.060642590455245e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3278112419357058e-05 }, { "epoch": 0.5120882687003083, "step": 1578, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2847090147261042e-05, "weight_min": 7.508524357824342e-10, "weight_std": 1.647475073696114e-05 }, { "epoch": 0.5124127859808535, "step": 1579, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.519818397355266e-05, "weight_min": 0.0, "weight_std": 2.1011817807448097e-05 }, { "epoch": 0.5124127859808535, "step": 1579, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2815726222470403e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8428405383019708e-05 }, { "epoch": 0.5124127859808535, "step": 1579, "weight_max": 4.187939703115262e-05, "weight_mean": 9.702017450763378e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.803749182727188e-05 }, { "epoch": 0.5124127859808535, "step": 1579, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.421017239219509e-06, "weight_min": 0.0, "weight_std": 1.745243207551539e-05 }, { "epoch": 0.5127373032613987, "grad_norm": 3.0267744064331055, "learning_rate": 6.987755102040817e-06, "loss": 0.1732, "step": 1580 }, { "epoch": 0.5127373032613987, "step": 1580, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.2443746527424082e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.743044049362652e-05 }, { "epoch": 0.5127373032613987, "step": 1580, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3648358617501799e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.910198807308916e-05 }, { "epoch": 0.5127373032613987, "step": 1580, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.036424120888114e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.184939330618363e-05 }, { "epoch": 0.5127373032613987, "step": 1580, "weight_max": 3.904611730831675e-06, "weight_mean": 5.500845645656227e-07, "weight_min": 0.0, "weight_std": 1.3662537412528764e-06 }, { "epoch": 0.5130618205419438, "step": 1581, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2784031241608318e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9296214304631576e-05 }, { "epoch": 0.5130618205419438, "step": 1581, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5122941476874985e-05, "weight_min": 0.0, "weight_std": 2.0920146198477596e-05 }, { "epoch": 0.5130618205419438, "step": 1581, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3547549315262586e-05, "weight_min": 0.0, "weight_std": 1.8864911908167414e-05 }, { "epoch": 0.5130618205419438, "step": 1581, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077464720670832e-06, "weight_min": 0.0, "weight_std": 1.435747526556952e-05 }, { "epoch": 0.5133863378224891, "step": 1582, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3763801689492539e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.912630432343576e-05 }, { "epoch": 0.5133863378224891, "step": 1582, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.15646388521418e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.458094448025804e-05 }, { "epoch": 0.5133863378224891, "step": 1582, "weight_max": 3.813097282545641e-05, "weight_mean": 4.76760715173441e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3480836059898138e-05 }, { "epoch": 0.5133863378224891, "step": 1582, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105578682356281e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1609564353420865e-05 }, { "epoch": 0.5137108551030343, "step": 1583, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.45991268963553e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7526326701045036e-05 }, { "epoch": 0.5137108551030343, "step": 1583, "weight_max": 3.813097282545641e-05, "weight_mean": 4.7670400817878544e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3481065252562985e-05 }, { "epoch": 0.5137108551030343, "step": 1583, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0314302016922738e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.91073431778932e-05 }, { "epoch": 0.5137108551030343, "step": 1583, "weight_max": 3.752332122530788e-05, "weight_mean": 6.9208317654556595e-06, "weight_min": 0.0, "weight_std": 1.3851637049810961e-05 }, { "epoch": 0.5140353723835794, "step": 1584, "weight_max": 2.088051951432135e-05, "weight_mean": 2.6106768018507864e-06, "weight_min": 0.0, "weight_std": 7.382131116173696e-06 }, { "epoch": 0.5140353723835794, "step": 1584, "weight_max": 3.116124207735993e-05, "weight_mean": 6.7439568738336675e-06, "weight_min": 0.0, "weight_std": 1.2551759937196039e-05 }, { "epoch": 0.5140353723835794, "step": 1584, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.974217391572893e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.8549879314377904e-05 }, { "epoch": 0.5140353723835794, "step": 1584, "weight_max": 3.843663216684945e-05, "weight_mean": 1.2917485946672969e-05, "weight_min": 0.0, "weight_std": 1.8103120964951813e-05 }, { "epoch": 0.5143598896641246, "step": 1585, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.572506153024733e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.835342300182674e-05 }, { "epoch": 0.5143598896641246, "step": 1585, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.837049421737902e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.94571512110997e-05 }, { "epoch": 0.5143598896641246, "step": 1585, "weight_max": 3.998322426923551e-05, "weight_mean": 9.068124199984595e-06, "weight_min": 0.0, "weight_std": 1.6905873053474352e-05 }, { "epoch": 0.5143598896641246, "step": 1585, "weight_max": 3.662098970380612e-05, "weight_mean": 4.5781330300087575e-06, "weight_min": 0.0, "weight_std": 1.2947269169671927e-05 }, { "epoch": 0.5146844069446698, "step": 1586, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3056647730991244e-05, "weight_min": 0.0, "weight_std": 1.8948954675579444e-05 }, { "epoch": 0.5146844069446698, "step": 1586, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8830553017323837e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0263794795027934e-05 }, { "epoch": 0.5146844069446698, "step": 1586, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.52405662246747e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7644104445935227e-05 }, { "epoch": 0.5146844069446698, "step": 1586, "weight_max": 3.752332122530788e-05, "weight_mean": 8.655300916871056e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6100311768241227e-05 }, { "epoch": 0.515008924225215, "step": 1587, "weight_max": 1.5955802235012584e-09, "weight_mean": 7.247084599093512e-10, "weight_min": 0.0, "weight_std": 4.575783363325314e-10 }, { "epoch": 0.515008924225215, "step": 1587, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3005021173739806e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.82825588126434e-05 }, { "epoch": 0.515008924225215, "step": 1587, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.33714909479022e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.740061998134479e-05 }, { "epoch": 0.515008924225215, "step": 1587, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965601536037866e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1212534445803612e-05 }, { "epoch": 0.5153334415057602, "step": 1588, "weight_max": 1.6735920427279893e-09, "weight_mean": 7.983369521014083e-10, "weight_min": 5.958802340444436e-10, "weight_std": 3.6717451301626625e-10 }, { "epoch": 0.5153334415057602, "step": 1588, "weight_max": 3.843663216684945e-05, "weight_mean": 2.334273449378088e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.627421261218842e-05 }, { "epoch": 0.5153334415057602, "step": 1588, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077098194306018e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4357623513205908e-05 }, { "epoch": 0.5153334415057602, "step": 1588, "weight_max": 4.187939703115262e-05, "weight_mean": 1.216765394929098e-05, "weight_min": 0.0, "weight_std": 1.8666687537916005e-05 }, { "epoch": 0.5156579587863054, "step": 1589, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965514679293847e-06, "weight_min": 0.0, "weight_std": 1.121256900660228e-05 }, { "epoch": 0.5156579587863054, "step": 1589, "weight_max": 4.252122016623616e-05, "weight_mean": 1.396089282934554e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.944390169228427e-05 }, { "epoch": 0.5156579587863054, "step": 1589, "weight_max": 3.9360926166409627e-05, "weight_mean": 5.894347395951627e-06, "weight_min": 0.0, "weight_std": 1.3794537153444253e-05 }, { "epoch": 0.5156579587863054, "step": 1589, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960446732293349e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8457019905326888e-05 }, { "epoch": 0.5159824760668505, "grad_norm": 0.040605753660202026, "learning_rate": 6.967346938775511e-06, "loss": 0.2893, "step": 1590 }, { "epoch": 0.5159824760668505, "step": 1590, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9151990272803232e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0669915102189407e-05 }, { "epoch": 0.5159824760668505, "step": 1590, "weight_max": 3.5141503758495674e-05, "weight_mean": 5.8148752941633575e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.2447178960428573e-05 }, { "epoch": 0.5159824760668505, "step": 1590, "weight_max": 2.4786623864514468e-09, "weight_mean": 9.369434117445508e-10, "weight_min": 5.679522963042416e-10, "weight_std": 6.507764882535128e-10 }, { "epoch": 0.5159824760668505, "step": 1590, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4110153642832302e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9636214346974157e-05 }, { "epoch": 0.5163069933473957, "step": 1591, "weight_max": 4.187939703115262e-05, "weight_mean": 8.534918379154988e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.633502142794896e-05 }, { "epoch": 0.5163069933473957, "step": 1591, "weight_max": 4.154041999271385e-09, "weight_mean": 1.3928616038327846e-09, "weight_min": 5.958802340444436e-10, "weight_std": 1.1731481341925587e-09 }, { "epoch": 0.5163069933473957, "step": 1591, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9221455659135245e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9775032342295162e-05 }, { "epoch": 0.5163069933473957, "step": 1591, "weight_max": 3.843663216684945e-05, "weight_mean": 1.8299244402442127e-05, "weight_min": 7.829137893544669e-10, "weight_std": 1.9580234948080033e-05 }, { "epoch": 0.516631510627941, "step": 1592, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729170086648082e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3373353795032017e-05 }, { "epoch": 0.516631510627941, "step": 1592, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232139175059274e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8945682313642465e-05 }, { "epoch": 0.516631510627941, "step": 1592, "weight_max": 3.782653220696375e-05, "weight_mean": 9.1218880697852e-06, "weight_min": 0.0, "weight_std": 1.690351200522855e-05 }, { "epoch": 0.516631510627941, "step": 1592, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3186578144086525e-05, "weight_min": 0.0, "weight_std": 1.8534017726778984e-05 }, { "epoch": 0.5169560279084862, "step": 1593, "weight_max": 3.003409743129737e-09, "weight_mean": 6.540867292237351e-10, "weight_min": 0.0, "weight_std": 1.0198051292320542e-09 }, { "epoch": 0.5169560279084862, "step": 1593, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0470242159499321e-05, "weight_min": 0.0, "weight_std": 1.9386152416700497e-05 }, { "epoch": 0.5169560279084862, "step": 1593, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920927494822536e-06, "weight_min": 0.0, "weight_std": 1.3915861018176656e-05 }, { "epoch": 0.5169560279084862, "step": 1593, "weight_max": 3.998322426923551e-05, "weight_mean": 1.9491320927045308e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0852015950367786e-05 }, { "epoch": 0.5172805451890313, "step": 1594, "weight_max": 2.7939937353949063e-05, "weight_mean": 3.4931013033201452e-06, "weight_min": 5.958802340444436e-10, "weight_std": 9.878013770503458e-06 }, { "epoch": 0.5172805451890313, "step": 1594, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0036365893029142e-05, "weight_min": 0.0, "weight_std": 1.858263749454636e-05 }, { "epoch": 0.5172805451890313, "step": 1594, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0233319699182175e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8953987819259055e-05 }, { "epoch": 0.5172805451890313, "step": 1594, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.1753812032111455e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.6766252883826382e-05 }, { "epoch": 0.5176050624695765, "step": 1595, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.3203708628425375e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2217949915793724e-05 }, { "epoch": 0.5176050624695765, "step": 1595, "weight_max": 2.0739674422287635e-09, "weight_mean": 8.435290799191364e-10, "weight_min": 5.958802340444436e-10, "weight_std": 5.072829933894241e-10 }, { "epoch": 0.5176050624695765, "step": 1595, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.5721790987299755e-05, "weight_min": 0.0, "weight_std": 1.725727088341955e-05 }, { "epoch": 0.5176050624695765, "step": 1595, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.622271136322524e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.787593100743834e-05 }, { "epoch": 0.5179295797501217, "step": 1596, "weight_max": 3.752332122530788e-05, "weight_mean": 1.218960460391827e-05, "weight_min": 0.0, "weight_std": 1.7179185306304134e-05 }, { "epoch": 0.5179295797501217, "step": 1596, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4631110389018431e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.0297064111218788e-05 }, { "epoch": 0.5179295797501217, "step": 1596, "weight_max": 2.877845961535286e-09, "weight_mean": 9.777310072678347e-10, "weight_min": 0.0, "weight_std": 8.682431440476535e-10 }, { "epoch": 0.5179295797501217, "step": 1596, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5222005458781496e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.114203743985854e-05 }, { "epoch": 0.5182540970306669, "step": 1597, "weight_max": 4.252122016623616e-05, "weight_mean": 1.673468796070665e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9070294001721777e-05 }, { "epoch": 0.5182540970306669, "step": 1597, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4237091818358749e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9706119928741828e-05 }, { "epoch": 0.5182540970306669, "step": 1597, "weight_max": 4.252122016623616e-05, "weight_mean": 8.427479770034552e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.629900543775875e-05 }, { "epoch": 0.5182540970306669, "step": 1597, "weight_max": 4.187939703115262e-05, "weight_mean": 1.926574987010099e-05, "weight_min": 0.0, "weight_std": 2.078678699035663e-05 }, { "epoch": 0.5185786143112121, "step": 1598, "weight_max": 4.187939703115262e-05, "weight_mean": 2.6301087928004563e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.7457117792218924e-05 }, { "epoch": 0.5185786143112121, "step": 1598, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284110218577553e-06, "weight_min": 0.0, "weight_std": 1.2115506251575425e-05 }, { "epoch": 0.5185786143112121, "step": 1598, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.078206413600128e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357176951307338e-05 }, { "epoch": 0.5185786143112121, "step": 1598, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8694656318984926e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0182151274639182e-05 }, { "epoch": 0.5189031315917573, "step": 1599, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3317959201231133e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8507158529246226e-05 }, { "epoch": 0.5189031315917573, "step": 1599, "weight_max": 3.998322426923551e-05, "weight_mean": 9.996319931815378e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8508319044485688e-05 }, { "epoch": 0.5189031315917573, "step": 1599, "weight_max": 4.252122016623616e-05, "weight_mean": 5.31696832695161e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.5032787814561743e-05 }, { "epoch": 0.5189031315917573, "step": 1599, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4545774320140481e-05, "weight_min": 0.0, "weight_std": 2.0149233023403212e-05 }, { "epoch": 0.5192276488723024, "grad_norm": 0.12749135494232178, "learning_rate": 6.9469387755102045e-06, "loss": 0.214, "step": 1600 }, { "epoch": 0.5192276488723024, "eval_accuracy": 0.9905, "eval_auc": 0.995909911027188, "eval_loss": 0.07020853459835052, "eval_runtime": 156.1136, "eval_samples_per_second": 12.811, "eval_steps_per_second": 1.601, "step": 1600 }, { "epoch": 0.5192276488723024, "step": 1600, "weight_max": 3.813097282545641e-05, "weight_mean": 8.801143849268556e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.637028981349431e-05 }, { "epoch": 0.5192276488723024, "step": 1600, "weight_max": 4.252122016623616e-05, "weight_mean": 1.482702100474853e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0579955162247643e-05 }, { "epoch": 0.5192276488723024, "step": 1600, "weight_max": 3.36925259034615e-05, "weight_mean": 9.33403498493135e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4906201613484882e-05 }, { "epoch": 0.5192276488723024, "step": 1600, "weight_max": 3.5141503758495674e-05, "weight_mean": 9.135459549725056e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5222143701976165e-05 }, { "epoch": 0.5195521661528476, "step": 1601, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156518454896286e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4580922652385198e-05 }, { "epoch": 0.5195521661528476, "step": 1601, "weight_max": 4.187939703115262e-05, "weight_mean": 1.941075606737286e-05, "weight_min": 0.0, "weight_std": 2.0907029465888627e-05 }, { "epoch": 0.5195521661528476, "step": 1601, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3454814506985713e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.86512443178799e-05 }, { "epoch": 0.5195521661528476, "step": 1601, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.348384992335923e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.7330799892079085e-05 }, { "epoch": 0.5198766834333929, "step": 1602, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156045972398715e-06, "weight_min": 0.0, "weight_std": 1.4581114555767272e-05 }, { "epoch": 0.5198766834333929, "step": 1602, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.032363772741519e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1791463950648904e-05 }, { "epoch": 0.5198766834333929, "step": 1602, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3995933588594198e-05, "weight_min": 7.614652242082798e-10, "weight_std": 2.0086044969502836e-05 }, { "epoch": 0.5198766834333929, "step": 1602, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5388561223517172e-05, "weight_min": 0.0, "weight_std": 2.1243555238470435e-05 }, { "epoch": 0.520201200713938, "step": 1603, "weight_max": 3.813097282545641e-05, "weight_mean": 4.7673161134298425e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3480952475219965e-05 }, { "epoch": 0.520201200713938, "step": 1603, "weight_max": 1.1511383846141143e-08, "weight_mean": 2.0450365845192664e-09, "weight_min": 0.0, "weight_std": 3.840715301350883e-09 }, { "epoch": 0.520201200713938, "step": 1603, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320616881159367e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2217851690365933e-05 }, { "epoch": 0.520201200713938, "step": 1603, "weight_max": 3.967146767536178e-05, "weight_mean": 8.993622031994164e-06, "weight_min": 0.0, "weight_std": 1.6768841305747628e-05 }, { "epoch": 0.5205257179944832, "step": 1604, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4651430319645442e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.02488499780884e-05 }, { "epoch": 0.5205257179944832, "step": 1604, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0792538887471892e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.9982519006589428e-05 }, { "epoch": 0.5205257179944832, "step": 1604, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3966477126814425e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9477602108963765e-05 }, { "epoch": 0.5205257179944832, "step": 1604, "weight_max": 9.309319537464944e-09, "weight_mean": 2.036303126118355e-09, "weight_min": 5.679522963042416e-10, "weight_std": 2.971270118834468e-09 }, { "epoch": 0.5208502352750284, "step": 1605, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0471189852978569e-05, "weight_min": 0.0, "weight_std": 1.9385566702112556e-05 }, { "epoch": 0.5208502352750284, "step": 1605, "weight_max": 4.252122016623616e-05, "weight_mean": 1.432242333976319e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0016164853586815e-05 }, { "epoch": 0.5208502352750284, "step": 1605, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5278757928172126e-05, "weight_min": 0.0, "weight_std": 2.1129739252501167e-05 }, { "epoch": 0.5208502352750284, "step": 1605, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5232177247526124e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1019210180384107e-05 }, { "epoch": 0.5211747525555736, "step": 1606, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8631490092957392e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0162196960882284e-05 }, { "epoch": 0.5211747525555736, "step": 1606, "weight_max": 4.252122016623616e-05, "weight_mean": 5.961085662420373e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.488260295445798e-05 }, { "epoch": 0.5211747525555736, "step": 1606, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.6904170479392633e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8691671357373707e-05 }, { "epoch": 0.5211747525555736, "step": 1606, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3955152098787948e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.953614264493808e-05 }, { "epoch": 0.5214992698361188, "step": 1607, "weight_max": 3.998322426923551e-05, "weight_mean": 8.079727194854058e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5510131561313756e-05 }, { "epoch": 0.5214992698361188, "step": 1607, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9643435734906234e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1163432393223047e-05 }, { "epoch": 0.5214992698361188, "step": 1607, "weight_max": 1.5955802235012584e-09, "weight_mean": 9.066358219733672e-10, "weight_min": 5.958802340444436e-10, "weight_std": 3.4469377352408515e-10 }, { "epoch": 0.5214992698361188, "step": 1607, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3625855899590533e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.913994856295176e-05 }, { "epoch": 0.521823787116664, "step": 1608, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4778146578464657e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0467987269512378e-05 }, { "epoch": 0.521823787116664, "step": 1608, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8825539882527664e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.053542084468063e-05 }, { "epoch": 0.521823787116664, "step": 1608, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2355658226588275e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4806344552198425e-05 }, { "epoch": 0.521823787116664, "step": 1608, "weight_max": 3.602553260861896e-05, "weight_mean": 1.3180000678403303e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.819494536903221e-05 }, { "epoch": 0.5221483043972092, "step": 1609, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.676463039591908e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6064584997366183e-05 }, { "epoch": 0.5221483043972092, "step": 1609, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9590330339269713e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1047357222414576e-05 }, { "epoch": 0.5221483043972092, "step": 1609, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.008775143418461e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.873639484983869e-05 }, { "epoch": 0.5221483043972092, "step": 1609, "weight_max": 3.572963396436535e-05, "weight_mean": 4.4668731788988225e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.263206377188908e-05 }, { "epoch": 0.5224728216777543, "grad_norm": 1.1035503149032593, "learning_rate": 6.9265306122448986e-06, "loss": 0.0899, "step": 1610 }, { "epoch": 0.5224728216777543, "step": 1610, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.3648241040064022e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9937999240937643e-05 }, { "epoch": 0.5224728216777543, "step": 1610, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.320374046073994e-06, "weight_min": 0.0, "weight_std": 1.2217949006299023e-05 }, { "epoch": 0.5224728216777543, "step": 1610, "weight_max": 3.998322426923551e-05, "weight_mean": 1.1334771443216596e-05, "weight_min": 0.0, "weight_std": 1.6353249520761892e-05 }, { "epoch": 0.5224728216777543, "step": 1610, "weight_max": 3.909020662717921e-09, "weight_mean": 1.3224028538871835e-09, "weight_min": 6.196655966128617e-10, "weight_std": 1.2115002334667224e-09 }, { "epoch": 0.5227973389582995, "step": 1611, "weight_max": 3.998322426923551e-05, "weight_mean": 9.53882590692956e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.768808942870237e-05 }, { "epoch": 0.5227973389582995, "step": 1611, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.438390078576049e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2524528756330255e-05 }, { "epoch": 0.5227973389582995, "step": 1611, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0236124580842443e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8970316887134686e-05 }, { "epoch": 0.5227973389582995, "step": 1611, "weight_max": 3.22740561387036e-05, "weight_mean": 4.034649009554414e-06, "weight_min": 0.0, "weight_std": 1.1410444130888209e-05 }, { "epoch": 0.5231218562388448, "step": 1612, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.1315873052808456e-05, "weight_min": 0.0, "weight_std": 1.6245692677330226e-05 }, { "epoch": 0.5231218562388448, "step": 1612, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4489644854620565e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0133371435804293e-05 }, { "epoch": 0.5231218562388448, "step": 1612, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8744713088381104e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.027784830715973e-05 }, { "epoch": 0.5231218562388448, "step": 1612, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2657586669083685e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.9625984350568615e-05 }, { "epoch": 0.5234463735193899, "step": 1613, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.393360985792242e-06, "weight_min": 0.0, "weight_std": 1.5071991583681665e-05 }, { "epoch": 0.5234463735193899, "step": 1613, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155925009603379e-06, "weight_min": 0.0, "weight_std": 1.4581161849491764e-05 }, { "epoch": 0.5234463735193899, "step": 1613, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3739228961640038e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9292103388579562e-05 }, { "epoch": 0.5234463735193899, "step": 1613, "weight_max": 4.187939703115262e-05, "weight_mean": 9.813616088649724e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.8222921426058747e-05 }, { "epoch": 0.5237708907999351, "step": 1614, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0552182175160851e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.95434367924463e-05 }, { "epoch": 0.5237708907999351, "step": 1614, "weight_max": 3.998322426923551e-05, "weight_mean": 9.880018296826165e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.8294293113285676e-05 }, { "epoch": 0.5237708907999351, "step": 1614, "weight_max": 4.187939703115262e-05, "weight_mean": 9.340306860394776e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.746127600199543e-05 }, { "epoch": 0.5237708907999351, "step": 1614, "weight_max": 4.252122016623616e-05, "weight_mean": 1.7380527424393222e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9910954506485723e-05 }, { "epoch": 0.5240954080804803, "step": 1615, "weight_max": 3.722132169059478e-05, "weight_mean": 5.029522526456276e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.3032587958150543e-05 }, { "epoch": 0.5240954080804803, "step": 1615, "weight_max": 8.716874999592505e-10, "weight_mean": 7.120536382743126e-10, "weight_min": 5.771870759119224e-10, "weight_std": 9.731317140992957e-11 }, { "epoch": 0.5240954080804803, "step": 1615, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.360664989799261e-06, "weight_min": 0.0, "weight_std": 1.7413278328604065e-05 }, { "epoch": 0.5240954080804803, "step": 1615, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.99702388071455e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.8512404494686052e-05 }, { "epoch": 0.5244199253610254, "step": 1616, "weight_max": 4.252122016623616e-05, "weight_mean": 1.775957935024053e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9924795196857303e-05 }, { "epoch": 0.5244199253610254, "step": 1616, "weight_max": 4.187939703115262e-05, "weight_mean": 5.272161160974065e-06, "weight_min": 0.0, "weight_std": 1.4791908142797183e-05 }, { "epoch": 0.5244199253610254, "step": 1616, "weight_max": 2.464847420924343e-05, "weight_mean": 3.086609240199323e-06, "weight_min": 6.053385570581327e-10, "weight_std": 8.712319868209306e-06 }, { "epoch": 0.5244199253610254, "step": 1616, "weight_max": 3.998322426923551e-05, "weight_mean": 5.3462049436348025e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4029204066901002e-05 }, { "epoch": 0.5247444426415707, "step": 1617, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.740239531500265e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.865978083515074e-05 }, { "epoch": 0.5247444426415707, "step": 1617, "weight_max": 3.5141503758495674e-05, "weight_mean": 7.352784450631589e-06, "weight_min": 0.0, "weight_std": 1.395414710714249e-05 }, { "epoch": 0.5247444426415707, "step": 1617, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691238700615941e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3266164387459867e-05 }, { "epoch": 0.5247444426415707, "step": 1617, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.6907106328289956e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9389541193959303e-05 }, { "epoch": 0.5250689599221159, "step": 1618, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3271197531139478e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8557206203695387e-05 }, { "epoch": 0.5250689599221159, "step": 1618, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.8434571908728685e-06, "weight_min": 0.0, "weight_std": 1.3697688700631261e-05 }, { "epoch": 0.5250689599221159, "step": 1618, "weight_max": 3.143761568935588e-05, "weight_mean": 3.931079845642671e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.1114319022453856e-05 }, { "epoch": 0.5250689599221159, "step": 1618, "weight_max": 4.187939703115262e-05, "weight_mean": 7.11511165718548e-06, "weight_min": 0.0, "weight_std": 1.5000163330114447e-05 }, { "epoch": 0.525393477202661, "step": 1619, "weight_max": 7.614652242082798e-10, "weight_mean": 6.656675211047514e-10, "weight_min": 6.196655966128617e-10, "weight_std": 4.9188333961502906e-11 }, { "epoch": 0.525393477202661, "step": 1619, "weight_max": 4.381950566312298e-05, "weight_mean": 2.5627559807617217e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.126049730577506e-05 }, { "epoch": 0.525393477202661, "step": 1619, "weight_max": 2.926079287135508e-05, "weight_mean": 9.234706340066623e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2971936484973412e-05 }, { "epoch": 0.525393477202661, "step": 1619, "weight_max": 4.187939703115262e-05, "weight_mean": 9.459186912863515e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.762006650096737e-05 }, { "epoch": 0.5257179944832062, "grad_norm": 117.31742095947266, "learning_rate": 6.906122448979592e-06, "loss": 0.128, "step": 1620 }, { "epoch": 0.5257179944832062, "step": 1620, "weight_max": 3.486749999837002e-09, "weight_mean": 1.220825440739759e-09, "weight_min": 0.0, "weight_std": 1.1207162975424012e-09 }, { "epoch": 0.5257179944832062, "step": 1620, "weight_max": 3.752332122530788e-05, "weight_mean": 9.157164640782867e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6961204892140813e-05 }, { "epoch": 0.5257179944832062, "step": 1620, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.892624823085498e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.0447138012968935e-05 }, { "epoch": 0.5257179944832062, "step": 1620, "weight_max": 2.7679425329552032e-05, "weight_mean": 6.634513283643173e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2297824468987528e-05 }, { "epoch": 0.5260425117637514, "step": 1621, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2245389370946214e-05, "weight_min": 0.0, "weight_std": 1.7384829334332608e-05 }, { "epoch": 0.5260425117637514, "step": 1621, "weight_max": 7.906837709015235e-07, "weight_mean": 9.96619391457898e-08, "weight_min": 0.0, "weight_std": 2.7921618084292277e-07 }, { "epoch": 0.5260425117637514, "step": 1621, "weight_max": 3.312147964606993e-05, "weight_mean": 4.141198587603867e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1709802492987365e-05 }, { "epoch": 0.5260425117637514, "step": 1621, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.579881978221238e-06, "weight_min": 0.0, "weight_std": 1.594663262949325e-05 }, { "epoch": 0.5263670290442967, "step": 1622, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396633241616655e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5261904991348274e-05 }, { "epoch": 0.5263670290442967, "step": 1622, "weight_max": 3.143761568935588e-05, "weight_mean": 3.930585535272257e-06, "weight_min": 0.0, "weight_std": 1.1114517292298842e-05 }, { "epoch": 0.5263670290442967, "step": 1622, "weight_max": 3.340639523230493e-05, "weight_mean": 7.767610441078432e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.443529617972672e-05 }, { "epoch": 0.5263670290442967, "step": 1622, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313478924217634e-05, "weight_min": 0.0, "weight_std": 1.910785067593679e-05 }, { "epoch": 0.5266915463248418, "step": 1623, "weight_max": 4.252122016623616e-05, "weight_mean": 1.298040479014162e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.895085006253794e-05 }, { "epoch": 0.5266915463248418, "step": 1623, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653403266274836e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3159426089259796e-05 }, { "epoch": 0.5266915463248418, "step": 1623, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232145541522186e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8945678675663657e-05 }, { "epoch": 0.5266915463248418, "step": 1623, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.569514648115728e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.6223229977185838e-05 }, { "epoch": 0.527016063605387, "step": 1624, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.018862531054765e-05, "weight_min": 0.0, "weight_std": 1.7765938537195325e-05 }, { "epoch": 0.527016063605387, "step": 1624, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8705461116041988e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.007506918744184e-05 }, { "epoch": 0.527016063605387, "step": 1624, "weight_max": 3.722132169059478e-05, "weight_mean": 1.2659301319217775e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.757464633556083e-05 }, { "epoch": 0.527016063605387, "step": 1624, "weight_max": 1.3870513626557113e-09, "weight_mean": 8.276194729539554e-10, "weight_min": 5.679522963042416e-10, "weight_std": 3.0261640415751856e-10 }, { "epoch": 0.5273405808859322, "step": 1625, "weight_max": 3.632264997577295e-05, "weight_mean": 4.540953341347631e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2841745956393424e-05 }, { "epoch": 0.5273405808859322, "step": 1625, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.131790648098104e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.7255482816835865e-05 }, { "epoch": 0.5273405808859322, "step": 1625, "weight_max": 4.252122016623616e-05, "weight_mean": 1.843532118073199e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0071460312465206e-05 }, { "epoch": 0.5273405808859322, "step": 1625, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4134738194115926e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7230428056791425e-05 }, { "epoch": 0.5276650981664773, "step": 1626, "weight_max": 4.252122016623616e-05, "weight_mean": 2.370836227783002e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9917470126529224e-05 }, { "epoch": 0.5276650981664773, "step": 1626, "weight_max": 4.381950566312298e-05, "weight_mean": 5.478860657603946e-06, "weight_min": 0.0, "weight_std": 1.5491959857172333e-05 }, { "epoch": 0.5276650981664773, "step": 1626, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653429641621187e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.3159415175323375e-05 }, { "epoch": 0.5276650981664773, "step": 1626, "weight_max": 3.813097282545641e-05, "weight_mean": 6.188456154632149e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.3267690519569442e-05 }, { "epoch": 0.5279896154470226, "step": 1627, "weight_max": 3.5434961318969727e-05, "weight_mean": 5.861711088073207e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2603185496118385e-05 }, { "epoch": 0.5279896154470226, "step": 1627, "weight_max": 3.722132169059478e-05, "weight_mean": 4.653996256820392e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.315918689215323e-05 }, { "epoch": 0.5279896154470226, "step": 1627, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5359953977167606e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.124945240211673e-05 }, { "epoch": 0.5279896154470226, "step": 1627, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.424919901182875e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.972759491764009e-05 }, { "epoch": 0.5283141327275678, "step": 1628, "weight_max": 3.0886087188264355e-05, "weight_mean": 3.861406185023952e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.0919619853666518e-05 }, { "epoch": 0.5283141327275678, "step": 1628, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.601966328569688e-06, "weight_min": 0.0, "weight_std": 1.6094320017145947e-05 }, { "epoch": 0.5283141327275678, "step": 1628, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5043930034153163e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.080983540508896e-05 }, { "epoch": 0.5283141327275678, "step": 1628, "weight_max": 2.464847420924343e-05, "weight_mean": 3.085054004259291e-06, "weight_min": 6.053385570581327e-10, "weight_std": 8.71294287208002e-06 }, { "epoch": 0.5286386500081129, "step": 1629, "weight_max": 3.813097282545641e-05, "weight_mean": 4.769800398207735e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.3479951121553313e-05 }, { "epoch": 0.5286386500081129, "step": 1629, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.115256605378818e-06, "weight_min": 0.0, "weight_std": 1.5079285731189884e-05 }, { "epoch": 0.5286386500081129, "step": 1629, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236825018073432e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4805835235165432e-05 }, { "epoch": 0.5286386500081129, "step": 1629, "weight_max": 3.397988257347606e-05, "weight_mean": 8.353035809705034e-06, "weight_min": 0.0, "weight_std": 1.5467785487999208e-05 }, { "epoch": 0.5289631672886581, "grad_norm": 0.7209358811378479, "learning_rate": 6.885714285714287e-06, "loss": 0.2703, "step": 1630 }, { "epoch": 0.5289631672886581, "step": 1630, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.8441779654240236e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3697396752831992e-05 }, { "epoch": 0.5289631672886581, "step": 1630, "weight_max": 4.381950566312298e-05, "weight_mean": 6.122524609963875e-06, "weight_min": 7.403143098549947e-10, "weight_std": 1.531114503450226e-05 }, { "epoch": 0.5289631672886581, "step": 1630, "weight_max": 3.843663216684945e-05, "weight_mean": 4.806345714314375e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.358868848910788e-05 }, { "epoch": 0.5289631672886581, "step": 1630, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5873552911216393e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1894165911362506e-05 }, { "epoch": 0.5292876845692033, "step": 1631, "weight_max": 2.021574685429073e-09, "weight_mean": 8.147109098466387e-10, "weight_min": 0.0, "weight_std": 5.774114519851992e-10 }, { "epoch": 0.5292876845692033, "step": 1631, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.830504697863944e-05, "weight_min": 9.062063877074422e-10, "weight_std": 1.970233097381424e-05 }, { "epoch": 0.5292876845692033, "step": 1631, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.746373062720522e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4437528989219572e-05 }, { "epoch": 0.5292876845692033, "step": 1631, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9551564037101343e-05, "weight_min": 0.0, "weight_std": 2.100959318340756e-05 }, { "epoch": 0.5296122018497486, "step": 1632, "weight_max": 3.1715208024252206e-05, "weight_mean": 7.622629254910862e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4128049770079087e-05 }, { "epoch": 0.5296122018497486, "step": 1632, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390952411398757e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.923927993630059e-05 }, { "epoch": 0.5296122018497486, "step": 1632, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0550482329563238e-05, "weight_min": 0.0, "weight_std": 1.9535451428964734e-05 }, { "epoch": 0.5296122018497486, "step": 1632, "weight_max": 3.632264997577295e-05, "weight_mean": 4.5409474296320695e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2841745956393424e-05 }, { "epoch": 0.5299367191302937, "step": 1633, "weight_max": 4.252122016623616e-05, "weight_mean": 1.789626548998058e-05, "weight_min": 0.0, "weight_std": 1.969602635654155e-05 }, { "epoch": 0.5299367191302937, "step": 1633, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5626963431714103e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1572228433797136e-05 }, { "epoch": 0.5299367191302937, "step": 1633, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3420532013697084e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8768081645248458e-05 }, { "epoch": 0.5299367191302937, "step": 1633, "weight_max": 3.36925259034615e-05, "weight_mean": 