resmlp_comparison / results.json
AmberLJC's picture
Upload results.json with huggingface_hub
9ec0e35 verified
{
"config": {
"num_layers": 20,
"hidden_dim": 64,
"num_samples": 1024,
"training_steps": 500,
"learning_rate": 0.001,
"batch_size": 64
},
"plain_mlp": {
"final_loss": 0.3122814893722534,
"initial_loss": 0.42214322090148926,
"loss_history": [
0.42214322090148926,
0.36282670497894287,
0.3482760190963745,
0.3494352102279663,
0.33545294404029846,
0.3272090554237366,
0.33767467737197876,
0.3396916091442108,
0.3351821303367615,
0.3310932219028473,
0.3351791501045227,
0.34008604288101196,
0.34631210565567017,
0.33455097675323486,
0.33241066336631775,
0.327256441116333,
0.3311437964439392,
0.3305869996547699,
0.3346586227416992,
0.340005487203598,
0.33456265926361084,
0.33543068170547485,
0.3316764235496521,
0.342144250869751,
0.3264315724372864,
0.32780060172080994,
0.3364536166191101,
0.3336724638938904,
0.33728402853012085,
0.33577609062194824,
0.33049479126930237,
0.329383909702301,
0.3264477550983429,
0.3307558298110962,
0.32680246233940125,
0.333177387714386,
0.33437252044677734,
0.32979515194892883,
0.33821630477905273,
0.3267170190811157,
0.3301219344139099,
0.3348636031150818,
0.3265611529350281,
0.32980096340179443,
0.33458641171455383,
0.3295944631099701,
0.32541388273239136,
0.33195948600769043,
0.33705854415893555,
0.32626599073410034,
0.32933109998703003,
0.3315778374671936,
0.3286918103694916,
0.335918128490448,
0.3317474126815796,
0.3240748643875122,
0.3326587975025177,
0.32749223709106445,
0.33983999490737915,
0.3330131769180298,
0.3279249370098114,
0.328784316778183,
0.3272261619567871,
0.32868239283561707,
0.34050124883651733,
0.3372110426425934,
0.33211958408355713,
0.3379274010658264,
0.3354094922542572,
0.337791383266449,
0.3335695266723633,
0.33175790309906006,
0.3285711705684662,
0.3294464349746704,
0.3280278742313385,
0.33241695165634155,
0.33453792333602905,
0.32833805680274963,
0.3319797217845917,
0.3291819393634796,
0.3294021785259247,
0.3406433165073395,
0.32606297731399536,
0.3226204812526703,
0.3310944437980652,
0.3309115469455719,
0.3352738618850708,
0.33406805992126465,
0.33366382122039795,
0.34170109033584595,
0.32491233944892883,
0.3298508822917938,
0.3299129009246826,
0.33908554911613464,
0.33246976137161255,
0.33066731691360474,
0.3304769694805145,
0.32929137349128723,
0.32822930812835693,
0.3333491086959839,
0.32887813448905945,
0.3329111337661743,
0.33009040355682373,
0.3283909261226654,
0.32943427562713623,
0.33743613958358765,
0.32562410831451416,
0.3262087106704712,
0.3281552195549011,
0.32880309224128723,
0.33024877309799194,
0.33133694529533386,
0.3343980014324188,
0.3297366499900818,
0.33445966243743896,
0.3308635652065277,
0.33346372842788696,
0.3338125944137573,
0.33411431312561035,
0.32354146242141724,
0.3296177089214325,
0.3204713761806488,
0.3324952721595764,
0.32344335317611694,
0.330871045589447,
0.32716238498687744,
0.337432861328125,
0.3403051495552063,
0.3313557505607605,
0.32797929644584656,
0.3364092707633972,
0.3307040333747864,
0.32038557529449463,
0.31208786368370056,
0.3315786123275757,
0.33295413851737976,
0.33243390917778015,
0.3210064172744751,
0.32867860794067383,
0.3323233425617218,
0.333433598279953,
0.326099157333374,
0.33212804794311523,
0.3348027765750885,
0.32956549525260925,
0.3299698233604431,
0.3317911624908447,
0.33178210258483887,
0.3308490216732025,
0.3286418914794922,
0.32811760902404785,