4.212227850075578e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1911839465028606e-05 }, { "epoch": 0.5302612364108389, "step": 1634, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396572760218987e-06, "weight_min": 0.0, "weight_std": 1.5261930457199924e-05 }, { "epoch": 0.5302612364108389, "step": 1634, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.7866355847218074e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.941591290233191e-05 }, { "epoch": 0.5302612364108389, "step": 1634, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0520321666263044e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9396293282625265e-05 }, { "epoch": 0.5302612364108389, "step": 1634, "weight_max": 3.813097282545641e-05, "weight_mean": 4.766943220602116e-06, "weight_min": 0.0, "weight_std": 1.3481104360835161e-05 }, { "epoch": 0.530585753691384, "step": 1635, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5194869774859399e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0979734472348355e-05 }, { "epoch": 0.530585753691384, "step": 1635, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.548465641273651e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7754195141606033e-05 }, { "epoch": 0.530585753691384, "step": 1635, "weight_max": 3.722132169059478e-05, "weight_mean": 1.0582491086097434e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.5221367902995553e-05 }, { "epoch": 0.530585753691384, "step": 1635, "weight_max": 4.252122016623616e-05, "weight_mean": 5.316059741744539e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5033154340926558e-05 }, { "epoch": 0.5309102709719292, "step": 1636, "weight_max": 4.187939703115262e-05, "weight_mean": 1.269501353817759e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8061575246974826e-05 }, { "epoch": 0.5309102709719292, "step": 1636, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236377091932809e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806015315116383e-05 }, { "epoch": 0.5309102709719292, "step": 1636, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0171412213821895e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.882803007902112e-05 }, { "epoch": 0.5309102709719292, "step": 1636, "weight_max": 3.752332122530788e-05, "weight_mean": 1.7128306353697553e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8435623132972978e-05 }, { "epoch": 0.5312347882524745, "step": 1637, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7746264347806573e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9265757146058604e-05 }, { "epoch": 0.5312347882524745, "step": 1637, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248238383297576e-06, "weight_min": 0.0, "weight_std": 1.2013399100396782e-05 }, { "epoch": 0.5312347882524745, "step": 1637, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.96158959827153e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.854130459832959e-05 }, { "epoch": 0.5312347882524745, "step": 1637, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0153782568522729e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8802751583280042e-05 }, { "epoch": 0.5315593055330197, "step": 1638, "weight_max": 3.813097282545641e-05, "weight_mean": 9.196072824124712e-06, "weight_min": 0.0, "weight_std": 1.7042239051079378e-05 }, { "epoch": 0.5315593055330197, "step": 1638, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.3701341231353581e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9791012164205313e-05 }, { "epoch": 0.5315593055330197, "step": 1638, "weight_max": 3.998322426923551e-05, "weight_mean": 6.696553100482561e-06, "weight_min": 0.0, "weight_std": 1.425967911927728e-05 }, { "epoch": 0.5315593055330197, "step": 1638, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.884202881949022e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.832331508921925e-05 }, { "epoch": 0.5318838228135648, "step": 1639, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3909277186030522e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.946296833921224e-05 }, { "epoch": 0.5318838228135648, "step": 1639, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0647790986695327e-05, "weight_min": 0.0, "weight_std": 1.967727075680159e-05 }, { "epoch": 0.5318838228135648, "step": 1639, "weight_max": 4.187939703115262e-05, "weight_mean": 9.85088263405487e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8286182239535265e-05 }, { "epoch": 0.5318838228135648, "step": 1639, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.8820807023730595e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3806573406327516e-05 }, { "epoch": 0.53220834009411, "grad_norm": 2.247953414916992, "learning_rate": 6.86530612244898e-06, "loss": 0.0722, "step": 1640 }, { "epoch": 0.53220834009411, "step": 1640, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315700946084689e-06, "weight_min": 0.0, "weight_std": 1.5033299860078841e-05 }, { "epoch": 0.53220834009411, "step": 1640, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5706964404671453e-05, "weight_min": 0.0, "weight_std": 2.1685020328732207e-05 }, { "epoch": 0.53220834009411, "step": 1640, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.0074616259080358e-05, "weight_min": 0.0, "weight_std": 2.1478890630532987e-05 }, { "epoch": 0.53220834009411, "step": 1640, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4913651284587104e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.930455255205743e-05 }, { "epoch": 0.5325328573746552, "step": 1641, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.455170877306955e-05, "weight_min": 0.0, "weight_std": 2.0179199054837227e-05 }, { "epoch": 0.5325328573746552, "step": 1641, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.686831617727876e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7935301002580673e-05 }, { "epoch": 0.5325328573746552, "step": 1641, "weight_max": 3.061215829802677e-05, "weight_mean": 3.8275520637398586e-06, "weight_min": 0.0, "weight_std": 1.0822614967764821e-05 }, { "epoch": 0.5325328573746552, "step": 1641, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504061053012265e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2736598364426754e-05 }, { "epoch": 0.5328573746552004, "step": 1642, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0581348760752007e-05, "weight_min": 0.0, "weight_std": 1.9155912013957277e-05 }, { "epoch": 0.5328573746552004, "step": 1642, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0919618691550568e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.193206819356419e-05 }, { "epoch": 0.5328573746552004, "step": 1642, "weight_max": 3.998322426923551e-05, "weight_mean": 4.998469194106292e-06, "weight_min": 0.0, "weight_std": 1.4135976016405039e-05 }, { "epoch": 0.5328573746552004, "step": 1642, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6492009308421984e-05, "weight_min": 0.0, "weight_std": 2.010053867707029e-05 }, { "epoch": 0.5331818919357456, "step": 1643, "weight_max": 3.255531191825867e-05, "weight_mean": 1.152119966718601e-05, "weight_min": 0.0, "weight_std": 1.5927114873193204e-05 }, { "epoch": 0.5331818919357456, "step": 1643, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.190353011945263e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7183274394483306e-05 }, { "epoch": 0.5331818919357456, "step": 1643, "weight_max": 3.9051607018336654e-05, "weight_mean": 7.435938641719986e-06, "weight_min": 0.0, "weight_std": 1.463912758481456e-05 }, { "epoch": 0.5331818919357456, "step": 1643, "weight_max": 3.5141503758495674e-05, "weight_mean": 6.675637450825889e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.315741701546358e-05 }, { "epoch": 0.5335064092162908, "step": 1644, "weight_max": 1.344265809777312e-09, "weight_mean": 6.653580464366371e-10, "weight_min": 0.0, "weight_std": 3.627740885470132e-10 }, { "epoch": 0.5335064092162908, "step": 1644, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235739990894217e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4806273611611687e-05 }, { "epoch": 0.5335064092162908, "step": 1644, "weight_max": 3.602553260861896e-05, "weight_mean": 5.368787242332473e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2621594578376971e-05 }, { "epoch": 0.5335064092162908, "step": 1644, "weight_max": 4.1242459701607004e-05, "weight_mean": 2.096898060699459e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9294891899335198e-05 }, { "epoch": 0.533830926496836, "step": 1645, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0329487597336993e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.153713103325572e-05 }, { "epoch": 0.533830926496836, "step": 1645, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.999137546401471e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.8524646293371916e-05 }, { "epoch": 0.533830926496836, "step": 1645, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8086204363498837e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9640434402390383e-05 }, { "epoch": 0.533830926496836, "step": 1645, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.965153609897243e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1212716344743967e-05 }, { "epoch": 0.5341554437773811, "step": 1646, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4263299817685038e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9856266590068117e-05 }, { "epoch": 0.5341554437773811, "step": 1646, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396700544224586e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.526187770650722e-05 }, { "epoch": 0.5341554437773811, "step": 1646, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0124924301635474e-05, "weight_min": 0.0, "weight_std": 1.8800468751578592e-05 }, { "epoch": 0.5341554437773811, "step": 1646, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827149612334324e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0822777767316438e-05 }, { "epoch": 0.5344799610579264, "step": 1647, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.886171154794283e-06, "weight_min": 0.0, "weight_std": 1.3804926311422605e-05 }, { "epoch": 0.5344799610579264, "step": 1647, "weight_max": 3.843663216684945e-05, "weight_mean": 8.4626444731839e-06, "weight_min": 0.0, "weight_std": 1.585937206982635e-05 }, { "epoch": 0.5344799610579264, "step": 1647, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882279426965397e-06, "weight_min": 0.0, "weight_std": 1.380649337079376e-05 }, { "epoch": 0.5344799610579264, "step": 1647, "weight_max": 1.8684177121031098e-05, "weight_mean": 2.336483248654986e-06, "weight_min": 6.053385570581327e-10, "weight_std": 6.6054662966053e-06 }, { "epoch": 0.5348044783384716, "step": 1648, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.1862528481287882e-05, "weight_min": 0.0, "weight_std": 1.6947851690929383e-05 }, { "epoch": 0.5348044783384716, "step": 1648, "weight_max": 3.752332122530788e-05, "weight_mean": 4.737501967611024e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3248100003693253e-05 }, { "epoch": 0.5348044783384716, "step": 1648, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.551332939532585e-06, "weight_min": 0.0, "weight_std": 1.6273184883175418e-05 }, { "epoch": 0.5348044783384716, "step": 1648, "weight_max": 3.602553260861896e-05, "weight_mean": 4.504615844780346e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2736373719235417e-05 }, { "epoch": 0.5351289956190167, "step": 1649, "weight_max": 3.998322426923551e-05, "weight_mean": 1.401705503667472e-05, "weight_min": 0.0, "weight_std": 1.9434439309407026e-05 }, { "epoch": 0.5351289956190167, "step": 1649, "weight_max": 4.187939703115262e-05, "weight_mean": 9.925970516633242e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8414608348393813e-05 }, { "epoch": 0.5351289956190167, "step": 1649, "weight_max": 4.187939703115262e-05, "weight_mean": 5.257845259620808e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4797473340877332e-05 }, { "epoch": 0.5351289956190167, "step": 1649, "weight_max": 3.813097282545641e-05, "weight_mean": 7.293191629287321e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.4327007193060126e-05 }, { "epoch": 0.5354535128995619, "grad_norm": 0.06970520317554474, "learning_rate": 6.844897959183674e-06, "loss": 0.1486, "step": 1650 }, { "epoch": 0.5354535128995619, "step": 1650, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4216092495189514e-05, "weight_min": 0.0, "weight_std": 1.9708324543898925e-05 }, { "epoch": 0.5354535128995619, "step": 1650, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4241384633351117e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9888533643097617e-05 }, { "epoch": 0.5354535128995619, "step": 1650, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077125933894422e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4357612599269487e-05 }, { "epoch": 0.5354535128995619, "step": 1650, "weight_max": 2.820166446326766e-05, "weight_mean": 3.5268681131128687e-06, "weight_min": 5.679522963042416e-10, "weight_std": 9.970123755920213e-06 }, { "epoch": 0.5357780301801071, "step": 1651, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2885537216789089e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.793309456843417e-05 }, { "epoch": 0.5357780301801071, "step": 1651, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4861541785649024e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0560462871799245e-05 }, { "epoch": 0.5357780301801071, "step": 1651, "weight_max": 2.4214691620727535e-06, "weight_mean": 3.032463098406879e-07, "weight_min": 0.0, "weight_std": 8.558913577871863e-07 }, { "epoch": 0.5357780301801071, "step": 1651, "weight_max": 3.967146767536178e-05, "weight_mean": 4.960077149007702e-06, "weight_min": 0.0, "weight_std": 1.402551970386412e-05 }, { "epoch": 0.5361025474606523, "step": 1652, "weight_max": 4.187939703115262e-05, "weight_mean": 9.927530300046783e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.8413646102999337e-05 }, { "epoch": 0.5361025474606523, "step": 1652, "weight_max": 4.187939703115262e-05, "weight_mean": 1.913588130264543e-05, "weight_min": 0.0, "weight_std": 2.0734360077767633e-05 }, { "epoch": 0.5361025474606523, "step": 1652, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3888377907278482e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9356228222022764e-05 }, { "epoch": 0.5361025474606523, "step": 1652, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5122730474104173e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.091956775984727e-05 }, { "epoch": 0.5364270647411975, "step": 1653, "weight_max": 4.252122016623616e-05, "weight_mean": 9.930888154485729e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.844761209213175e-05 }, { "epoch": 0.5364270647411975, "step": 1653, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4618516615882982e-05, "weight_min": 0.0, "weight_std": 2.0229645087965764e-05 }, { "epoch": 0.5364270647411975, "step": 1653, "weight_max": 3.28377827827353e-05, "weight_mean": 5.211074039834784e-06, "weight_min": 0.0, "weight_std": 1.1450013516878244e-05 }, { "epoch": 0.5364270647411975, "step": 1653, "weight_max": 4.187939703115262e-05, "weight_mean": 1.031264582707081e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9094813978881575e-05 }, { "epoch": 0.5367515820217427, "step": 1654, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.725376003189012e-06, "weight_min": 0.0, "weight_std": 1.8007909602602012e-05 }, { "epoch": 0.5367515820217427, "step": 1654, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2013527339149732e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.712676021270454e-05 }, { "epoch": 0.5367515820217427, "step": 1654, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5077537682373077e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0829906134167686e-05 }, { "epoch": 0.5367515820217427, "step": 1654, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.0778057912830263e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.641539711272344e-05 }, { "epoch": 0.5370760993022878, "step": 1655, "weight_max": 4.252122016623616e-05, "weight_mean": 5.847289685334545e-06, "weight_min": 0.0, "weight_std": 1.489305759605486e-05 }, { "epoch": 0.5370760993022878, "step": 1655, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4308622667158488e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.630473525437992e-05 }, { "epoch": 0.5370760993022878, "step": 1655, "weight_max": 3.843663216684945e-05, "weight_mean": 7.58717033022549e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4698021004733164e-05 }, { "epoch": 0.5370760993022878, "step": 1655, "weight_max": 3.998322426923551e-05, "weight_mean": 9.727490578370634e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.801789585442748e-05 }, { "epoch": 0.537400616582833, "step": 1656, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0243249562336132e-05, "weight_min": 0.0, "weight_std": 1.881664320535492e-05 }, { "epoch": 0.537400616582833, "step": 1656, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.3933770787843969e-05, "weight_min": 0.0, "weight_std": 1.928669007611461e-05 }, { "epoch": 0.537400616582833, "step": 1656, "weight_max": 4.252122016623616e-05, "weight_mean": 9.818785656534601e-06, "weight_min": 0.0, "weight_std": 1.826246989367064e-05 }, { "epoch": 0.537400616582833, "step": 1656, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.677422556502279e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6063992006820627e-05 }, { "epoch": 0.5377251338633783, "step": 1657, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1048392480006441e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.863133547885809e-05 }, { "epoch": 0.5377251338633783, "step": 1657, "weight_max": 3.843663216684945e-05, "weight_mean": 9.571443115419243e-06, "weight_min": 0.0, "weight_std": 1.7721842596074566e-05 }, { "epoch": 0.5377251338633783, "step": 1657, "weight_max": 4.187939703115262e-05, "weight_mean": 2.3896780476206914e-05, "weight_min": 0.0, "weight_std": 1.9903456632164307e-05 }, { "epoch": 0.5377251338633783, "step": 1657, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4064141396374907e-05, "weight_min": 0.0, "weight_std": 1.9683351638377644e-05 }, { "epoch": 0.5380496511439234, "step": 1658, "weight_max": 2.9469142646121327e-06, "weight_mean": 3.6937214531462814e-07, "weight_min": 0.0, "weight_std": 1.0414854614282376e-06 }, { "epoch": 0.5380496511439234, "step": 1658, "weight_max": 1.076157385426768e-09, "weight_mean": 7.525595702162491e-10, "weight_min": 5.864964625068581e-10, "weight_std": 2.0588569737967077e-10 }, { "epoch": 0.5380496511439234, "step": 1658, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.214757594425464e-05, "weight_min": 0.0, "weight_std": 1.9212959159631282e-05 }, { "epoch": 0.5380496511439234, "step": 1658, "weight_max": 3.843663216684945e-05, "weight_mean": 1.3833090633852407e-05, "weight_min": 0.0, "weight_std": 1.9058223188039847e-05 }, { "epoch": 0.5383741684244686, "step": 1659, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5946718121995218e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.2009819076629356e-05 }, { "epoch": 0.5383741684244686, "step": 1659, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.499647169111995e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0703058908111416e-05 }, { "epoch": 0.5383741684244686, "step": 1659, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3531718397862278e-05, "weight_min": 6.686325937366178e-10, "weight_std": 1.8175787772634067e-05 }, { "epoch": 0.5383741684244686, "step": 1659, "weight_max": 1.3956888267330214e-07, "weight_mean": 1.801853422023214e-08, "weight_min": 0.0, "weight_std": 4.9114696309970896e-08 }, { "epoch": 0.5386986857050138, "grad_norm": 0.283355712890625, "learning_rate": 6.8244897959183675e-06, "loss": 0.0538, "step": 1660 }, { "epoch": 0.5386986857050138, "step": 1660, "weight_max": 3.397988257347606e-05, "weight_mean": 4.249205630912911e-06, "weight_min": 0.0, "weight_std": 1.201300801767502e-05 }, { "epoch": 0.5386986857050138, "step": 1660, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.087600114464294e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0872697859886102e-05 }, { "epoch": 0.5386986857050138, "step": 1660, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0695870514609851e-05, "weight_min": 6.586902134841921e-10, "weight_std": 1.617423549760133e-05 }, { "epoch": 0.5386986857050138, "step": 1660, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.8111149984179065e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9456039808574133e-05 }, { "epoch": 0.539023202985559, "step": 1661, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.790615877136588e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.825042454584036e-05 }, { "epoch": 0.539023202985559, "step": 1661, "weight_max": 4.187939703115262e-05, "weight_mean": 1.043423981172964e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9212935512769036e-05 }, { "epoch": 0.539023202985559, "step": 1661, "weight_max": 4.187939703115262e-05, "weight_mean": 1.968674587260466e-05, "weight_min": 0.0, "weight_std": 2.1224856027401984e-05 }, { "epoch": 0.539023202985559, "step": 1661, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2511574823292904e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.795012940419838e-05 }, { "epoch": 0.5393477202661042, "step": 1662, "weight_max": 1.384178176522255e-07, "weight_mean": 1.8094759468567645e-08, "weight_min": 5.864964625068581e-10, "weight_std": 4.862014080231347e-08 }, { "epoch": 0.5393477202661042, "step": 1662, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.8925313270301558e-05, "weight_min": 6.586902134841921e-10, "weight_std": 2.044416578428354e-05 }, { "epoch": 0.5393477202661042, "step": 1662, "weight_max": 3.967146767536178e-05, "weight_mean": 5.264851097308565e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3928558473708108e-05 }, { "epoch": 0.5393477202661042, "step": 1662, "weight_max": 4.2156173662988294e-08, "weight_mean": 7.1357484365819346e-09, "weight_min": 0.0, "weight_std": 1.43613840819512e-08 }, { "epoch": 0.5396722375466494, "step": 1663, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4735151125933044e-05, "weight_min": 0.0, "weight_std": 2.0394529201439582e-05 }, { "epoch": 0.5396722375466494, "step": 1663, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3006461813347414e-05, "weight_min": 0.0, "weight_std": 1.815725590859074e-05 }, { "epoch": 0.5396722375466494, "step": 1663, "weight_max": 1.868417864869798e-09, "weight_mean": 8.819219798894551e-10, "weight_min": 5.864964625068581e-10, "weight_std": 4.74263295302535e-10 }, { "epoch": 0.5396722375466494, "step": 1663, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315743237588322e-06, "weight_min": 0.0, "weight_std": 1.5033282579679508e-05 }, { "epoch": 0.5399967548271946, "step": 1664, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9429828171269037e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0853896785411052e-05 }, { "epoch": 0.5399967548271946, "step": 1664, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.612169040134177e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6159325241460465e-05 }, { "epoch": 0.5399967548271946, "step": 1664, "weight_max": 2.8994179956498556e-05, "weight_mean": 8.005570634850301e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2251836778887082e-05 }, { "epoch": 0.5399967548271946, "step": 1664, "weight_max": 4.3167921830900013e-05, "weight_mean": 8.063402674451936e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6028901882236823e-05 }, { "epoch": 0.5403212721077397, "step": 1665, "weight_max": 3.340639523230493e-05, "weight_mean": 4.17643741457141e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1810686373792123e-05 }, { "epoch": 0.5403212721077397, "step": 1665, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.773225999902934e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.636034903640393e-05 }, { "epoch": 0.5403212721077397, "step": 1665, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4401945009012707e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9981787772849202e-05 }, { "epoch": 0.5403212721077397, "step": 1665, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691210506280186e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.326617621089099e-05 }, { "epoch": 0.5406457893882849, "step": 1666, "weight_max": 3.998322426923551e-05, "weight_mean": 9.282527571485844e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7253005353268236e-05 }, { "epoch": 0.5406457893882849, "step": 1666, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.396616415964672e-06, "weight_min": 0.0, "weight_std": 1.5261912267305888e-05 }, { "epoch": 0.5406457893882849, "step": 1666, "weight_max": 4.252122016623616e-05, "weight_mean": 9.74504291662015e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.81421528395731e-05 }, { "epoch": 0.5406457893882849, "step": 1666, "weight_max": 3.4558255720185116e-05, "weight_mean": 4.321864253142849e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2217348739795852e-05 }, { "epoch": 0.5409703066688302, "step": 1667, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.385463272337802e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7395286704413593e-05 }, { "epoch": 0.5409703066688302, "step": 1667, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.476497436669888e-05, "weight_min": 0.0, "weight_std": 2.039585160673596e-05 }, { "epoch": 0.5409703066688302, "step": 1667, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471274435985833e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9390034140087664e-05 }, { "epoch": 0.5409703066688302, "step": 1667, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4412700693355873e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.00507329282118e-05 }, { "epoch": 0.5412948239493753, "step": 1668, "weight_max": 3.752332122530788e-05, "weight_mean": 1.6215133655350655e-05, "weight_min": 0.0, "weight_std": 1.7520094843348488e-05 }, { "epoch": 0.5412948239493753, "step": 1668, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.217797934164992e-06, "weight_min": 0.0, "weight_std": 1.4306213415693492e-05 }, { "epoch": 0.5412948239493753, "step": 1668, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156010956852697e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.4581128198187798e-05 }, { "epoch": 0.5412948239493753, "step": 1668, "weight_max": 4.252122016623616e-05, "weight_mean": 1.791516478988342e-05, "weight_min": 7.403143098549947e-10, "weight_std": 1.937515480676666e-05 }, { "epoch": 0.5416193412299205, "step": 1669, "weight_max": 3.602553260861896e-05, "weight_mean": 1.3109998690197244e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8112747056875378e-05 }, { "epoch": 0.5416193412299205, "step": 1669, "weight_max": 2.926079287135508e-05, "weight_mean": 7.261806786118541e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.3414688510238193e-05 }, { "epoch": 0.5416193412299205, "step": 1669, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.304159923165571e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8345259377383627e-05 }, { "epoch": 0.5416193412299205, "step": 1669, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0153358743991703e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8798496967065148e-05 }, { "epoch": 0.5419438585104657, "grad_norm": 0.2778269350528717, "learning_rate": 6.8040816326530625e-06, "loss": 0.1612, "step": 1670 }, { "epoch": 0.5419438585104657, "step": 1670, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3557680176745635e-05, "weight_min": 7.194614903838215e-10, "weight_std": 1.8912354789790697e-05 }, { "epoch": 0.5419438585104657, "step": 1670, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.078751655673841e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4356956853589509e-05 }, { "epoch": 0.5419438585104657, "step": 1670, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0518915334832855e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.200148446718231e-05 }, { "epoch": 0.5419438585104657, "step": 1670, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4694086530653294e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.0324527213233523e-05 }, { "epoch": 0.5422683757910108, "step": 1671, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.1100761184934527e-05, "weight_min": 0.0, "weight_std": 1.6227902960963547e-05 }, { "epoch": 0.5422683757910108, "step": 1671, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4061964975553565e-05, "weight_min": 0.0, "weight_std": 1.9829369193757884e-05 }, { "epoch": 0.5422683757910108, "step": 1671, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4853623724775389e-05, "weight_min": 0.0, "weight_std": 2.056776065728627e-05 }, { "epoch": 0.5422683757910108, "step": 1671, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.159443854703568e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.7497686712886207e-05 }, { "epoch": 0.5425928930715561, "step": 1672, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.430936188437045e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.770207381923683e-05 }, { "epoch": 0.5425928930715561, "step": 1672, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4244342310121283e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.969725599337835e-05 }, { "epoch": 0.5425928930715561, "step": 1672, "weight_max": 1.4598501296703148e-09, "weight_mean": 8.22706458514233e-10, "weight_min": 0.0, "weight_std": 4.320373503396979e-10 }, { "epoch": 0.5425928930715561, "step": 1672, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9372399037820287e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9259441614849493e-05 }, { "epoch": 0.5429174103521013, "step": 1673, "weight_max": 1.5345109627418196e-09, "weight_mean": 7.537360735554444e-10, "weight_min": 0.0, "weight_std": 4.3269765548359373e-10 }, { "epoch": 0.5429174103521013, "step": 1673, "weight_max": 3.998322426923551e-05, "weight_mean": 5.004469585401239e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4133559488982428e-05 }, { "epoch": 0.5429174103521013, "step": 1673, "weight_max": 4.0610404539620504e-05, "weight_mean": 6.3458119257120416e-06, "weight_min": 0.0, "weight_std": 1.4293483218352776e-05 }, { "epoch": 0.5429174103521013, "step": 1673, "weight_max": 3.5141503758495674e-05, "weight_mean": 5.157433861313621e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2301881724852137e-05 }, { "epoch": 0.5432419276326464, "step": 1674, "weight_max": 9.087236740867866e-09, "weight_mean": 2.045156488605926e-09, "weight_min": 5.771870759119224e-10, "weight_std": 2.9162501302693045e-09 }, { "epoch": 0.5432419276326464, "step": 1674, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0548093743855134e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1990595996612683e-05 }, { "epoch": 0.5432419276326464, "step": 1674, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8821250705514103e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0494708223850466e-05 }, { "epoch": 0.5432419276326464, "step": 1674, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4329213627206627e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.9907185560441576e-05 }, { "epoch": 0.5435664449131916, "step": 1675, "weight_max": 2.872879122151062e-05, "weight_mean": 3.591897439036984e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0156839380215388e-05 }, { "epoch": 0.5435664449131916, "step": 1675, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5007725778559688e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.989147494896315e-05 }, { "epoch": 0.5435664449131916, "step": 1675, "weight_max": 3.998322426923551e-05, "weight_mean": 9.575989679433405e-06, "weight_min": 0.0, "weight_std": 1.7752900021150708e-05 }, { "epoch": 0.5435664449131916, "step": 1675, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.569360943511128e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.587199221830815e-05 }, { "epoch": 0.5438909621937368, "step": 1676, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.292003596783616e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.750004776113201e-05 }, { "epoch": 0.5438909621937368, "step": 1676, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0783610276994295e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4357113286678214e-05 }, { "epoch": 0.5438909621937368, "step": 1676, "weight_max": 9.062063877074422e-10, "weight_mean": 6.679227171346724e-10, "weight_min": 5.771870759119224e-10, "weight_std": 1.0155662422128842e-10 }, { "epoch": 0.5438909621937368, "step": 1676, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235670869296882e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.480630089645274e-05 }, { "epoch": 0.5442154794742821, "step": 1677, "weight_max": 3.662098970380612e-05, "weight_mean": 4.5784295252815355e-06, "weight_min": 0.0, "weight_std": 1.2947149116371293e-05 }, { "epoch": 0.5442154794742821, "step": 1677, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4182128325046506e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9723005607374944e-05 }, { "epoch": 0.5442154794742821, "step": 1677, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3897863027523272e-05, "weight_min": 0.0, "weight_std": 1.9400715245865285e-05 }, { "epoch": 0.5442154794742821, "step": 1677, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4478648154181428e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0097913875360973e-05 }, { "epoch": 0.5445399967548272, "step": 1678, "weight_max": 1.2334475663067224e-09, "weight_mean": 6.96131152722046e-10, "weight_min": 5.771870759119224e-10, "weight_std": 2.1934212779406437e-10 }, { "epoch": 0.5445399967548272, "step": 1678, "weight_max": 2.690521432668902e-05, "weight_mean": 3.3637254546192707e-06, "weight_min": 0.0, "weight_std": 9.512197721051052e-06 }, { "epoch": 0.5445399967548272, "step": 1678, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.998975656344555e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.852474633778911e-05 }, { "epoch": 0.5445399967548272, "step": 1678, "weight_max": 3.28377827827353e-05, "weight_mean": 6.6313232309767045e-06, "weight_min": 0.0, "weight_std": 1.2733084986393806e-05 }, { "epoch": 0.5448645140353724, "step": 1679, "weight_max": 4.252122016623616e-05, "weight_mean": 1.722958404570818e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.921566581586376e-05 }, { "epoch": 0.5448645140353724, "step": 1679, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0472864232724532e-05, "weight_min": 0.0, "weight_std": 1.940260517585557e-05 }, { "epoch": 0.5448645140353724, "step": 1679, "weight_max": 3.967146767536178e-05, "weight_mean": 9.242933629138861e-06, "weight_min": 0.0, "weight_std": 1.7174308595713228e-05 }, { "epoch": 0.5448645140353724, "step": 1679, "weight_max": 4.252122016623616e-05, "weight_mean": 1.135549337050179e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8130254829884507e-05 }, { "epoch": 0.5451890313159176, "grad_norm": 0.5280314683914185, "learning_rate": 6.783673469387756e-06, "loss": 0.1584, "step": 1680 }, { "epoch": 0.5451890313159176, "step": 1680, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.5244662790792063e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.079161822621245e-05 }, { "epoch": 0.5451890313159176, "step": 1680, "weight_max": 4.252122016623616e-05, "weight_mean": 2.2878517484059557e-05, "weight_min": 0.0, "weight_std": 1.9226985386922024e-05 }, { "epoch": 0.5451890313159176, "step": 1680, "weight_max": 3.397988257347606e-05, "weight_mean": 7.644210199941881e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.4267539881984703e-05 }, { "epoch": 0.5451890313159176, "step": 1680, "weight_max": 3.255531191825867e-05, "weight_mean": 4.0705108403926715e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.1509598152770195e-05 }, { "epoch": 0.5455135485964627, "step": 1681, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5869758499320596e-05, "weight_min": 0.0, "weight_std": 2.0881872842437588e-05 }, { "epoch": 0.5455135485964627, "step": 1681, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.4695415100432e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.759363840392325e-05 }, { "epoch": 0.5455135485964627, "step": 1681, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843579517910257e-06, "weight_min": 0.0, "weight_std": 1.3697639587917365e-05 }, { "epoch": 0.5455135485964627, "step": 1681, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.7628017303650267e-05, "weight_min": 0.0, "weight_std": 1.965421142813284e-05 }, { "epoch": 0.545838065877008, "step": 1682, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.39676829148084e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.7472770196036436e-05 }, { "epoch": 0.545838065877008, "step": 1682, "weight_max": 3.602553260861896e-05, "weight_mean": 8.896417057258077e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6473331925226375e-05 }, { "epoch": 0.545838065877008, "step": 1682, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7427581042284146e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.893998523883056e-05 }, { "epoch": 0.545838065877008, "step": 1682, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077262812847039e-06, "weight_min": 0.0, "weight_std": 1.4357557120092679e-05 }, { "epoch": 0.5461625831575532, "step": 1683, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1949050531256944e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.7830381693784148e-05 }, { "epoch": 0.5461625831575532, "step": 1683, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.431971774465637e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.252710080734687e-05 }, { "epoch": 0.5461625831575532, "step": 1683, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.2324429917498492e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9219338355469517e-05 }, { "epoch": 0.5461625831575532, "step": 1683, "weight_max": 3.843663216684945e-05, "weight_mean": 8.576421350880992e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.600458381290082e-05 }, { "epoch": 0.5464871004380983, "step": 1684, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0314073733752593e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9107483240077272e-05 }, { "epoch": 0.5464871004380983, "step": 1684, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9547809642972425e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0929126549162902e-05 }, { "epoch": 0.5464871004380983, "step": 1684, "weight_max": 4.381950566312298e-05, "weight_mean": 1.5212878679449204e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1080571968923323e-05 }, { "epoch": 0.5464871004380983, "step": 1684, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.016836884256918e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.692193836788647e-05 }, { "epoch": 0.5468116177186435, "step": 1685, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.001236705633346e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.8611119230627082e-05 }, { "epoch": 0.5468116177186435, "step": 1685, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.363146657240577e-05, "weight_min": 0.0, "weight_std": 1.9016844817087986e-05 }, { "epoch": 0.5468116177186435, "step": 1685, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0349540136521682e-05, "weight_min": 0.0, "weight_std": 2.1773714252049103e-05 }, { "epoch": 0.5468116177186435, "step": 1685, "weight_max": 9.230136129190214e-06, "weight_mean": 1.1543452274054289e-06, "weight_min": 5.771870759119224e-10, "weight_std": 3.2631126032356406e-06 }, { "epoch": 0.5471361349991887, "step": 1686, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0631273653416429e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.9682902347994968e-05 }, { "epoch": 0.5471361349991887, "step": 1686, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4285415090853348e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9975104805780575e-05 }, { "epoch": 0.5471361349991887, "step": 1686, "weight_max": 3.662098970380612e-05, "weight_mean": 1.2472168236854486e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.728267307044007e-05 }, { "epoch": 0.5471361349991887, "step": 1686, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236048309598118e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4806149010837544e-05 }, { "epoch": 0.547460652279734, "step": 1687, "weight_max": 3.843663216684945e-05, "weight_mean": 4.80529160995502e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.358911413262831e-05 }, { "epoch": 0.547460652279734, "step": 1687, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2201729987282306e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.752109164954163e-05 }, { "epoch": 0.547460652279734, "step": 1687, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471051609783899e-05, "weight_min": 0.0, "weight_std": 1.9390172383282334e-05 }, { "epoch": 0.547460652279734, "step": 1687, "weight_max": 3.998322426923551e-05, "weight_mean": 1.4569568520528264e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.011532887991052e-05 }, { "epoch": 0.5477851695602791, "step": 1688, "weight_max": 3.0339444492710754e-05, "weight_mean": 3.794683379965136e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0725703759817407e-05 }, { "epoch": 0.5477851695602791, "step": 1688, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4284589269664139e-05, "weight_min": 0.0, "weight_std": 1.9814153347397223e-05 }, { "epoch": 0.5477851695602791, "step": 1688, "weight_max": 4.252122016623616e-05, "weight_mean": 1.8635801097843796e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.0183502783766016e-05 }, { "epoch": 0.5477851695602791, "step": 1688, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0018487955676392e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8656761312740855e-05 }, { "epoch": 0.5481096868408243, "step": 1689, "weight_max": 4.252122016623616e-05, "weight_mean": 9.895104085444473e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8384180293651298e-05 }, { "epoch": 0.5481096868408243, "step": 1689, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105560037714895e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.1609571629378479e-05 }, { "epoch": 0.5481096868408243, "step": 1689, "weight_max": 3.722132169059478e-05, "weight_mean": 1.298500683333259e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.799643723643385e-05 }, { "epoch": 0.5481096868408243, "step": 1689, "weight_max": 4.252122016623616e-05, "weight_mean": 