0.326610267162323,
0.3255402743816376,
0.33278921246528625,
0.3288092315196991,
0.333260178565979,
0.32541632652282715,
0.325713574886322,
0.3272123336791992,
0.3295033872127533,
0.3211234211921692,
0.33132150769233704,
0.32147130370140076,
0.33328524231910706,
0.3304368853569031,
0.32660651206970215,
0.32244348526000977,
0.3221438229084015,
0.32539036870002747,
0.3355017304420471,
0.33238741755485535,
0.31443461775779724,
0.32446616888046265,
0.3186056911945343,
0.3275202512741089,
0.32419508695602417,
0.3279516100883484,
0.3259829878807068,
0.32992708683013916,
0.32248663902282715,
0.3220922350883484,
0.3303144872188568,
0.332166850566864,
0.32141345739364624,
0.31897056102752686,
0.32363876700401306,
0.3221549689769745,
0.33142802119255066,
0.3282598853111267,
0.31797027587890625,
0.33446645736694336,
0.33268487453460693,
0.3175439238548279,
0.33158648014068604,
0.324335515499115,
0.3174867630004883,
0.32891330122947693,
0.3220325708389282,
0.3250204026699066,
0.32564860582351685,
0.32088494300842285,
0.32883039116859436,
0.32697242498397827,
0.33850207924842834,
0.3279131054878235,
0.3202984929084778,
0.3261798620223999,
0.3341549038887024,
0.3264288604259491,
0.32558804750442505,
0.3242926001548767,
0.3234439492225647,
0.31936201453208923,
0.32065972685813904,
0.3229916989803314,
0.3290161192417145,
0.31866925954818726,
0.3291019797325134,
0.3215341567993164,
0.3218863010406494,
0.3229762017726898,
0.3182288408279419,
0.3278171420097351,
0.31954365968704224,
0.32973426580429077,
0.3156335949897766,
0.3162286579608917,
0.32882851362228394,
0.3199450373649597,
0.32140082120895386,
0.3241465985774994,
0.3269677758216858,
0.3188479244709015,
0.31919217109680176,
0.3209821581840515,
0.3129628300666809,
0.3245271146297455,
0.3315494656562805,
0.32561102509498596,
0.32373708486557007,
0.3241332769393921,
0.3227849304676056,
0.3221299648284912,
0.3213585615158081,
0.3290410339832306,
0.3342036008834839,
0.3285410404205322,
0.3231137990951538,
0.31683844327926636,
0.3211103677749634,
0.33177798986434937,
0.32102054357528687,
0.3156767189502716,
0.32156461477279663,
0.3223722577095032,
0.3139723539352417,
0.32709231972694397,
0.3173341751098633,
0.31532400846481323,
0.3170064687728882,
0.31997907161712646,
0.32745033502578735,
0.32697030901908875,
0.3204156756401062,
0.3271084129810333,
0.32671791315078735,
0.3246278762817383,
0.3164779245853424,
0.3271852433681488,
0.3196042776107788,
0.32026350498199463,
0.3120955228805542,
0.318584680557251,
0.32811853289604187,
0.32301777601242065,
0.32394924759864807,
0.3186284899711609,
0.33235689997673035,
0.3238418698310852,
0.32303276658058167,
0.3165513277053833,
0.3227938115596771,
0.3216084837913513,
0.32230186462402344,
0.31495195627212524,
0.322005957365036,
0.3129628300666809,
0.316558837890625,
0.3285377323627472,
0.32414621114730835,
0.31914806365966797,
0.3209497332572937,
0.31563228368759155,
0.315194308757782,
0.32207655906677246,
0.3177405595779419,
0.31551986932754517,
0.31802260875701904,
0.32021886110305786,
0.3208439350128174,
0.31313204765319824,
0.3136066198348999,
0.31712469458580017,
0.31795984506607056,
0.3181302547454834,
0.322174072265625,
0.32505011558532715,
0.31750059127807617,
0.3222911059856415,
0.3154448866844177,
0.3248714208602905,
0.3258054554462433,
0.3222883343696594,
0.3219612240791321,
0.30858346819877625,
0.30579590797424316,
0.31194090843200684,
0.30980226397514343,
0.3115063011646271,
0.3166201710700989,