9.527142537990585e-06, "weight_min": 0.0, "weight_std": 1.779690501280129e-05 }, { "epoch": 0.5484342041213695, "grad_norm": 0.2536866068840027, "learning_rate": 6.76326530612245e-06, "loss": 0.0975, "step": 1690 }, { "epoch": 0.5484342041213695, "step": 1690, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.144766040146351e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.9503831936162896e-05 }, { "epoch": 0.5484342041213695, "step": 1690, "weight_max": 4.666427244615079e-09, "weight_mean": 1.15360387908936e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.4208491050382577e-09 }, { "epoch": 0.5484342041213695, "step": 1690, "weight_max": 3.4849264920921996e-05, "weight_mean": 8.349625204573385e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.541172241559252e-05 }, { "epoch": 0.5484342041213695, "step": 1690, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6026751836761832e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.2118911147117615e-05 }, { "epoch": 0.5487587214019146, "step": 1691, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.491117629688233e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0790226699318737e-05 }, { "epoch": 0.5487587214019146, "step": 1691, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0471098903508391e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9390143279451877e-05 }, { "epoch": 0.5487587214019146, "step": 1691, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3920922356192023e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.963896647794172e-05 }, { "epoch": 0.5487587214019146, "step": 1691, "weight_max": 6.2595857386327225e-09, "weight_mean": 1.4319245789096158e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.9737358414317896e-09 }, { "epoch": 0.5490832386824599, "step": 1692, "weight_max": 3.998322426923551e-05, "weight_mean": 9.879757271846756e-06, "weight_min": 0.0, "weight_std": 1.829445500334259e-05 }, { "epoch": 0.5490832386824599, "step": 1692, "weight_max": 4.772478412462533e-09, "weight_mean": 1.1834996316295587e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.451484044068252e-09 }, { "epoch": 0.5490832386824599, "step": 1692, "weight_max": 1.4469024790741969e-05, "weight_mean": 1.8098334066962707e-06, "weight_min": 6.488223291967188e-10, "weight_std": 5.115085514262319e-06 }, { "epoch": 0.5490832386824599, "step": 1692, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4921963156666607e-05, "weight_min": 7.194614903838215e-10, "weight_std": 2.0615414541680366e-05 }, { "epoch": 0.5494077559630051, "step": 1693, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2922284188098274e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.928896745084785e-05 }, { "epoch": 0.5494077559630051, "step": 1693, "weight_max": 4.187939703115262e-05, "weight_mean": 9.96364360617008e-06, "weight_min": 0.0, "weight_std": 1.8479860955267213e-05 }, { "epoch": 0.5494077559630051, "step": 1693, "weight_max": 9.893572627817093e-10, "weight_mean": 6.966594523483138e-10, "weight_min": 5.864964625068581e-10, "weight_std": 1.3017500966050477e-10 }, { "epoch": 0.5494077559630051, "step": 1693, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3335305993678048e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8446224203216843e-05 }, { "epoch": 0.5497322732435502, "step": 1694, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0712604307627771e-05, "weight_min": 0.0, "weight_std": 1.9833032638416626e-05 }, { "epoch": 0.5497322732435502, "step": 1694, "weight_max": 4.252122016623616e-05, "weight_mean": 2.1702355297748e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.072999814117793e-05 }, { "epoch": 0.5497322732435502, "step": 1694, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23545440955786e-06, "weight_min": 0.0, "weight_std": 1.4806389117438812e-05 }, { "epoch": 0.5497322732435502, "step": 1694, "weight_max": 4.381950566312298e-05, "weight_mean": 1.4943374480935745e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0897816284559667e-05 }, { "epoch": 0.5500567905240954, "step": 1695, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.329367119586095e-06, "weight_min": 0.0, "weight_std": 1.5993537090253085e-05 }, { "epoch": 0.5500567905240954, "step": 1695, "weight_max": 4.187939703115262e-05, "weight_mean": 1.1827807611552998e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.818453529267572e-05 }, { "epoch": 0.5500567905240954, "step": 1695, "weight_max": 3.9051607018336654e-05, "weight_mean": 1.6700898413546383e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.5150631952565163e-05 }, { "epoch": 0.5500567905240954, "step": 1695, "weight_max": 3.4849264920921996e-05, "weight_mean": 1.2183960279799066e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.6846726794028655e-05 }, { "epoch": 0.5503813078046406, "step": 1696, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.467849511362147e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.589288149261847e-05 }, { "epoch": 0.5503813078046406, "step": 1696, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0074850251839962e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8654154700925574e-05 }, { "epoch": 0.5503813078046406, "step": 1696, "weight_max": 3.486749999837002e-09, "weight_mean": 1.0574456865697357e-09, "weight_min": 5.679522963042416e-10, "weight_std": 1.0040953624113058e-09 }, { "epoch": 0.5503813078046406, "step": 1696, "weight_max": 3.397988257347606e-05, "weight_mean": 8.807643098407425e-06, "weight_min": 0.0, "weight_std": 1.5560541214654222e-05 }, { "epoch": 0.5507058250851858, "step": 1697, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0077170372824185e-05, "weight_min": 0.0, "weight_std": 1.8192506104242057e-05 }, { "epoch": 0.5507058250851858, "step": 1697, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.71393785928376e-06, "weight_min": 0.0, "weight_std": 1.6491179849253967e-05 }, { "epoch": 0.5507058250851858, "step": 1697, "weight_max": 4.381950566312298e-05, "weight_mean": 2.006000795518048e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.156374102924019e-05 }, { "epoch": 0.5507058250851858, "step": 1697, "weight_max": 4.187939703115262e-05, "weight_mean": 1.538911965326406e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.1243093215161934e-05 }, { "epoch": 0.551030342365731, "step": 1698, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.843608166993363e-06, "weight_min": 0.0, "weight_std": 1.3697627764486242e-05 }, { "epoch": 0.551030342365731, "step": 1698, "weight_max": 4.0610404539620504e-05, "weight_mean": 2.2684642317472026e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.90230075531872e-05 }, { "epoch": 0.551030342365731, "step": 1698, "weight_max": 3.998322426923551e-05, "weight_mean": 2.4789100280031562e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.648701618250925e-05 }, { "epoch": 0.551030342365731, "step": 1698, "weight_max": 1.1429997215373078e-07, "weight_mean": 1.8566195691960274e-08, "weight_min": 0.0, "weight_std": 3.9464165269009754e-08 }, { "epoch": 0.5513548596462762, "step": 1699, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.653274901211262e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.773576695995871e-05 }, { "epoch": 0.5513548596462762, "step": 1699, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076936758996453e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4357688996824436e-05 }, { "epoch": 0.5513548596462762, "step": 1699, "weight_max": 3.5141503758495674e-05, "weight_mean": 8.786093530943617e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6266911188722588e-05 }, { "epoch": 0.5513548596462762, "step": 1699, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.1967598766204901e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.7581394786247984e-05 }, { "epoch": 0.5516793769268213, "grad_norm": 0.1294732391834259, "learning_rate": 6.742857142857143e-06, "loss": 0.1152, "step": 1700 }, { "epoch": 0.5516793769268213, "eval_accuracy": 0.98, "eval_auc": 0.9959679193803908, "eval_loss": 0.09139815717935562, "eval_runtime": 156.0216, "eval_samples_per_second": 12.819, "eval_steps_per_second": 1.602, "step": 1700 }, { "epoch": 0.5516793769268213, "step": 1700, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3612176189781167e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.899070048239082e-05 }, { "epoch": 0.5516793769268213, "step": 1700, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3889841284253635e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9504277588566765e-05 }, { "epoch": 0.5516793769268213, "step": 1700, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5546031136182137e-05, "weight_min": 0.0, "weight_std": 2.145537655451335e-05 }, { "epoch": 0.5516793769268213, "step": 1700, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9589890143834054e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.119451528415084e-05 }, { "epoch": 0.5520038942073665, "step": 1701, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0729803761932999e-05, "weight_min": 0.0, "weight_std": 1.871568565547932e-05 }, { "epoch": 0.5520038942073665, "step": 1701, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827531145361718e-06, "weight_min": 0.0, "weight_std": 1.0822623153217137e-05 }, { "epoch": 0.5520038942073665, "step": 1701, "weight_max": 1.0469849257788155e-05, "weight_mean": 1.3093180086798384e-06, "weight_min": 5.958802340444436e-10, "weight_std": 3.701413561429945e-06 }, { "epoch": 0.5520038942073665, "step": 1701, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.948355571192224e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.682191577856429e-05 }, { "epoch": 0.5523284114879118, "step": 1702, "weight_max": 4.252122016623616e-05, "weight_mean": 9.751071047503501e-06, "weight_min": 0.0, "weight_std": 1.81384584720945e-05 }, { "epoch": 0.5523284114879118, "step": 1702, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3499824490281753e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9181787138222717e-05 }, { "epoch": 0.5523284114879118, "step": 1702, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0418495548947249e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9222428818466142e-05 }, { "epoch": 0.5523284114879118, "step": 1702, "weight_max": 4.381950566312298e-05, "weight_mean": 1.575468559167348e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.17661199712893e-05 }, { "epoch": 0.552652928768457, "step": 1703, "weight_max": 3.061215829802677e-05, "weight_mean": 7.303077836695593e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3504452908819076e-05 }, { "epoch": 0.552652928768457, "step": 1703, "weight_max": 3.5434961318969727e-05, "weight_mean": 4.43003409600351e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.252788115380099e-05 }, { "epoch": 0.552652928768457, "step": 1703, "weight_max": 7.2748242985198885e-09, "weight_mean": 1.388639869759345e-09, "weight_min": 0.0, "weight_std": 2.389259012858247e-09 }, { "epoch": 0.552652928768457, "step": 1703, "weight_max": 2.202445102739148e-05, "weight_mean": 2.754208708211081e-06, "weight_min": 5.679522963042416e-10, "weight_std": 7.786353307892568e-06 }, { "epoch": 0.5529774460490021, "step": 1704, "weight_max": 3.4268457966391e-05, "weight_mean": 6.951271643629298e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3325184227142017e-05 }, { "epoch": 0.5529774460490021, "step": 1704, "weight_max": 3.813097282545641e-05, "weight_mean": 9.171447345579509e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6476733435411006e-05 }, { "epoch": 0.5529774460490021, "step": 1704, "weight_max": 3.998322426923551e-05, "weight_mean": 1.9078848708886653e-05, "weight_min": 0.0, "weight_std": 2.0445871996344067e-05 }, { "epoch": 0.5529774460490021, "step": 1704, "weight_max": 3.998322426923551e-05, "weight_mean": 9.464774848311208e-06, "weight_min": 0.0, "weight_std": 1.7560547348693945e-05 }, { "epoch": 0.5533019633295473, "step": 1705, "weight_max": 3.28377827827353e-05, "weight_mean": 4.107037511857925e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1608975000854116e-05 }, { "epoch": 0.5533019633295473, "step": 1705, "weight_max": 3.967146767536178e-05, "weight_mean": 1.4158720659906976e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9587172573665157e-05 }, { "epoch": 0.5533019633295473, "step": 1705, "weight_max": 3.967146767536178e-05, "weight_mean": 2.185555422329344e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.854805123002734e-05 }, { "epoch": 0.5533019633295473, "step": 1705, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.1821870430139825e-05, "weight_min": 0.0, "weight_std": 1.7187343473779038e-05 }, { "epoch": 0.5536264806100925, "step": 1706, "weight_max": 4.187939703115262e-05, "weight_mean": 9.482843779551331e-06, "weight_min": 0.0, "weight_std": 1.7684220438241027e-05 }, { "epoch": 0.5536264806100925, "step": 1706, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882484063273296e-06, "weight_min": 0.0, "weight_std": 1.3806410606775898e-05 }, { "epoch": 0.5536264806100925, "step": 1706, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.15692590852268e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4580758033844177e-05 }, { "epoch": 0.5536264806100925, "step": 1706, "weight_max": 4.252122016623616e-05, "weight_mean": 9.563254025124479e-06, "weight_min": 0.0, "weight_std": 1.7852731616585515e-05 }, { "epoch": 0.5539509978906377, "step": 1707, "weight_max": 3.6920544516760856e-05, "weight_mean": 5.40121664016624e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.2924213478981983e-05 }, { "epoch": 0.5539509978906377, "step": 1707, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156231054570526e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4581038158212323e-05 }, { "epoch": 0.5539509978906377, "step": 1707, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3135704648448154e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.870133201009594e-05 }, { "epoch": 0.5539509978906377, "step": 1707, "weight_max": 1.1534618815645103e-09, "weight_mean": 8.238619786382628e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.260210213433922e-10 }, { "epoch": 0.5542755151711829, "step": 1708, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4926731637388002e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.064584805339109e-05 }, { "epoch": 0.5542755151711829, "step": 1708, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4101962733548135e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9592082026065327e-05 }, { "epoch": 0.5542755151711829, "step": 1708, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631542863848154e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9687255189637654e-05 }, { "epoch": 0.5542755151711829, "step": 1708, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.2553982742247172e-05, "weight_min": 0.0, "weight_std": 1.785642780305352e-05 }, { "epoch": 0.5546000324517281, "step": 1709, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0392265721748117e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9238470486016013e-05 }, { "epoch": 0.5546000324517281, "step": 1709, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.989992238639388e-06, "weight_min": 0.0, "weight_std": 1.674417217145674e-05 }, { "epoch": 0.5546000324517281, "step": 1709, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.04839964985149e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5307408830267377e-05 }, { "epoch": 0.5546000324517281, "step": 1709, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156270162842702e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4581022696802393e-05 }, { "epoch": 0.5549245497322732, "grad_norm": 0.3923422694206238, "learning_rate": 6.7224489795918364e-06, "loss": 0.0718, "step": 1710 }, { "epoch": 0.5549245497322732, "step": 1710, "weight_max": 3.312147964606993e-05, "weight_mean": 4.140652890782803e-06, "weight_min": 0.0, "weight_std": 1.1710022590705194e-05 }, { "epoch": 0.5549245497322732, "step": 1710, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.920757874060655e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.3915930139773991e-05 }, { "epoch": 0.5549245497322732, "step": 1710, "weight_max": 3.397988257347606e-05, "weight_mean": 4.248265213391278e-06, "weight_min": 0.0, "weight_std": 1.2013387276965659e-05 }, { "epoch": 0.5549245497322732, "step": 1710, "weight_max": 3.998322426923551e-05, "weight_mean": 1.561016688356176e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.0023842807859182e-05 }, { "epoch": 0.5552490670128184, "step": 1711, "weight_max": 3.998322426923551e-05, "weight_mean": 9.765170034370385e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.8086147974827327e-05 }, { "epoch": 0.5552490670128184, "step": 1711, "weight_max": 8.926079999582726e-07, "weight_mean": 1.1213975881219085e-07, "weight_min": 5.771870759119224e-10, "weight_std": 3.1535680022898305e-07 }, { "epoch": 0.5552490670128184, "step": 1711, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3703549484489486e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9081922800978646e-05 }, { "epoch": 0.5552490670128184, "step": 1711, "weight_max": 4.252122016623616e-05, "weight_mean": 6.584672973986017e-06, "weight_min": 0.0, "weight_std": 1.4948798707337119e-05 }, { "epoch": 0.5555735842933637, "step": 1712, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.0229055078525562e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.62845135491807e-05 }, { "epoch": 0.5555735842933637, "step": 1712, "weight_max": 3.998322426923551e-05, "weight_mean": 1.3417455193120986e-05, "weight_min": 7.937498991417158e-10, "weight_std": 1.889221130113583e-05 }, { "epoch": 0.5555735842933637, "step": 1712, "weight_max": 3.998322426923551e-05, "weight_mean": 9.764806236489676e-06, "weight_min": 0.0, "weight_std": 1.8086371710523963e-05 }, { "epoch": 0.5555735842933637, "step": 1712, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3442561794363428e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.955327570613008e-05 }, { "epoch": 0.5558981015739088, "step": 1713, "weight_max": 9.413956281179026e-10, "weight_mean": 5.960094084933587e-10, "weight_min": 0.0, "weight_std": 2.6825425214482834e-10 }, { "epoch": 0.5558981015739088, "step": 1713, "weight_max": 3.843663216684945e-05, "weight_mean": 1.134601279773051e-05, "weight_min": 0.0, "weight_std": 1.612990308785811e-05 }, { "epoch": 0.5558981015739088, "step": 1713, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0155534255318344e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.881522075564135e-05 }, { "epoch": 0.5558981015739088, "step": 1713, "weight_max": 4.252122016623616e-05, "weight_mean": 9.280201084038708e-06, "weight_min": 0.0, "weight_std": 1.7423019016860053e-05 }, { "epoch": 0.556222618854454, "step": 1714, "weight_max": 3.6920544516760856e-05, "weight_mean": 7.969338184921071e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.476588749937946e-05 }, { "epoch": 0.556222618854454, "step": 1714, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.516573871020228e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.58289767568931e-05 }, { "epoch": 0.556222618854454, "step": 1714, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.9354289179318585e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.0774306904058903e-05 }, { "epoch": 0.556222618854454, "step": 1714, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235749085841235e-06, "weight_min": 0.0, "weight_std": 1.480626997363288e-05 }, { "epoch": 0.5565471361349992, "step": 1715, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0313809070794377e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.910764694912359e-05 }, { "epoch": 0.5565471361349992, "step": 1715, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0245754058123566e-05, "weight_min": 7.508524357824342e-10, "weight_std": 2.1721794837503694e-05 }, { "epoch": 0.5565471361349992, "step": 1715, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.91418858373072e-06, "weight_min": 0.0, "weight_std": 1.6772937669884413e-05 }, { "epoch": 0.5565471361349992, "step": 1715, "weight_max": 3.752332122530788e-05, "weight_mean": 9.083720215130597e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6830299500725232e-05 }, { "epoch": 0.5568716534155443, "step": 1716, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.076926299807383e-06, "weight_min": 0.0, "weight_std": 1.4357693544297945e-05 }, { "epoch": 0.5568716534155443, "step": 1716, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3129090550355613e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.9548102500266396e-05 }, { "epoch": 0.5568716534155443, "step": 1716, "weight_max": 3.967146767536178e-05, "weight_mean": 1.3191908692533616e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.6953639715211466e-05 }, { "epoch": 0.5568716534155443, "step": 1716, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5075818737386726e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0807412511203438e-05 }, { "epoch": 0.5571961706960896, "step": 1717, "weight_max": 3.8743506593164057e-05, "weight_mean": 8.537086614524014e-06, "weight_min": 0.0, "weight_std": 1.5990823158062994e-05 }, { "epoch": 0.5571961706960896, "step": 1717, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8599354007164948e-05, "weight_min": 