0.32646316289901733,
0.32494890689849854,
0.3148566484451294,
0.31464236974716187,
0.32489898800849915,
0.3286982774734497,
0.31258541345596313,
0.32561126351356506,
0.32239222526550293,
0.31925585865974426,
0.31524786353111267,
0.3155617117881775,
0.30894869565963745,
0.31647616624832153,
0.3181927800178528,
0.3155575394630432,
0.31147250533103943,
0.3121526539325714,
0.3178417980670929,
0.31777507066726685,
0.3277454078197479,
0.3230074644088745,
0.3155066967010498,
0.3123127222061157,
0.321414977312088,
0.31344032287597656,
0.31667494773864746,
0.32026779651641846,
0.32438141107559204,
0.3171403110027313,
0.3197984993457794,
0.32138901948928833,
0.3184692859649658,
0.31916412711143494,
0.3180121183395386,
0.32657110691070557,
0.31782758235931396,
0.3241843581199646,
0.31732508540153503,
0.317064106464386,
0.3212159276008606,
0.31150341033935547,
0.3197249472141266,
0.31177613139152527,
0.3193328082561493,
0.3192586302757263,
0.3233141303062439,
0.3107929229736328,
0.3158963918685913,
0.32218313217163086,
0.3179759979248047,
0.3117298483848572,
0.3206039369106293,
0.31486546993255615,
0.3219609260559082,
0.30939263105392456,
0.3182328939437866,
0.3162742853164673,
0.32427945733070374,
0.3155043125152588,
0.31323426961898804,
0.3223169147968292,
0.33166787028312683,
0.3266363739967346,
0.318408727645874,
0.3162022829055786,
0.316178560256958,
0.3218512237071991,
0.32263171672821045,
0.3201322853565216,
0.3166625499725342,
0.3264351487159729,
0.3150608539581299,
0.3099004626274109,
0.311903178691864,
0.315491259098053,
0.3139011859893799,
0.3073548674583435,
0.3136225640773773,
0.3125903010368347,
0.3150739371776581,
0.31659185886383057,
0.32262271642684937,
0.3133242130279541,
0.31148239970207214,
0.32298412919044495,
0.31778499484062195,
0.30843785405158997,
0.3154827952384949,
0.31984445452690125,
0.3228866457939148,
0.31940460205078125,
0.315796822309494,
0.31098097562789917,
0.31478509306907654,
0.3137901723384857,
0.3185809850692749,
0.3246801793575287,
0.30861347913742065,
0.3175472021102905,
0.3125755190849304,
0.317693829536438,
0.31665486097335815,
0.31839433312416077,
0.31122976541519165,
0.3198891580104828,
0.3025595247745514,
0.31865018606185913,
0.31307554244995117,
0.310791939496994,
0.3189123570919037,
0.31337010860443115,
0.3129110038280487,
0.3109862804412842,
0.3105536699295044,
0.3100142478942871,
0.30681681632995605,
0.30979853868484497,
0.32108548283576965,
0.31359052658081055,
0.32146626710891724,
0.3074720501899719,
0.30723342299461365,
0.3087216317653656,
0.31279802322387695,
0.30798858404159546,
0.3053668141365051,
0.3111911118030548,
0.3119443953037262,
0.31192684173583984,
0.3080732822418213,
0.31404170393943787,
0.3138525187969208,
0.31473881006240845,
0.31580841541290283,
0.3151828646659851,
0.31342148780822754,
0.3052767515182495,
0.3119245171546936,
0.3106262683868408,
0.3038669228553772,
0.3183707296848297,
0.3203107714653015,
0.309036523103714,
0.31399810314178467,
0.31323930621147156,
0.306890606880188,
0.30548590421676636,
0.3077350854873657,
0.3146387040615082,
0.3109719455242157,
0.31250497698783875,
0.3004187345504761,
0.30847451090812683,
0.31020867824554443,
0.31392747163772583,
0.31991061568260193,
0.3089701533317566,
0.3094186782836914,
0.316642165184021,
0.3136940002441406,
0.31746378540992737,
0.3059760332107544,
0.30767661333084106,
0.31823939085006714,
0.3136705756187439,
0.31073373556137085,
0.31243982911109924,
0.31552764773368835,