0.0, "weight_std": 2.0136774764978327e-05 }, { "epoch": 0.5571961706960896, "step": 1717, "weight_max": 4.252122016623616e-05, "weight_mean": 1.498009714850923e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.077893259411212e-05 }, { "epoch": 0.5571961706960896, "step": 1717, "weight_max": 4.252122016623616e-05, "weight_mean": 1.9363335013622418e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.11104670597706e-05 }, { "epoch": 0.5575206879766348, "step": 1718, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0075318641611375e-05, "weight_min": 0.0, "weight_std": 1.865386548161041e-05 }, { "epoch": 0.5575206879766348, "step": 1718, "weight_max": 3.4268457966391e-05, "weight_mean": 4.285539489501389e-06, "weight_min": 0.0, "weight_std": 1.2114929631934501e-05 }, { "epoch": 0.5575206879766348, "step": 1718, "weight_max": 3.967146767536178e-05, "weight_mean": 9.280866834160406e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7233782273251563e-05 }, { "epoch": 0.5575206879766348, "step": 1718, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.94932144950144e-05, "weight_min": 0.0, "weight_std": 2.086290442093741e-05 }, { "epoch": 0.55784520525718, "step": 1719, "weight_max": 2.2961557988310233e-05, "weight_mean": 2.98674399346055e-06, "weight_min": 5.958802340444436e-10, "weight_std": 8.077562597463839e-06 }, { "epoch": 0.55784520525718, "step": 1719, "weight_max": 3.967146767536178e-05, "weight_mean": 8.55141752253985e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.6098601918201894e-05 }, { "epoch": 0.55784520525718, "step": 1719, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.938276106258854e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.390892703057034e-05 }, { "epoch": 0.55784520525718, "step": 1719, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.2954821613675449e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8015451132669114e-05 }, { "epoch": 0.5581697225377251, "grad_norm": 0.40615031123161316, "learning_rate": 6.702040816326531e-06, "loss": 0.1697, "step": 1720 }, { "epoch": 0.5581697225377251, "step": 1720, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.812774806050584e-06, "weight_min": 0.0, "weight_std": 1.8164844732382335e-05 }, { "epoch": 0.5581697225377251, "step": 1720, "weight_max": 4.187939703115262e-05, "weight_mean": 9.971121471608058e-06, "weight_min": 0.0, "weight_std": 1.8475264369044453e-05 }, { "epoch": 0.5581697225377251, "step": 1720, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4614053725381382e-05, "weight_min": 0.0, "weight_std": 2.0200996004859917e-05 }, { "epoch": 0.5581697225377251, "step": 1720, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.922416211338714e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.838973730627913e-05 }, { "epoch": 0.5584942398182703, "step": 1721, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1563765737228096e-06, "weight_min": 0.0, "weight_std": 1.4580979041056708e-05 }, { "epoch": 0.5584942398182703, "step": 1721, "weight_max": 3.722132169059478e-05, "weight_mean": 8.548436198907439e-06, "weight_min": 0.0, "weight_std": 1.590976353327278e-05 }, { "epoch": 0.5584942398182703, "step": 1721, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393460585561115e-06, "weight_min": 0.0, "weight_std": 1.2424085070961155e-05 }, { "epoch": 0.5584942398182703, "step": 1721, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9589766452554613e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1006213501095772e-05 }, { "epoch": 0.5588187570988156, "step": 1722, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077241439721547e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4357566215039697e-05 }, { "epoch": 0.5588187570988156, "step": 1722, "weight_max": 3.255531191825867e-05, "weight_mean": 8.069183422776405e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4940701476007234e-05 }, { "epoch": 0.5588187570988156, "step": 1722, "weight_max": 8.378387983398738e-10, "weight_mean": 6.803472230032526e-10, "weight_min": 5.958802340444436e-10, "weight_std": 8.634380016525611e-11 }, { "epoch": 0.5588187570988156, "step": 1722, "weight_max": 4.252122016623616e-05, "weight_mean": 1.000608790491242e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.857448114606086e-05 }, { "epoch": 0.5591432743793607, "step": 1723, "weight_max": 4.187939703115262e-05, "weight_mean": 7.845781510695815e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.557208815938793e-05 }, { "epoch": 0.5591432743793607, "step": 1723, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5467594494111836e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.969227014342323e-05 }, { "epoch": 0.5591432743793607, "step": 1723, "weight_max": 4.381950566312298e-05, "weight_mean": 5.478294042404741e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.549218904983718e-05 }, { "epoch": 0.5591432743793607, "step": 1723, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0478071089892182e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.9421962861088105e-05 }, { "epoch": 0.5594677916599059, "step": 1724, "weight_max": 3.967146767536178e-05, "weight_mean": 4.959377747582039e-06, "weight_min": 0.0, "weight_std": 1.402580164722167e-05 }, { "epoch": 0.5594677916599059, "step": 1724, "weight_max": 3.9360926166409627e-05, "weight_mean": 8.954895747592673e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.668778895691503e-05 }, { "epoch": 0.5594677916599059, "step": 1724, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5153435015236028e-05, "weight_min": 0.0, "weight_std": 2.092439535772428e-05 }, { "epoch": 0.5594677916599059, "step": 1724, "weight_max": 3.843663216684945e-05, "weight_mean": 1.858160612755455e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9577108105295338e-05 }, { "epoch": 0.5597923089404511, "step": 1725, "weight_max": 3.9360926166409627e-05, "weight_mean": 4.921162599202944e-06, "weight_min": 6.686325937366178e-10, "weight_std": 1.391576552123297e-05 }, { "epoch": 0.5597923089404511, "step": 1725, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0138042853213847e-05, "weight_min": 0.0, "weight_std": 1.5686018741689622e-05 }, { "epoch": 0.5597923089404511, "step": 1725, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.3595579730463214e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9114260794594884e-05 }, { "epoch": 0.5597923089404511, "step": 1725, "weight_max": 4.187939703115262e-05, "weight_mean": 2.351429793634452e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9707358660525642e-05 }, { "epoch": 0.5601168262209962, "step": 1726, "weight_max": 4.187939703115262e-05, "weight_mean": 8.63084369484568e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6456568118883297e-05 }, { "epoch": 0.5601168262209962, "step": 1726, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.886691097752191e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.057422352663707e-05 }, { "epoch": 0.5601168262209962, "step": 1726, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9695980881806463e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.1116327843628824e-05 }, { "epoch": 0.5601168262209962, "step": 1726, "weight_max": 7.194614903838215e-10, "weight_mean": 5.508172806756306e-10, "weight_min": 0.0, "weight_std": 2.2760775209018647e-10 }, { "epoch": 0.5604413435015415, "step": 1727, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.0283452613512054e-05, "weight_min": 6.100955851628953e-10, "weight_std": 2.1755800844402984e-05 }, { "epoch": 0.5604413435015415, "step": 1727, "weight_max": 4.252122016623616e-05, "weight_mean": 2.3962500563357025e-05, "weight_min": 5.771870759119224e-10, "weight_std": 2.023140223172959e-05 }, { "epoch": 0.5604413435015415, "step": 1727, "weight_max": 3.813097282545641e-05, "weight_mean": 7.726692274445668e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.4817782357567921e-05 }, { "epoch": 0.5604413435015415, "step": 1727, "weight_max": 4.252122016623616e-05, "weight_mean": 1.2732522009173408e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8013457520282827e-05 }, { "epoch": 0.5607658607820867, "step": 1728, "weight_max": 9.230136129190214e-06, "weight_mean": 1.1563080306586926e-06, "weight_min": 0.0, "weight_std": 3.2623240713292034e-06 }, { "epoch": 0.5607658607820867, "step": 1728, "weight_max": 4.252122016623616e-05, "weight_mean": 9.745779607328586e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8141699911211617e-05 }, { "epoch": 0.5607658607820867, "step": 1728, "weight_max": 3.006794759130571e-05, "weight_mean": 3.759947958315024e-06, "weight_min": 0.0, "weight_std": 1.0630036740622018e-05 }, { "epoch": 0.5607658607820867, "step": 1728, "weight_max": 3.967146767536178e-05, "weight_mean": 6.792747626604978e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4240473319659941e-05 }, { "epoch": 0.5610903780626318, "step": 1729, "weight_max": 2.133443376806099e-05, "weight_mean": 4.4292391976341605e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.425302439718507e-06 }, { "epoch": 0.5610903780626318, "step": 1729, "weight_max": 4.187939703115262e-05, "weight_mean": 2.2672447812510654e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8977569197886623e-05 }, { "epoch": 0.5610903780626318, "step": 1729, "weight_max": 4.252122016623616e-05, "weight_mean": 1.6095782484626397e-05, "weight_min": 0.0, "weight_std": 2.138094168913085e-05 }, { "epoch": 0.5610903780626318, "step": 1729, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.53513790591387e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.7659898730926216e-05 }, { "epoch": 0.561414895343177, "grad_norm": 0.17683692276477814, "learning_rate": 6.6816326530612255e-06, "loss": 0.0685, "step": 1730 }, { "epoch": 0.561414895343177, "step": 1730, "weight_max": 2.872879122151062e-05, "weight_mean": 3.5919849779020296e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.0156803000427317e-05 }, { "epoch": 0.561414895343177, "step": 1730, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.871956374088768e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0149942429270595e-05 }, { "epoch": 0.561414895343177, "step": 1730, "weight_max": 3.4268457966391e-05, "weight_mean": 1.6120775399031118e-05, "weight_min": 0.0, "weight_std": 1.7309932445641607e-05 }, { "epoch": 0.561414895343177, "step": 1730, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.396418136020657e-06, "weight_min": 0.0, "weight_std": 1.7472986655775458e-05 }, { "epoch": 0.5617394126237222, "step": 1731, "weight_max": 3.199401908204891e-05, "weight_mean": 1.0211247172264848e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.4457328688877169e-05 }, { "epoch": 0.5617394126237222, "step": 1731, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5635503334342502e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.129598578903824e-05 }, { "epoch": 0.5617394126237222, "step": 1731, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0390689567429945e-05, "weight_min": 0.0, "weight_std": 1.9239441826357506e-05 }, { "epoch": 0.5617394126237222, "step": 1731, "weight_max": 4.447596495538164e-07, "weight_mean": 5.622852583542226e-08, "weight_min": 0.0, "weight_std": 1.5699060895713046e-07 }, { "epoch": 0.5620639299042675, "step": 1732, "weight_max": 3.843663216684945e-05, "weight_mean": 8.840017471811734e-06, "weight_min": 0.0, "weight_std": 1.6448282622150145e-05 }, { "epoch": 0.5620639299042675, "step": 1732, "weight_max": 3.116124207735993e-05, "weight_mean": 3.897715942002833e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.101612997445045e-05 }, { "epoch": 0.5620639299042675, "step": 1732, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4745826774742454e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.0610756109817885e-05 }, { "epoch": 0.5620639299042675, "step": 1732, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.29175439523533e-06, "weight_min": 0.0, "weight_std": 1.7500200556241907e-05 }, { "epoch": 0.5623884471848126, "step": 1733, "weight_max": 3.813097282545641e-05, "weight_mean": 1.3553224562201649e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.8728631403064355e-05 }, { "epoch": 0.5623884471848126, "step": 1733, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3172907428233884e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5032658666314092e-05 }, { "epoch": 0.5623884471848126, "step": 1733, "weight_max": 3.843663216684945e-05, "weight_mean": 1.885646634036675e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.6936695828917436e-05 }, { "epoch": 0.5623884471848126, "step": 1733, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.8271901353728026e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9836332285194658e-05 }, { "epoch": 0.5627129644653578, "step": 1734, "weight_max": 2.8994179956498556e-05, "weight_mean": 3.6251860819902504e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.0250621016894002e-05 }, { "epoch": 0.5627129644653578, "step": 1734, "weight_max": 3.752332122530788e-05, "weight_mean": 8.484206773573533e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.582325239724014e-05 }, { "epoch": 0.5627129644653578, "step": 1734, "weight_max": 3.602553260861896e-05, "weight_mean": 9.007171684061177e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.6676094674039632e-05 }, { "epoch": 0.5627129644653578, "step": 1734, "weight_max": 2.820166446326766e-05, "weight_mean": 3.526079353832756e-06, "weight_min": 6.053385570581327e-10, "weight_std": 9.970442079065833e-06 }, { "epoch": 0.563037481745903, "step": 1735, "weight_max": 4.252122016623616e-05, "weight_mean": 2.060512633761391e-05, "weight_min": 6.786494144428445e-10, "weight_std": 2.201087772846222e-05 }, { "epoch": 0.563037481745903, "step": 1735, "weight_max": 3.5434961318969727e-05, "weight_mean": 1.013854762277333e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.492250885348767e-05 }, { "epoch": 0.563037481745903, "step": 1735, "weight_max": 1.5802011921195458e-09, "weight_mean": 5.975558936555103e-10, "weight_min": 0.0, "weight_std": 5.818066584062365e-10 }, { "epoch": 0.563037481745903, "step": 1735, "weight_max": 1.7219337678398006e-05, "weight_mean": 2.183172682634904e-06, "weight_min": 5.771870759119224e-10, "weight_std": 6.076045337977121e-06 }, { "epoch": 0.5633619990264481, "step": 1736, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.156325187272159e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.4580999959434848e-05 }, { "epoch": 0.5633619990264481, "step": 1736, "weight_max": 5.26445717241586e-07, "weight_mean": 6.740324920428975e-08, "weight_min": 5.864964625068581e-10, "weight_std": 1.85500411475914e-07 }, { "epoch": 0.5633619990264481, "step": 1736, "weight_max": 2.3835209361777743e-09, "weight_mean": 1.1699859969738213e-09, "weight_min": 0.0, "weight_std": 7.363751830524734e-10 }, { "epoch": 0.5633619990264481, "step": 1736, "weight_max": 4.252122016623616e-05, "weight_mean": 6.726266292389482e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4992612705100328e-05 }, { "epoch": 0.5636865163069934, "step": 1737, "weight_max": 3.255531191825867e-05, "weight_mean": 4.070208888151683e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.1509719115565531e-05 }, { "epoch": 0.5636865163069934, "step": 1737, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.4102248314884491e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.920152317325119e-05 }, { "epoch": 0.5636865163069934, "step": 1737, "weight_max": 3.9360926166409627e-05, "weight_mean": 9.025434337672777e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.680108289292548e-05 }, { "epoch": 0.5636865163069934, "step": 1737, "weight_max": 3.722132169059478e-05, "weight_mean": 1.1119165719719604e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.572963628859725e-05 }, { "epoch": 0.5640110335875386, "step": 1738, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0776802709151525e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4357387954078149e-05 }, { "epoch": 0.5640110335875386, "step": 1738, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4578919945051894e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.0168921764707193e-05 }, { "epoch": 0.5640110335875386, "step": 1738, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3980989024275914e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.9391804016777314e-05 }, { "epoch": 0.5640110335875386, "step": 1738, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5171592167462222e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1023883164161816e-05 }, { "epoch": 0.5643355508680837, "step": 1739, "weight_max": 3.752332122530788e-05, "weight_mean": 1.2510849046520889e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7632468370720744e-05 }, { "epoch": 0.5643355508680837, "step": 1739, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.5485311198281124e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.120564204233233e-05 }, { "epoch": 0.5643355508680837, "step": 1739, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.960367606254295e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.8457069018040784e-05 }, { "epoch": 0.5643355508680837, "step": 1739, "weight_max": 3.662098970380612e-05, "weight_mean": 6.22617153567262e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.3119131835992448e-05 }, { "epoch": 0.5646600681486289, "grad_norm": 19.235923767089844, "learning_rate": 6.661224489795919e-06, "loss": 0.1675, "step": 1740 }, { "epoch": 0.5646600681486289, "step": 1740, "weight_max": 3.998322426923551e-05, "weight_mean": 8.490852451359387e-06, "weight_min": 0.0, "weight_std": 1.6046991731855087e-05 }, { "epoch": 0.5646600681486289, "step": 1740, "weight_max": 3.397988257347606e-05, "weight_mean": 8.074535799096338e-06, "weight_min": 0.0, "weight_std": 1.4976893908169586e-05 }, { "epoch": 0.5646600681486289, "step": 1740, "weight_max": 2.690521432668902e-05, "weight_mean": 3.3641561003605602e-06, "weight_min": 0.0, "weight_std": 9.512023098068312e-06 }, { "epoch": 0.5646600681486289, "step": 1740, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8880382413044572e-05, "weight_min": 6.488223291967188e-10, "weight_std": 2.0636940462281927e-05 }, { "epoch": 0.5649845854291741, "step": 1741, "weight_max": 3.782653220696375e-05, "weight_mean": 4.729276952275541e-06, "weight_min": 0.0, "weight_std": 1.3373310139286332e-05 }, { "epoch": 0.5649845854291741, "step": 1741, "weight_max": 2.690521432668902e-05, "weight_mean": 3.3640988021943485e-06, "weight_min": 0.0, "weight_std": 9.51204765442526e-06 }, { "epoch": 0.5649845854291741, "step": 1741, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105394054931821e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.160963893198641e-05 }, { "epoch": 0.5649845854291741, "step": 1741, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0006477168644778e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8574241039459594e-05 }, { "epoch": 0.5653091027097193, "step": 1742, "weight_max": 3.998322426923551e-05, "weight_mean": 1.8579736206447706e-05, "weight_min": 6.488223291967188e-10, "weight_std": 1.8952863683807664e-05 }, { "epoch": 0.5653091027097193, "step": 1742, "weight_max": 6.077719216790456e-09, "weight_mean": 1.5025310995397945e-09, "weight_min": 5.864964625068581e-10, "weight_std": 1.8726582506900513e-09 }, { "epoch": 0.5653091027097193, "step": 1742, "weight_max": 1.2744457711377777e-09, "weight_mean": 7.067705309893313e-10, "weight_min": 5.679522963042416e-10, "weight_std": 2.320870162719757e-10 }, { "epoch": 0.5653091027097193, "step": 1742, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.68531788000837e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.6122286979225464e-05 }, { "epoch": 0.5656336199902645, "step": 1743, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3966310689284e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.946882730408106e-05 }, { "epoch": 0.5656336199902645, "step": 1743, "weight_max": 4.252122016623616e-05, "weight_mean": 2.366072658332996e-05, "weight_min": 0.0, "weight_std": 2.011940705415327e-05 }, { "epoch": 0.5656336199902645, "step": 1743, "weight_max": 4.187939703115262e-05, "weight_mean": 1.575896385475062e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.846523991844151e-05 }, { "epoch": 0.5656336199902645, "step": 1743, "weight_max": 3.998322426923551e-05, "weight_mean": 1.22545652629924e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.7648062566877343e-05 }, { "epoch": 0.5659581372708097, "step": 1744, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.464793058403302e-05, "weight_min": 6.989064882390039e-10, "weight_std": 2.022747867158614e-05 }, { "epoch": 0.5659581372708097, "step": 1744, "weight_max": 4.187939703115262e-05, "weight_mean": 5.690410944225732e-06, "weight_min": 0.0, "weight_std": 1.467710990255e-05 }, { "epoch": 0.5659581372708097, "step": 1744, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3785600458504632e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.908879858092405e-05 }, { "epoch": 0.5659581372708097, "step": 1744, "weight_max": 4.187939703115262e-05, "weight_mean": 1.019463252305286e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.8884104065364227e-05 }, { "epoch": 0.5662826545513548, "step": 1745, "weight_max": 3.813097282545641e-05, "weight_mean": 