0.3138889670372009,
0.30521848797798157,
0.2989374101161957,
0.3153800964355469,
0.31432241201400757,
0.31000563502311707,
0.3153458833694458,
0.31128472089767456,
0.310610294342041,
0.3141220808029175,
0.3122814893722534
],
"gradient_norms": [
0.010068393312394619,
0.009067831560969353,
0.008841422386467457,
0.008336398750543594,
0.007635441143065691,
0.007993700914084911,
0.008473974652588367,
0.008481012657284737,
0.00877982284873724,
0.009291221387684345,
0.009246684610843658,
0.010227610357105732,
0.009622696787118912,
0.008872134611010551,
0.009406423196196556,
0.009590999223291874,
0.010016469284892082,
0.010378911159932613,
0.010058541782200336,
0.009685155935585499
],
"activation_means": [
0.02235810086131096,
-0.0006043091416358948,
0.027520379051566124,
0.09677311778068542,
-0.09256714582443237,
-0.0832536593079567,
-0.03931272402405739,
-0.10891453176736832,
-0.1333763301372528,
-0.01697702705860138,
-0.24202276766300201,
-0.022496212273836136,
0.0013794712722301483,
-0.07350999861955643,
-0.09732113778591156,
-0.0586143396794796,
-0.09032783657312393,
-0.010199513286352158,
0.010252870619297028,
-0.1822284758090973
],
"activation_stds": [
0.8081836700439453,
0.8764128684997559,
0.9008104205131531,
0.9063840508460999,
0.9467869400978088,
0.9418801069259644,
0.8313923478126526,
0.7985774874687195,
0.843100905418396,
0.7745543718338013,
0.7544355392456055,
0.5160097479820251,
0.5802003741264343,
0.5734189748764038,
0.48425760865211487,
0.44677266478538513,
0.44068723917007446,
0.4156236946582794,
0.424014687538147,
0.35631605982780457
]
},
"res_mlp": {
"final_loss": 0.0629926323890686,
"initial_loss": 13.825733184814453,
"loss_history": [
13.825733184814453,
6.929441928863525,
4.056141376495361,
2.5541770458221436,
1.7824699878692627,
1.416144847869873,
1.1620547771453857,
0.9483454823493958,
0.7843164801597595,
0.697279155254364,
0.6453202366828918,
0.5573611855506897,
0.5255516171455383,
0.5015190839767456,
0.45454704761505127,
0.41940945386886597,
0.3973648250102997,
0.3921322822570801,
0.3478171229362488,
0.3742031157016754,
0.33467820286750793,
0.33183223009109497,
0.3318370580673218,
0.3350353240966797,
0.3137206435203552,
0.30033594369888306,
0.3070758879184723,
0.28984498977661133,
0.2851133346557617,
0.26924896240234375,
0.2905859053134918,
0.279144287109375,
0.2638997435569763,
0.2549370229244232,
0.24959874153137207,
0.2570638060569763,
0.2534509301185608,
0.25506916642189026,
0.24851691722869873,
0.23037107288837433,
0.24302563071250916,
0.22616733610630035,
0.24495507776737213,
0.23248232901096344,
0.24176537990570068,
0.2333303689956665,
0.22003066539764404,
0.245095893740654,
0.23193161189556122,
0.2175668478012085,
0.22363689541816711,
0.2120945304632187,
0.23737770318984985,
0.23073768615722656,
0.21438273787498474,
0.21955570578575134,
0.21080368757247925,
0.21987861394882202,
0.21614855527877808,
0.21649761497974396,
0.19733500480651855,
0.2083965241909027,
0.20562587678432465,
0.22173789143562317,
0.2059110701084137,
0.1945328712463379,
0.21662525832653046,
0.19985336065292358,
0.19271507859230042,
0.21030962467193604,
0.2069288194179535,
0.1867353767156601,
0.20643624663352966,
0.1913994699716568,
0.18623554706573486,
0.20420098304748535,
0.20161986351013184,
0.1983901560306549,
0.19239430129528046,
0.1878090798854828,
0.19025678932666779,
0.1812693476676941,
0.19511495530605316,
0.19247011840343475,
0.19345001876354218,
0.18723750114440918,
0.17752212285995483,
0.18105608224868774,
0.17320361733436584,
0.17695054411888123,
0.18027013540267944,
0.18198877573013306,
0.1806012988090515,
0.17939069867134094,
0.19096586108207703,
0.17916780710220337,
0.1848122477531433,
0.16437095403671265,
0.1749158799648285,
0.1854352056980133,
0.17872394621372223,
0.1727355718612671,
0.17188334465026855,
0.17672254145145416,
0.17822399735450745,
0.17592446506023407,
0.1675281673669815,
0.16010603308677673,
0.17089994251728058,
0.1601085066795349,
0.1725354939699173,
0.1574731022119522,
0.1598099172115326,
0.16953468322753906,
0.16409367322921753,
0.1595795452594757,
0.15564961731433868,
0.1542244702577591,
0.14573857188224792,
0.1609399914741516,
0.15494975447654724,
0.15977708995342255,
0.1628899723291397,
0.15699170529842377,
0.15361425280570984,
0.16150403022766113,
0.15789395570755005,
0.13668665289878845,
0.16025841236114502,
0.14876320958137512,
0.15640254318714142,
0.13928887248039246,
0.15516719222068787,
0.14445939660072327,
0.1433185338973999,
0.14129473268985748,
0.14215922355651855,
0.15496797859668732,
0.1444668471813202,
0.14378705620765686,
0.14088605344295502,
0.12942053377628326,
0.1482706069946289,
0.13720951974391937,
0.1429380476474762,
0.14560531079769135,
0.13521014153957367,
0.1422801911830902,
0.1372361034154892,
0.13660244643688202,
0.13699297606945038,
0.14125284552574158,
0.1396929919719696,
0.13849902153015137,
0.14266842603683472,
0.14512087404727936,
0.14272311329841614,
0.13200822472572327,
0.12528839707374573,
0.1362495869398117,
0.13740304112434387,
0.14004293084144592,
0.13355493545532227,
0.13202989101409912,
0.12513452768325806,
0.12320512533187866,
0.13207478821277618,
0.12799137830734253,
0.12362419813871384,
0.13280878961086273,
0.1266961544752121,
0.13285361230373383,
0.13146376609802246,
0.1272575557231903,
0.13236036896705627,
0.13058793544769287,
0.12451867759227753,
0.13076333701610565,
0.12537893652915955,
0.1286524534225464,
0.1255934089422226,
0.12199349701404572,
0.12703447043895721,
0.1205616444349289,
0.11638176441192627,
0.12982048094272614,
0.11749467253684998,
0.11279859393835068,
0.11866491287946701,
0.12125350534915924,
0.12608124315738678,
0.1334507018327713,
0.12499742209911346,
0.12244322896003723,
0.12269967049360275,
0.11862020939588547,
0.11574031412601471,
0.12193585187196732,
0.11600715667009354,
0.11964584141969681,
0.1162799671292305,
0.12539267539978027,
0.12248687446117401,
0.12070564925670624,
0.13140465319156647,
0.11640572547912598,
0.10755001753568649,
0.11134535819292068,
0.11165831983089447,
0.1097593680024147,
0.11635623127222061,
0.11294370889663696,
0.10825790464878082,
0.10611490160226822,
0.11304027587175369,
0.11056037247180939,
0.11218193918466568,
0.11071833968162537,
0.11433180421590805,
0.1080167144536972,
0.11098785698413849,
0.10625725984573364,
0.10640506446361542,
0.11352894455194473,
0.11021991074085236,
0.10663466155529022,
0.10082174837589264,
0.10823401808738708,
0.10888013243675232,
0.10405822843313217,
0.10449113696813583,
0.10663808882236481,
0.1043190062046051,
0.10356761515140533,
0.10988475382328033,
0.11231439560651779,
0.10832472890615463,
0.10288867354393005,
0.11017926037311554,
0.10268239676952362,
0.10657632350921631,
0.10329350084066391,
0.11197119206190109,
0.10926295816898346,
0.1126130148768425,
0.0998806282877922,
0.10103444755077362,
0.09290547668933868,
0.09799303859472275,
0.10779248923063278,
0.10077505558729172,