8.710025213076733e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.6192545444937423e-05 }, { "epoch": 0.5662826545513548, "step": 1745, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.654442692408338e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.790704482118599e-05 }, { "epoch": 0.5662826545513548, "step": 1745, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.96036487777019e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.8457070837030187e-05 }, { "epoch": 0.5662826545513548, "step": 1745, "weight_max": 3.813097282545641e-05, "weight_mean": 9.269971997127868e-06, "weight_min": 0.0, "weight_std": 1.7172866137116216e-05 }, { "epoch": 0.5666071718319, "step": 1746, "weight_max": 3.722132169059478e-05, "weight_mean": 1.3115602996549569e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.8142483895644546e-05 }, { "epoch": 0.5666071718319, "step": 1746, "weight_max": 3.998322426923551e-05, "weight_mean": 8.62592423800379e-06, "weight_min": 0.0, "weight_std": 1.6230784240178764e-05 }, { "epoch": 0.5666071718319, "step": 1746, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.927632729348261e-05, "weight_min": 6.686325937366178e-10, "weight_std": 2.069054426101502e-05 }, { "epoch": 0.5666071718319, "step": 1746, "weight_max": 4.381950566312298e-05, "weight_mean": 1.3764120922132861e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.9644810890895315e-05 }, { "epoch": 0.5669316891124453, "step": 1747, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3120670701027848e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.8486418412066996e-05 }, { "epoch": 0.5669316891124453, "step": 1747, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.0122656021849252e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6469601177959703e-05 }, { "epoch": 0.5669316891124453, "step": 1747, "weight_max": 3.28377827827353e-05, "weight_mean": 4.391458787722513e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.1521900887601078e-05 }, { "epoch": 0.5669316891124453, "step": 1747, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284424448997015e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.2115378922317177e-05 }, { "epoch": 0.5672562063929905, "step": 1748, "weight_max": 3.967146767536178e-05, "weight_mean": 8.452272595604882e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.595949288457632e-05 }, { "epoch": 0.5672562063929905, "step": 1748, "weight_max": 4.252122016623616e-05, "weight_mean": 2.0586128812283278e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.2021607946953736e-05 }, { "epoch": 0.5672562063929905, "step": 1748, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.155885446583852e-06, "weight_min": 0.0, "weight_std": 1.4581178220396396e-05 }, { "epoch": 0.5672562063929905, "step": 1748, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.8575256035546772e-05, "weight_min": 0.0, "weight_std": 2.015934842347633e-05 }, { "epoch": 0.5675807236735356, "step": 1749, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.4017829168296885e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.2420739949448034e-05 }, { "epoch": 0.5675807236735356, "step": 1749, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.6876290828804485e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.846506529545877e-05 }, { "epoch": 0.5675807236735356, "step": 1749, "weight_max": 3.662098970380612e-05, "weight_mean": 5.887118277314585e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.2947417417308316e-05 }, { "epoch": 0.5675807236735356, "step": 1749, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.536801371723413e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7860160369309597e-05 }, { "epoch": 0.5679052409540808, "grad_norm": 0.15648913383483887, "learning_rate": 6.640816326530612e-06, "loss": 0.0956, "step": 1750 }, { "epoch": 0.5679052409540808, "step": 1750, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4217224816093221e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9910128685296513e-05 }, { "epoch": 0.5679052409540808, "step": 1750, "weight_max": 2.714597657771378e-09, "weight_mean": 1.0350427182004296e-09, "weight_min": 6.390288853630466e-10, "weight_std": 6.941563990281452e-10 }, { "epoch": 0.5679052409540808, "step": 1750, "weight_max": 3.6920544516760856e-05, "weight_mean": 4.615675607055891e-06, "weight_min": 0.0, "weight_std": 1.3053137990937103e-05 }, { "epoch": 0.5679052409540808, "step": 1750, "weight_max": 4.187939703115262e-05, "weight_mean": 1.015559428196866e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.8815182556863874e-05 }, { "epoch": 0.568229758234626, "step": 1751, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5310839444282465e-05, "weight_min": 7.091466858177853e-10, "weight_std": 2.1129415472387336e-05 }, { "epoch": 0.568229758234626, "step": 1751, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0631478289724328e-05, "weight_min": 0.0, "weight_std": 1.9687293388415128e-05 }, { "epoch": 0.568229758234626, "step": 1751, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3920814126322512e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9381011952646077e-05 }, { "epoch": 0.568229758234626, "step": 1751, "weight_max": 3.998322426923551e-05, "weight_mean": 7.212007403722964e-06, "weight_min": 6.989064882390039e-10, "weight_std": 1.460644489270635e-05 }, { "epoch": 0.5685542755151712, "step": 1752, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4935623767087236e-05, "weight_min": 6.390288853630466e-10, "weight_std": 1.701985274848994e-05 }, { "epoch": 0.5685542755151712, "step": 1752, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4271137843024917e-05, "weight_min": 0.0, "weight_std": 1.9714234440471046e-05 }, { "epoch": 0.5685542755151712, "step": 1752, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0312827726011164e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.909018465084955e-05 }, { "epoch": 0.5685542755151712, "step": 1752, "weight_max": 3.813097282545641e-05, "weight_mean": 1.3666951417690143e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.8897262634709477e-05 }, { "epoch": 0.5688787927957164, "step": 1753, "weight_max": 2.2888118564878823e-06, "weight_mean": 2.867610646717367e-07, "weight_min": 6.053385570581327e-10, "weight_std": 8.089507446129574e-07 }, { "epoch": 0.5688787927957164, "step": 1753, "weight_max": 1.5038247511256486e-05, "weight_mean": 1.8810007986758137e-06, "weight_min": 6.390288853630466e-10, "weight_std": 5.316330771165667e-06 }, { "epoch": 0.5688787927957164, "step": 1753, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.901527684822213e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.043274616880808e-05 }, { "epoch": 0.5688787927957164, "step": 1753, "weight_max": 3.998322426923551e-05, "weight_mean": 1.7819844288169406e-05, "weight_min": 0.0, "weight_std": 1.9698869436979294e-05 }, { "epoch": 0.5692033100762616, "step": 1754, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.077110017737141e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.43576189657324e-05 }, { "epoch": 0.5692033100762616, "step": 1754, "weight_max": 1.9767097910516895e-05, "weight_mean": 4.352375071903225e-06, "weight_min": 0.0, "weight_std": 8.153367161867209e-06 }, { "epoch": 0.5692033100762616, "step": 1754, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.324443453806452e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7354601368424483e-05 }, { "epoch": 0.5692033100762616, "step": 1754, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0110973562404979e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.8427188479108736e-05 }, { "epoch": 0.5695278273568067, "step": 1755, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5626119420630857e-05, "weight_min": 0.0, "weight_std": 2.15689269680297e-05 }, { "epoch": 0.5695278273568067, "step": 1755, "weight_max": 3.813097282545641e-05, "weight_mean": 8.942694876168389e-06, "weight_min": 0.0, "weight_std": 1.6605437849648297e-05 }, { "epoch": 0.5695278273568067, "step": 1755, "weight_max": 4.252122016623616e-05, "weight_mean": 1.4970054508012254e-05, "weight_min": 0.0, "weight_std": 2.0721059627248906e-05 }, { "epoch": 0.5695278273568067, "step": 1755, "weight_max": 3.843663216684945e-05, "weight_mean": 5.946230885456316e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.3308617781149223e-05 }, { "epoch": 0.5698523446373519, "step": 1756, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.996885637519881e-06, "weight_min": 0.0, "weight_std": 1.851248998718802e-05 }, { "epoch": 0.5698523446373519, "step": 1756, "weight_max": 1.076157385426768e-09, "weight_mean": 7.644231914127886e-10, "weight_min": 5.864964625068581e-10, "weight_std": 2.0308185688655556e-10 }, { "epoch": 0.5698523446373519, "step": 1756, "weight_max": 3.782653220696375e-05, "weight_mean": 8.62535944179399e-06, "weight_min": 7.937498991417158e-10, "weight_std": 1.6065469026216306e-05 }, { "epoch": 0.5698523446373519, "step": 1756, "weight_max": 3.6920544516760856e-05, "weight_mean": 9.052890163729899e-06, "weight_min": 6.786494144428445e-10, "weight_std": 1.6658488675602712e-05 }, { "epoch": 0.5701768619178972, "step": 1757, "weight_max": 2.716206836339552e-05, "weight_mean": 4.62569369119592e-06, "weight_min": 0.0, "weight_std": 9.735294042911846e-06 }, { "epoch": 0.5701768619178972, "step": 1757, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.430929821974132e-06, "weight_min": 0.0, "weight_std": 1.7702075638226233e-05 }, { "epoch": 0.5701768619178972, "step": 1757, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.59146927925758e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7711823602439836e-05 }, { "epoch": 0.5701768619178972, "step": 1757, "weight_max": 3.572963396436535e-05, "weight_mean": 4.466847713047173e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2632072866836097e-05 }, { "epoch": 0.5705013791984423, "step": 1758, "weight_max": 2.133443376806099e-05, "weight_mean": 2.6675386379793053e-06, "weight_min": 0.0, "weight_std": 7.542564162577037e-06 }, { "epoch": 0.5705013791984423, "step": 1758, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1117385838588234e-05, "weight_min": 0.0, "weight_std": 1.7192882296512835e-05 }, { "epoch": 0.5705013791984423, "step": 1758, "weight_max": 3.306819484905077e-09, "weight_mean": 9.23741749758733e-10, "weight_min": 0.0, "weight_std": 9.952020318948485e-10 }, { "epoch": 0.5705013791984423, "step": 1758, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3493534424924292e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.747822534525767e-05 }, { "epoch": 0.5708258964789875, "step": 1759, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.2630770470423158e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.7735759684001096e-05 }, { "epoch": 0.5708258964789875, "step": 1759, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2623761904251296e-05, "weight_min": 0.0, "weight_std": 1.7841670342022553e-05 }, { "epoch": 0.5708258964789875, "step": 1759, "weight_max": 3.4268457966391e-05, "weight_mean": 1.2113272532587871e-05, "weight_min": 0.0, "weight_std": 1.673245060374029e-05 }, { "epoch": 0.5708258964789875, "step": 1759, "weight_max": 3.752332122530788e-05, "weight_mean": 8.382196028833278e-06, "weight_min": 0.0, "weight_std": 1.5665475075365975e-05 }, { "epoch": 0.5711504137595327, "grad_norm": 0.052877746522426605, "learning_rate": 6.620408163265306e-06, "loss": 0.1813, "step": 1760 }, { "epoch": 0.5711504137595327, "step": 1760, "weight_max": 4.252122016623616e-05, "weight_mean": 2.334935197723098e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9547631382010877e-05 }, { "epoch": 0.5711504137595327, "step": 1760, "weight_max": 4.0610404539620504e-05, "weight_mean": 5.0768630899256095e-06, "weight_min": 0.0, "weight_std": 1.4357719010149594e-05 }, { "epoch": 0.5711504137595327, "step": 1760, "weight_max": 4.3167921830900013e-05, "weight_mean": 5.397522272687638e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.5261546650435776e-05 }, { "epoch": 0.5711504137595327, "step": 1760, "weight_max": 1.179825903641074e-09, "weight_mean": 7.934658485808654e-10, "weight_min": 6.196655966128617e-10, "weight_std": 1.9010874818814472e-10 }, { "epoch": 0.5714749310400778, "step": 1761, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0274696251144633e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9038774553337134e-05 }, { "epoch": 0.5714749310400778, "step": 1761, "weight_max": 4.252122016623616e-05, "weight_mean": 1.0552190360613167e-05, "weight_min": 0.0, "weight_std": 1.9534396415110677e-05 }, { "epoch": 0.5714749310400778, "step": 1761, "weight_max": 1.1403916699848082e-09, "weight_mean": 7.031701887427744e-10, "weight_min": 0.0, "weight_std": 3.5585326352283175e-10 }, { "epoch": 0.5714749310400778, "step": 1761, "weight_max": 4.187939703115262e-05, "weight_mean": 2.290480915689841e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.9529778001015075e-05 }, { "epoch": 0.5717994483206231, "step": 1762, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0232414751953911e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.8945513147627935e-05 }, { "epoch": 0.5717994483206231, "step": 1762, "weight_max": 3.116124207735993e-05, "weight_mean": 3.895969257428078e-06, "weight_min": 0.0, "weight_std": 1.1016833923349623e-05 }, { "epoch": 0.5717994483206231, "step": 1762, "weight_max": 2.8994179956498556e-05, "weight_mean": 5.701127520296723e-06, "weight_min": 0.0, "weight_std": 1.1061689292546362e-05 }, { "epoch": 0.5717994483206231, "step": 1762, "weight_max": 3.1715208024252206e-05, "weight_mean": 3.96500354327145e-06, "weight_min": 0.0, "weight_std": 1.1212776371394284e-05 }, { "epoch": 0.5721239656011683, "step": 1763, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.7791828213375993e-05, "weight_min": 0.0, "weight_std": 1.9180508388672024e-05 }, { "epoch": 0.5721239656011683, "step": 1763, "weight_max": 2.3835209361777743e-09, "weight_mean": 9.963112557187515e-10, "weight_min": 6.488223291967188e-10, "weight_std": 5.781427003803685e-10 }, { "epoch": 0.5721239656011683, "step": 1763, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.15585361426929e-06, "weight_min": 0.0, "weight_std": 1.458119095332222e-05 }, { "epoch": 0.5721239656011683, "step": 1763, "weight_max": 4.3167921830900013e-05, "weight_mean": 9.6088178906939e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.7968308384297416e-05 }, { "epoch": 0.5724484828817135, "step": 1764, "weight_max": 2.846461757144425e-05, "weight_mean": 3.5588686841947492e-06, "weight_min": 0.0, "weight_std": 1.0063442459795624e-05 }, { "epoch": 0.5724484828817135, "step": 1764, "weight_max": 3.624825550829769e-09, "weight_mean": 1.0593144139647848e-09, "weight_min": 5.771870759119224e-10, "weight_std": 1.0474775491431387e-09 }, { "epoch": 0.5724484828817135, "step": 1764, "weight_max": 3.28377827827353e-05, "weight_mean": 4.105438620172208e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.1609619832597673e-05 }, { "epoch": 0.5724484828817135, "step": 1764, "weight_max": 4.187939703115262e-05, "weight_mean": 1.805338979465887e-05, "weight_min": 0.0, "weight_std": 1.989759948628489e-05 }, { "epoch": 0.5727730001622586, "step": 1765, "weight_max": 3.722132169059478e-05, "weight_mean": 1.0296873369952664e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.6841760952956975e-05 }, { "epoch": 0.5727730001622586, "step": 1765, "weight_max": 3.967146767536178e-05, "weight_mean": 9.46397540246835e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.754638651618734e-05 }, { "epoch": 0.5727730001622586, "step": 1765, "weight_max": 4.187939703115262e-05, "weight_mean": 1.7214439139934257e-05, "weight_min": 0.0, "weight_std": 1.9882822016370483e-05 }, { "epoch": 0.5727730001622586, "step": 1765, "weight_max": 3.752332122530788e-05, "weight_mean": 4.691340109275188e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.3266123460198287e-05 }, { "epoch": 0.5730975174428038, "step": 1766, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.882414815365337e-06, "weight_min": 0.0, "weight_std": 1.830410838010721e-05 }, { "epoch": 0.5730975174428038, "step": 1766, "weight_max": 4.187939703115262e-05, "weight_mean": 1.9861148757627234e-05, "weight_min": 9.295914038531805e-10, "weight_std": 2.106322608597111e-05 }, { "epoch": 0.5730975174428038, "step": 1766, "weight_max": 3.28377827827353e-05, "weight_mean": 5.53664631297579e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.1736432497855276e-05 }, { "epoch": 0.5730975174428038, "step": 1766, "weight_max": 3.632264997577295e-05, "weight_mean": 8.369368515559472e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.5565845387754962e-05 }, { "epoch": 0.5734220347233491, "step": 1767, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.7255654206092004e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0534725333855022e-05 }, { "epoch": 0.5734220347233491, "step": 1767, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.45103367685806e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.0105688236071728e-05 }, { "epoch": 0.5734220347233491, "step": 1767, "weight_max": 1.3727149417164242e-09, "weight_mean": 7.169677074259084e-10, "weight_min": 0.0, "weight_std": 4.135818632011734e-10 }, { "epoch": 0.5734220347233491, "step": 1767, "weight_max": 4.187939703115262e-05, "weight_mean": 1.8920403817901388e-05, "weight_min": 6.29309993005478e-10, "weight_std": 2.0394900275277905e-05 }, { "epoch": 0.5737465520038942, "step": 1768, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.39906203711871e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7471360479248688e-05 }, { "epoch": 0.5737465520038942, "step": 1768, "weight_max": 3.4268457966391e-05, "weight_mean": 5.040368250774918e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.1997424735454842e-05 }, { "epoch": 0.5737465520038942, "step": 1768, "weight_max": 9.772551656794803e-10, "weight_mean": 6.847260536346766e-10, "weight_min": 0.0, "weight_std": 3.151414407209785e-10 }, { "epoch": 0.5737465520038942, "step": 1768, "weight_max": 2.2961557988310233e-05, "weight_mean": 2.8710130663966993e-06, "weight_min": 6.100955851628953e-10, "weight_std": 8.117805919027887e-06 }, { "epoch": 0.5740710692844394, "step": 1769, "weight_max": 4.252122016623616e-05, "weight_mean": 5.3188687161309645e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.503202202002285e-05 }, { "epoch": 0.5740710692844394, "step": 1769, "weight_max": 1.6209629393415526e-05, "weight_mean": 2.652629518706817e-06, "weight_min": 5.958802340444436e-10, "weight_std": 5.749894171458436e-06 }, { "epoch": 0.5740710692844394, "step": 1769, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.1190132681804243e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.6270139894913882e-05 }, { "epoch": 0.5740710692844394, "step": 1769, "weight_max": 1.7424939869670197e-05, "weight_mean": 2.1921455299889203e-06, "weight_min": 6.196655966128617e-10, "weight_std": 6.15508133705589e-06 }, { "epoch": 0.5743955865649846, "grad_norm": 0.10346938669681549, "learning_rate": 6.600000000000001e-06, "loss": 0.3255, "step": 1770 }, { "epoch": 0.5743955865649846, "step": 1770, "weight_max": 2.5141427613561973e-05, "weight_mean": 3.143507683489588e-06, "weight_min": 5.864964625068581e-10, "weight_std": 8.888501724868547e-06 }, { "epoch": 0.5743955865649846, "step": 1770, "weight_max": 3.722132169059478e-05, "weight_mean": 7.827007721061818e-06, "weight_min": 0.0, "weight_std": 1.4832248780294321e-05 }, { "epoch": 0.5743955865649846, "step": 1770, "weight_max": 5.874315434084565e-08, "weight_mean": 8.404488660573861e-09, "weight_min": 6.053385570581327e-10, "weight_std": 2.03541787868744e-08 }, { "epoch": 0.5743955865649846, "step": 1770, "weight_max": 4.252122016623616e-05, "weight_mean": 1.890157363959588e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.046476220129989e-05 }, { "epoch": 0.5747201038455297, "step": 1771, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355406629969366e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.9196053472114727e-05 }, { "epoch": 0.5747201038455297, "step": 1771, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.082367796509061e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.5213778169709258e-05 }, { "epoch": 0.5747201038455297, "step": 1771, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.9731223801500164e-05, "weight_min": 5.679522963042416e-10, "weight_std": 2.114148264809046e-05 }, { "epoch": 0.5747201038455297, "step": 1771, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5429113773279823e-05, "weight_min": 0.0, "weight_std": 2.130436769220978e-05 }, { "epoch": 0.575044621126075, "step": 1772, "weight_max": 3.602553260861896e-05, "weight_mean": 4.503671789279906e-06, "weight_min": 0.0, "weight_std": 1.273675479751546e-05 }, { "epoch": 0.575044621126075, "step": 1772, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.578696901560761e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.1792744519189e-05 }, { "epoch": 0.575044621126075, "step": 1772, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4967862625780981e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.0712590412585996e-05 }, { "epoch": 0.575044621126075, "step": 1772, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5090818124008365e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0905947167193517e-05 }, { "epoch": 0.5753691384066202, "step": 1773, "weight_max": 3.967146767536178e-05, "weight_mean": 9.30557416722877e-06, "weight_min": 0.0, "weight_std": 1.7055928765330464e-05 }, { "epoch": 0.5753691384066202, "step": 1773, "weight_max": 3.4268457966391e-05, "weight_mean": 1.1533350516401697e-05, "weight_min": 0.0, "weight_std": 1.6001589756342582e-05 }, { "epoch": 0.5753691384066202, "step": 1773, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5501716916332953e-05, "weight_min": 7.091466858177853e-10, "weight_std": 1.9517559849191457e-05 }, { "epoch": 0.5753691384066202, "step": 1773, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.436898674000986e-05, "weight_min": 0.0, "weight_std": 1.9969358618254773e-05 }, { "epoch": 0.5756936556871654, "step": 1774, "weight_max": 4.187939703115262e-05, "weight_mean": 1.5467307093786076e-05, "weight_min": 5.864964625068581e-10, "weight_std": 2.1347896108636633e-05 }, { "epoch": 0.5756936556871654, "step": 1774, "weight_max": 4.187939703115262e-05, "weight_mean": 5.295945356920129e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.4782907783228438e-05 }, { "epoch": 0.5756936556871654, "step": 1774, "weight_max": 4.252122016623616e-05, "weight_mean": 1.328853795712348e-05, "weight_min": 6.29309993005478e-10, "weight_std": 1.8800636098603718e-05 }, { "epoch": 0.5756936556871654, "step": 1774, "weight_max": 3.632264997577295e-05, "weight_mean": 1.3328270142665133e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.8393711798125878e-05 }, { "epoch": 0.5760181729677105, "step": 1775, "weight_max": 3.782653220696375e-05, "weight_mean": 8.386936315218918e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.5694924513809383e-05 }, { "epoch": 0.5760181729677105, "step": 1775, "weight_max": 3.632264997577295e-05, "weight_mean": 4.540812369668856e-06, "weight_min": 0.0, "weight_std": 1.2841802345064934e-05 }, { "epoch": 0.5760181729677105, "step": 1775, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.0075164937006775e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.865395825006999e-05 }, { "epoch": 0.5760181729677105, "step": 1775, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.996927474276163e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.8512462702346966e-05 }, { "epoch": 0.5763426902482557, "step": 1776, "weight_max": 4.187939703115262e-05, "weight_mean": 2.0038609363837168e-05, "weight_min": 6.196655966128617e-10, "weight_std": 2.1448366169352084e-05 }, { "epoch": 0.5763426902482557, "step": 1776, "weight_max": 3.572963396436535e-05, "weight_mean": 5.25995483258157e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.2505879567470402e-05 }, { "epoch": 0.5763426902482557, "step": 1776, "weight_max": 1.6735920427279893e-09, "weight_mean": 8.774044824022553e-10, "weight_min": 0.0, "weight_std": 5.348910203650803e-10 }, { "epoch": 0.5763426902482557, "step": 1776, "weight_max": 4.187939703115262e-05, "weight_mean": 9.888157364912331e-06, "weight_min": 0.0, "weight_std": 1.8350187019677833e-05 }, { "epoch": 0.576667207528801, "step": 1777, "weight_max": 3.006794759130571e-05, "weight_mean": 3.758919319807319e-06, "weight_min": 0.0, "weight_std": 1.063045328919543e-05 }, { "epoch": 0.576667207528801, "step": 1777, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.9943428924307227e-05, "weight_min": 0.0, "weight_std": 2.1441901481011882e-05 }, { "epoch": 0.576667207528801, "step": 1777, "weight_max": 2.639517333591357e-05, "weight_mean": 3.3003750559146283e-06, "weight_min": 5.771870759119224e-10, "weight_std": 9.331708497484215e-06 }, { "epoch": 0.576667207528801, "step": 1777, "weight_max": 3.998322426923551e-05, "weight_mean": 6.579999535460956e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.420429180143401e-05 }, { "epoch": 0.5769917248093461, "step": 1778, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.7255033475958044e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.0534750799706671e-05 }, { "epoch": 0.5769917248093461, "step": 1778, "weight_max": 4.252122016623616e-05, "weight_mean": 1.574886846356094e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.1642492356477305e-05 }, { "epoch": 0.5769917248093461, "step": 1778, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4323590221465565e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9869337847921997e-05 }, { "epoch": 0.5769917248093461, "step": 1778, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.4922675291018095e-05, "weight_min": 0.0, "weight_std": 2.0621260773623362e-05 }, { "epoch": 0.5773162420898913, "step": 1779, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.291804417443927e-06, "weight_min": 0.0, "weight_std": 1.7294471035711467e-05 }, { "epoch": 0.5773162420898913, "step": 1779, "weight_max": 3.752332122530788e-05, "weight_mean": 1.351378159597516e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.866675120254513e-05 }, { "epoch": 0.5773162420898913, "step": 1779, "weight_max": 3.199401908204891e-05, "weight_mean": 7.792126780259423e-06, "weight_min": 0.0, "weight_std": 1.4433894648391288e-05 }, { "epoch": 0.5773162420898913, "step": 1779, "weight_max": 4.252122016623616e-05, "weight_mean": 1.5080485354701523e-05, "weight_min": 0.0, "weight_std": 2.0844028767896816e-05 }, { "epoch": 0.5776407593704365, "grad_norm": 2.315532684326172, "learning_rate": 6.5795918367346945e-06, "loss": 0.1114, "step": 1780 }, { "epoch": 0.5776407593704365, "step": 1780, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.572550879966002e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.772793439158704e-05 }, { "epoch": 0.5776407593704365, "step": 1780, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0201320037594996e-05, "weight_min": 0.0, "weight_std": 1.8929600628325716e-05 }, { "epoch": 0.5776407593704365, "step": 1780, "weight_max": 4.187939703115262e-05, "weight_mean": 1.4243446457840037e-05, "weight_min": 5.864964625068581e-10, "weight_std": 1.9744546079891734e-05 }, { "epoch": 0.5776407593704365, "step": 1780, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.532612441224046e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.8386728697805665e-05 }, { "epoch": 0.5779652766509816, "step": 1781, "weight_max": 3.782653220696375e-05, "weight_mean": 8.123971383611206e-06, "weight_min": 0.0, "weight_std": 1.5309347872971557e-05 }, { "epoch": 0.5779652766509816, "step": 1781, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.1565612011472695e-06, "weight_min": 0.0, "weight_std": 1.4580905371985864e-05 }, { "epoch": 0.5779652766509816, "step": 1781, "weight_max": 4.1242459701607004e-05, "weight_mean": 5.380376933317166e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.450407307856949e-05 }, { "epoch": 0.5779652766509816, "step": 1781, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0286918950441759e-05, "weight_min": 7.829137893544669e-10, "weight_std": 1.6256766684819013e-05 }, { "epoch": 0.5782897939315269, "step": 1782, "weight_max": 4.187939703115262e-05, "weight_mean": 1.3033495633862913e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.83152351382887e-05 }, { "epoch": 0.5782897939315269, "step": 1782, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235580374574056e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.480633727624081e-05 }, { "epoch": 0.5782897939315269, "step": 1782, "weight_max": 4.187939703115262e-05, "weight_mean": 1.28121337183984e-05, "weight_min": 0.0, "weight_std": 1.8634369553183205e-05 }, { "epoch": 0.5782897939315269, "step": 1782, "weight_max": 4.187939703115262e-05, "weight_mean": 1.0117910278495401e-05, "weight_min": 6.786494144428445e-10, "weight_std": 1.8746181012829766e-05 }, { "epoch": 0.5786143112120721, "step": 1783, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4701588952448219e-05, "weight_min": 6.053385570581327e-10, "weight_std": 2.052214040304534e-05 }, { "epoch": 0.5786143112120721, "step": 1783, "weight_max": 2.2169586166853605e-09, "weight_mean": 7.611090091508288e-10, "weight_min": 0.0, "weight_std": 7.023051029619864e-10 }, { "epoch": 0.5786143112120721, "step": 1783, "weight_max": 3.602553260861896e-05, "weight_mean": 1.198804056912195e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.665375202719588e-05 }, { "epoch": 0.5786143112120721, "step": 1783, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.075202064588666e-05, "weight_min": 6.390288853630466e-10, "weight_std": 2.2207854271982796e-05 }, { "epoch": 0.5789388284926172, "step": 1784, "weight_max": 4.187939703115262e-05, "weight_mean": 5.983432856737636e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.4654239748779219e-05 }, { "epoch": 0.5789388284926172, "step": 1784, "weight_max": 3.967146767536178e-05, "weight_mean": 9.650158062868286e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.787584369594697e-05 }, { "epoch": 0.5789388284926172, "step": 1784, "weight_max": 4.381950566312298e-05, "weight_mean": 1.0321011359337717e-05, "weight_min": 0.0, "weight_std": 1.9157327187713236e-05 }, { "epoch": 0.5789388284926172, "step": 1784, "weight_max": 2.3459858500274322e-09, "weight_mean": 1.2885634781412136e-09, "weight_min": 6.390288853630466e-10, "weight_std": 6.985120815095058e-10 }, { "epoch": 0.5792633457731624, "step": 1785, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.425912139704451e-06, "weight_min": 0.0, "weight_std": 1.7216367268702015e-05 }, { "epoch": 0.5792633457731624, "step": 1785, "weight_max": 4.0610404539620504e-05, "weight_mean": 9.617222531232983e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.7843014575191773e-05 }, { "epoch": 0.5792633457731624, "step": 1785, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.0311749065294862e-05, "weight_min": 0.0, "weight_std": 1.9090848581981845e-05 }, { "epoch": 0.5792633457731624, "step": 1785, "weight_max": 1.3544732610171195e-05, "weight_mean": 3.1043391572893597e-06, "weight_min": 6.29309993005478e-10, "weight_std": 5.778043032478308e-06 }, { "epoch": 0.5795878630537076, "step": 1786, "weight_max": 3.8743506593164057e-05, "weight_mean": 1.3816200407745782e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9099959899904206e-05 }, { "epoch": 0.5795878630537076, "step": 1786, "weight_max": 4.0610404539620504e-05, "weight_mean": 8.801646799838636e-06, "weight_min": 0.0, "weight_std": 1.6550160580663942e-05 }, { "epoch": 0.5795878630537076, "step": 1786, "weight_max": 3.967146767536178e-05, "weight_mean": 8.99375845619943e-06, "weight_min": 6.586902134841921e-10, "weight_std": 1.676875945122447e-05 }, { "epoch": 0.5795878630537076, "step": 1786, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3206527000875212e-05, "weight_min": 0.0, "weight_std": 1.833273927331902e-05 }, { "epoch": 0.5799123803342529, "step": 1787, "weight_max": 3.4268457966391e-05, "weight_mean": 4.284249371266924e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.2115449862903915e-05 }, { "epoch": 0.5799123803342529, "step": 1787, "weight_max": 4.187939703115262e-05, "weight_mean": 9.70308065006975e-06, "weight_min": 7.194614903838215e-10, "weight_std": 1.8036836991086602e-05 }, { "epoch": 0.5799123803342529, "step": 1787, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3898779798182659e-05, "weight_min": 6.100955851628953e-10, "weight_std": 1.983540460059885e-05 }, { "epoch": 0.5799123803342529, "step": 1787, "weight_max": 3.116124207735993e-05, "weight_mean": 3.8965840758464765e-06, "weight_min": 0.0, "weight_std": 1.1016585631296039e-05 }, { "epoch": 0.580236897614798, "step": 1788, "weight_max": 4.1242459701607004e-05, "weight_mean": 9.051159395312425e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.697298648650758e-05 }, { "epoch": 0.580236897614798, "step": 1788, "weight_max": 4.187939703115262e-05, "weight_mean": 5.23561311638332e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.4806325452809688e-05 }, { "epoch": 0.580236897614798, "step": 1788, "weight_max": 2.4894337911973707e-05, "weight_mean": 5.5286468523263466e-06, "weight_min": 6.053385570581327e-10, "weight_std": 1.03431120805908e-05 }, { "epoch": 0.580236897614798, "step": 1788, "weight_max": 3.8743506593164057e-05, "weight_mean": 9.421519280294888e-06, "weight_min": 6.390288853630466e-10, "weight_std": 1.7452122847316787e-05 }, { "epoch": 0.5805614148953432, "step": 1789, "weight_max": 3.061215829802677e-05, "weight_mean": 3.827333330264082e-06, "weight_min": 5.864964625068581e-10, "weight_std": 1.0822703188750893e-05 }, { "epoch": 0.5805614148953432, "step": 1789, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0316594853065908e-05, "weight_min": 6.053385570581327e-10, "weight_std": 1.912849438667763e-05 }, { "epoch": 0.5805614148953432, "step": 1789, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.45503200253006e-06, "weight_min": 0.0, "weight_std": 1.614994471310638e-05 }, { "epoch": 0.5805614148953432, "step": 1789, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.1446354619693011e-05, "weight_min": 0.0, "weight_std": 1.9313565644552e-05 }, { "epoch": 0.5808859321758884, "grad_norm": 0.20603279769420624, "learning_rate": 6.559183673469388e-06, "loss": 0.0711, "step": 1790 }, { "epoch": 0.5808859321758884, "step": 1790, "weight_max": 4.187939703115262e-05, "weight_mean": 5.2361388043209445e-06, "weight_min": 0.0, "weight_std": 1.4806111721554771e-05 }, { "epoch": 0.5808859321758884, "step": 1790, "weight_max": 4.252122016623616e-05, "weight_mean": 1.3401283467828762e-05, "weight_min": 0.0, "weight_std": 1.8962657122756355e-05 }, { "epoch": 0.5808859321758884, "step": 1790, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235601292952197e-06, "weight_min": 0.0, "weight_std": 1.4806330000283197e-05 }, { "epoch": 0.5808859321758884, "step": 1790, "weight_max": 3.998322426923551e-05, "weight_mean": 9.765540198714007e-06, "weight_min": 6.29309993005478e-10, "weight_std": 1.808591878216248e-05 }, { "epoch": 0.5812104494564335, "step": 1791, "weight_max": 4.187939703115262e-05, "weight_mean": 9.06226341612637e-06, "weight_min": 6.196655966128617e-10, "weight_std": 1.704611895547714e-05 }, { "epoch": 0.5812104494564335, "step": 1791, "weight_max": 4.1242459701607004e-05, "weight_mean": 8.273326784546953e-06, "weight_min": 0.0, "weight_std": 1.5915342373773456e-05 }, { "epoch": 0.5812104494564335, "step": 1791, "weight_max": 3.8743506593164057e-05, "weight_mean": 4.844003342441283e-06, "weight_min": 0.0, "weight_std": 1.3697466783924028e-05 }, { "epoch": 0.5812104494564335, "step": 1791, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.393737981445156e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2423974112607539e-05 }, { "epoch": 0.5815349667369788, "step": 1792, "weight_max": 3.782653220696375e-05, "weight_mean": 8.658594197186176e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.612185951671563e-05 }, { "epoch": 0.5815349667369788, "step": 1792, "weight_max": 3.843663216684945e-05, "weight_mean": 9.45782994676847e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.7514064893475734e-05 }, { "epoch": 0.5815349667369788, "step": 1792, "weight_max": 4.0610404539620504e-05, "weight_mean": 1.3283981388667598e-05, "weight_min": 5.958802340444436e-10, "weight_std": 1.84612836164888e-05 }, { "epoch": 0.5815349667369788, "step": 1792, "weight_max": 3.572963396436535e-05, "weight_mean": 8.293376595247537e-06, "weight_min": 6.100955851628953e-10, "weight_std": 1.5415531379403546e-05 }, { "epoch": 0.581859484017524, "step": 1793, "weight_max": 3.6920544516760856e-05, "weight_mean": 8.140857062244322e-06, "weight_min": 0.0, "weight_std": 1.5251566765073221e-05 }, { "epoch": 0.581859484017524, "step": 1793, "weight_max": 3.967146767536178e-05, "weight_mean": 1.6378735381294973e-05, "weight_min": 0.0, "weight_std": 1.692592923063785e-05 }, { "epoch": 0.581859484017524, "step": 1793, "weight_max": 3.782653220696375e-05, "weight_mean": 1.5704668840044178e-05, "weight_min": 6.989064882390039e-10, "weight_std": 1.7828424461185932e-05 }, { "epoch": 0.581859484017524, "step": 1793, "weight_max": 2.872879122151062e-05, "weight_mean": 6.612260222027544e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.2301522474444937e-05 }, { "epoch": 0.5821840012980691, "step": 1794, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.5151845218497328e-05, "weight_min": 5.958802340444436e-10, "weight_std": 2.0913528715027496e-05 }, { "epoch": 0.5821840012980691, "step": 1794, "weight_max": 4.187939703115262e-05, "weight_mean": 5.235712251305813e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.4806284525548108e-05 }, { "epoch": 0.5821840012980691, "step": 1794, "weight_max": 2.9797680326737463e-05, "weight_mean": 3.725704345924896e-06, "weight_min": 0.0, "weight_std": 1.0534669854678214e-05 }, { "epoch": 0.5821840012980691, "step": 1794, "weight_max": 4.3167921830900013e-05, "weight_mean": 2.239395871583838e-05, "weight_min": 6.196655966128617e-10, "weight_std": 1.9740924471989274e-05 }, { "epoch": 0.5825085185786143, "step": 1795, "weight_max": 7.653039574506693e-06, "weight_mean": 9.576178854331374e-07, "weight_min": 6.100955851628953e-10, "weight_std": 2.7053588382841554e-06 }, { "epoch": 0.5825085185786143, "step": 1795, "weight_max": 4.1242459701607004e-05, "weight_mean": 1.2935449376527686e-05, "weight_min": 5.679522963042416e-10, "weight_std": 1.8215994714410044e-05 }, { "epoch": 0.5825085185786143, "step": 1795, "weight_max": 4.187939703115262e-05, "weight_mean": 9.611700079403818e-06, "weight_min": 5.771870759119224e-10, "weight_std": 1.6720843632356264e-05 }, { "epoch": 0.5825085185786143, "step": 1795, "weight_max": 4.252122016623616e-05, "weight_mean": 5.315955604601186e-06, "weight_min": 6.488223291967188e-10, "weight_std": 1.5033197087177541e-05 }, { "epoch": 0.5828330358591595, "step": 1796, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.4627971722802613e-05, "weight_min": 0.0, "weight_std": 2.0284880520193838e-05 }, { "epoch": 0.5828330358591595, "step": 1796, "weight_max": 4.187939703115262e-05, "weight_mean": 5.236940978647908e-06, "weight_min": 0.0, "weight_std": 1.4805788850935642e-05 }, { "epoch": 0.5828330358591595, "step": 1796, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.883152087131748e-06, "weight_min": 5.958802340444436e-10, "weight_std": 1.3806141396344174e-05 }, { "epoch": 0.5828330358591595, "step": 1796, "weight_max": 3.6920544516760856e-05, "weight_mean": 1.1569142770895269e-05, "weight_min": 5.771870759119224e-10, "weight_std": 1.6214537026826292e-05 }, { "epoch": 0.5831575531397047, "step": 1797, "weight_max": 3.9051607018336654e-05, "weight_mean": 4.882191660726676e-06, "weight_min": 0.0, "weight_std": 1.3806528841087129e-05 }, { "epoch": 0.5831575531397047, "step": 1797, "weight_max": 1.7859370649020434e-09, "weight_mean": 7.262456191980959e-10, "weight_min": 0.0, "weight_std": 4.929929797725663e-10 }, { "epoch": 0.5831575531397047, "step": 1797, "weight_max": 3.9051607018336654e-05, "weight_mean": 9.728810255182907e-06, "weight_min": 0.0, "weight_std": 1.8003531295107678e-05 }, { "epoch": 0.5831575531397047, "step": 1797, "weight_max": 3.782653220696375e-05, "weight_mean": 1.3848515663994476e-05, "weight_min": 0.0, "weight_std": 1.9124521713820286e-05 }, { "epoch": 0.5834820704202499, "step": 1798, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.6877562302397564e-05, "weight_min": 0.0, "weight_std": 1.9286682800156996e-05 }, { "epoch": 0.5834820704202499, "step": 1798, "weight_max": 1.0890587418543873e-06, "weight_mean": 1.368537141388515e-07, "weight_min": 5.958802340444436e-10, "weight_std": 3.8474900065921247e-07 }, { "epoch": 0.5834820704202499, "step": 1798, "weight_max": 3.5141503758495674e-05, "weight_mean": 4.3933519009442534e-06, "weight_min": 5.679522963042416e-10, "weight_std": 1.2424129636201542e-05 }, { "epoch": 0.5834820704202499, "step": 1798, "weight_max": 4.187939703115262e-05, "weight_mean": 1.439193329133559e-05, "weight_min": 0.0, "weight_std": 1.993370460695587e-05 }, { "epoch": 0.5838065877007951, "step": 1799, "weight_max": 2.3835209361777743e-09, "weight_mean": 1.0710476949782333e-09, "weight_min": 5.771870759119224e-10, "weight_std": 6.115663531147675e-10 }, { "epoch": 0.5838065877007951, "step": 1799, "weight_max": 3.9360926166409627e-05, "weight_mean": 1.3363113794184756e-05, "weight_min": 7.298507354036587e-10, "weight_std": 1.8596971131046303e-05 }, { "epoch": 0.5838065877007951, "step": 1799, "weight_max": 4.772478412462533e-09, "weight_mean": 1.5348367021772447e-09, "weight_min": 6.100955851628953e-10, "weight_std": 1.67347169455212e-09 }, { "epoch": 0.5838065877007951, "step": 1799, "weight_max": 4.3167921830900013e-05, "weight_mean": 1.0355441190768033e-05, "weight_min": 0.0, "weight_std": 1.9196031644241884e-05 }, { "epoch": 0.5841311049813402, "grad_norm": 10.022843360900879, "learning_rate": 6.538775510204082e-06, "loss": 0.1773, "step": 1800 }, { "epoch": 0.5841311049813402, "eval_accuracy": 0.991, "eval_auc": 0.9962809644588821, "eval_loss": 0.04703560471534729, "eval_runtime": 156.9644, "eval_samples_per_second": 12.742, "eval_steps_per_second": 1.593, "step": 1800 } ], "logging_steps": 10, "max_steps": 5000, "num_input_tokens_seen": 0, "num_train_epochs": 2, "save_steps": 100, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 1.066481202774649e+17, "train_batch_size": 8, "trial_name": null, "trial_params": null }