0.10144646465778351,
0.10275106132030487,
0.09488312900066376,
0.0980915054678917,
0.09439370036125183,
0.10310178995132446,
0.09724435210227966,
0.10685458779335022,
0.09623485803604126,
0.09753909707069397,
0.08963052928447723,
0.09267763793468475,
0.09678862243890762,
0.09540798515081406,
0.08834634721279144,
0.09270018339157104,
0.08261752128601074,
0.1081431582570076,
0.10075649619102478,
0.09674010425806046,
0.09663750231266022,
0.09644731134176254,
0.09872166812419891,
0.09951721131801605,
0.088838130235672,
0.09299120306968689,
0.09496532380580902,
0.09245827049016953,
0.09697187691926956,
0.0878148078918457,
0.09270801395177841,
0.09216231107711792,
0.09456576406955719,
0.08298899233341217,
0.09676386415958405,
0.09488122165203094,
0.09232550114393234,
0.09934262931346893,
0.08506445586681366,
0.09329526126384735,
0.09657076001167297,
0.09500131011009216,
0.09519918262958527,
0.09765627980232239,
0.08732327073812485,
0.09301722794771194,
0.09543395042419434,
0.0873609408736229,
0.08974739909172058,
0.08848126232624054,
0.08153203874826431,
0.0837751030921936,
0.09327980130910873,
0.0937112420797348,
0.0836104080080986,
0.08840271085500717,
0.08613412827253342,
0.08702630549669266,
0.0854334831237793,
0.07984519749879837,
0.0783020630478859,
0.0839296504855156,
0.0771411657333374,
0.07718954980373383,
0.08612189441919327,
0.08547349274158478,
0.08440866321325302,
0.0895165503025055,
0.08419612795114517,
0.08556754887104034,
0.08557409048080444,
0.08724275976419449,
0.08401279896497726,
0.08161499351263046,
0.08135215193033218,
0.07858897000551224,
0.07412116229534149,
0.09073606133460999,
0.0884423777461052,
0.07584438472986221,
0.08999098837375641,
0.07833802700042725,
0.08642788231372833,
0.08511155843734741,
0.07571419328451157,
0.07647889107465744,
0.0770968496799469,
0.07909407466650009,
0.07677635550498962,
0.07856360077857971,
0.07897358387708664,
0.08044072985649109,
0.0849953144788742,
0.07985112816095352,
0.07496027648448944,
0.0693586990237236,
0.077299565076828,
0.07525025308132172,
0.07886706292629242,
0.07089076936244965,
0.0745236724615097,
0.08298902958631516,
0.07434681057929993,
0.07863019406795502,
0.07296870648860931,
0.07038713246583939,
0.07498430460691452,
0.0731402337551117,
0.0791475921869278,
0.07742582261562347,
0.07537870854139328,
0.0774664357304573,
0.08444315940141678,
0.06756129115819931,
0.07491543889045715,
0.07476594299077988,
0.07262799143791199,
0.0735686719417572,
0.07892186939716339,
0.07961255311965942,
0.07117517292499542,
0.06840993463993073,
0.06531023234128952,
0.07480637729167938,
0.07482421398162842,
0.07697328180074692,
0.08092670142650604,
0.07015441358089447,
0.07309134304523468,
0.07465572655200958,
0.07515942305326462,
0.07721630483865738,
0.07378698885440826,
0.0710270032286644,
0.07413457334041595,
0.07682187855243683,
0.08446906507015228,
0.07762646675109863,
0.07136626541614532,
0.07230590283870697,
0.07373417913913727,
0.07236794382333755,
0.07123907655477524,
0.08195364475250244,
0.06900320947170258,
0.060011740773916245,
0.07414371520280838,
0.07555309683084488,
0.0644296333193779,
0.07632855325937271,
0.07520553469657898,
0.07276205718517303,
0.07271023094654083,
0.06531301140785217,
0.07297447323799133,
0.07270961999893188,
0.07454632222652435,
0.07276542484760284,
0.06848332285881042,
0.06939152628183365,
0.07199845463037491,
0.07199475914239883,
0.06776336580514908,
0.06876210123300552,
0.07206545770168304,
0.06846804171800613,
0.07429170608520508,
0.0714627206325531,
0.07270139455795288,
0.07588784396648407,
0.07119745016098022,
0.06948363780975342,
0.06362828612327576,
0.07597526907920837,
0.07059313356876373,
0.0640760287642479,
0.06041271239519119,
0.062441278249025345,
0.06848485767841339,
0.06524111330509186,
0.06564353406429291,
0.07082067430019379,
0.0676988959312439,
0.07045383006334305,
0.06132864952087402,
0.0676838830113411,
0.064274862408638,
0.06932217627763748,
0.06570309400558472,
0.06711895018815994,
0.06243062764406204,
0.0685991570353508,
0.0648145079612732,
0.0689878761768341,
0.06215982884168625,
0.06463310867547989,
0.055849675089120865,
0.06691331416368484,
0.06402920186519623,
0.05663374066352844,
0.056076034903526306,
0.06126324087381363,
0.06105763465166092,
0.06767643988132477,
0.06495392322540283,
0.06322813034057617,
0.0626610741019249,
0.05588827282190323,
0.06816142797470093,
0.06441586464643478,
0.06622708588838577,
0.053589969873428345,
0.06528521329164505,
0.06313464045524597,
0.06422115117311478,
0.06403939425945282,
0.05603744089603424,
0.06019753962755203,
0.05978955700993538,
0.058692023158073425,
0.05943136289715767,
0.06272856146097183,
0.06204809248447418,
0.05660928413271904,
0.06422382593154907,
0.06855501979589462,
0.05489934980869293,
0.05919265374541283,
0.06072087585926056,
0.06199055165052414,
0.06212494522333145,
0.0532408244907856,
0.05850880220532417,
0.06332679837942123,
0.06473984569311142,
0.05620533972978592,
0.056782789528369904,
0.05984317883849144,
0.06279785186052322,
0.0680396780371666,
0.06252189725637436,
0.057086262851953506,
0.057004414498806,
0.05471985042095184,
0.05932869017124176,
0.06498537957668304,
0.06377285718917847,
0.06319063156843185,
0.0629926323890686
],
"gradient_norms": [
0.0037803652230650187,
0.003631350351497531,
0.003396314335986972,
0.00274146581068635,
0.002563745016232133,
0.0024723419919610023,
0.002229638397693634,
0.002245080890133977,
0.002163525903597474,
0.0021340805105865,
0.0021220329217612743,
0.0020243690814822912,
0.0020883611869066954,
0.001933824154548347,
0.0018491963855922222,
0.0018003054428845644,
0.0017963032005354762,
0.0018056104891002178,
0.0018354005878791213,
0.0019063110230490565
],
"activation_means": [
-0.02131376974284649,
-0.03601019084453583,
-0.06662210822105408,
-0.12518449127674103,
-0.1524452120065689,
-0.1727040708065033,
-0.1917235553264618,
-0.20844022929668427,
-0.21533320844173431,
-0.2277694195508957,
-0.2347966730594635,
-0.2530445158481598,
-0.2579406797885895,
-0.2545703649520874,
-0.2688005566596985,
-0.2855533957481384,
-0.28739380836486816,
-0.3037635087966919,
-0.29087066650390625,
-0.3035217523574829
],
"activation_stds": [
0.13483233749866486,
0.1384648084640503,
0.14461848139762878,
0.14690853655338287,
0.155510812997818,
0.15813173353672028,
0.1588330715894699,
0.16174864768981934,
0.1593213975429535,
0.16204914450645447,
0.16705702245235443,
0.16576600074768066,
0.16414247453212738,
0.16486325860023499,
0.16198192536830902,
0.16898144781589508,
0.16573883593082428,
0.17666961252689362,
0.16819415986537933,
0.17242176830768585
]
},
"summary": {
"loss_improvement": 4.957428790139671,
"plain_grad_range": [
0.007635441143065691,
0.010378911159932613
],
"res_grad_range": [
0.0017963032005354762,
0.0037803652230650187
],
"plain_std_range": [
0.35631605982780457,
0.9467869400978088
],
"res_std_range": [
0.13483233749866486,
0.17666961252689362
]
}
}