Transformers faafinta
Beddelayaasha Fidinta (DiTs) waxay ku beddelaan U-Net-ka is-bedbeddelka leh ee udub-dhexaadka sawirka iyo soo-saareyaasha fiidyaha ee laf dhabarta Transformer.
Dulmar
This architecture powers leading systems like Stable Diffusion 3 and OpenAI's Sora, and it scales remarkably well as you add compute.
quusid qoto dheer
Moodooyinka faafinta waxay abuuraan sawirro iyagoo ka soo bilaabaya buuq saafi ah oo si isdaba joog ah u diidaya sawir isku xiran. Muddo sanado ah shabakadu waxay samaynaysay diidmadaas waxay ahayd U-Net, qaab-dhismeedka convolutional. Transformer-ka Diffusion, oo ay soo bandhigeen Peebles iyo Xie sanadka 2022, ayaa U-Net ku bedelay Transformer. Sawirka ayaa marka hore la isku dhejiyaa meel qarsoon, waxaana loo kala qaybiyaa balamo yaryar, oo balastar kastaa wuxuu noqdaa calaamad, sida erayada qaabka luqadda. Transformer-ku wuxuu markaa ku socodsiiyaa calaamadahan isagoo is-fiiro gaar ah u leh tallaabo kasta oo diidmo ah. Natiijooyinka muhiimka ah ayaa ahaa in waxqabadka DiT uu u fiicnaado si la saadaalin karo marka aad kordhiso cabbirka moodeelka oo aad yarayso cabbirka balastarka, adoo raacaya sharciyo miisaan nadiif ah. Miisaankani waa sababta qoraal-ilaa-fidiyoow iyo nidaamyada-dhamaadka-sare ee qoraalka-muuqalka ay si weyn ugu guureen laf-dhabarka Transformer.
Aragtida Farsamada
Hal-abuurnimada udub-dhexaadka ah waa sida DiTs-ku u duraan qaboojinta sida wakhtiga iyo qoraalka qoraalka. Halkii ay ka ahaan lahaayeen isku xidhka fudud, waxay isticmaalaan laqabsiga lakabka caadiga ah (adaLN), halkaas oo shabakadu ay saadaaliso cabirka iyo cabirka wareejinta lakabyada caadiga ah ee calaamada qaboojinta. Kala duwanaanshiyaha adaLN-eberku wuxuu bilaabaa kuwan si xannibaad kastaa u bilaabato sidii shaqo aqoonsi, xasilinta tababarka. Baakadaha waxaa loo fidiyaa calaamado, waxaa lagu farsameeyaa baloogyada Transformer-ka caadiga ah iyadoo is-fiirsasho leh, ka dibna dib ayaa la isugu soo ururiyaa oo dib loogu dejiyaa pixels.
Saamaynta Istiraatijiyadeed
Xawaaraha iyo miisaanka
Visual AI wuxuu si otomaatig ah u samayn karaa baadhista, ogaanshaha, iyo sumadaynta hawlaha miisaanka.
Xulashada dhismayaasha
Kooxaha hal-abuurka leh waxay hindise karaan fikradaha si dhakhso leh iyagoo leh dib-u-eegis buugeed yar.
Kooxda iyo socodka shaqada
Hawlgalladu waxay isticmaali karaan calaamadaha muuqaalka iyo muuqaalka kuwaas oo markii hore adkeyd in la farsameeyo.
Mustaqbalka Transformers faafinta
Beddelayaasha fidinta waxay noqonayaan laf-dhabarta aasaasiga ah ee warbaahinta wax-soo-saarka. Naqshaddooda calaamad ku salaysan ayaa ka dhigaysa mid dabiici ah midaynta sawirada, fiidyaha, iyo xitaa jiilka badan ee hal dhisme oo miisaan leh. Cilmi baaristu waxay u riixaysaa muuqaal dheer, xallin sare, iyo feejignaan hufan si loo xakameeyo qiimaha afar geesoodka ah ee calaamado badan. Filo isku dhafka luqadda iyo moodooyinka aragtida, halkaas oo cuntooyinka isku midka ah ee Transformer-ka iyo kaabayaasha ay u adeegaan labadaba, dardargelinta horumarka moodooyinka adduunka iyo fiidiyowga is-dhexgalka.
Dhaqangelinta Adduunka-dhabta ah
OpenAI's Sora waxay isticmaashaa Transformer laf dhabarta inta lagu guda jiro balastar wakhtiga spacetime si ay u soo saaraan fiidyowyo daacad ah oo daqiiqado ah oo ka yimid qoraalada.
Stable Diffusion 3 waxa ay qabataa bedelaadaha kala-duwanaanshaha hab-badan (MMDiT) si ay si wanagsan isugu waafajiyaan sawirada la soo saaray oo leh sharaxaad faahfaahsan oo qoraal ah.
Cilmi baadhayaashu waxay cabbiraan DiT balaayiin halbeegyo ah waxayna u fiirsadaan tayada sawirka oo si la saadaalin karo u soo hagaagayso, hagaya go'aamada xisaabinta-miisaaniyadda.
Istuudiyaha waxa uu isticmaalaa qaab ku salaysan DiT si uu u kordhiyo cajaladaha gaagaaban, isaga oo u daawaynaya fiidiyooga dheeraadka ah sida calaamado balastar dheeraad ah si loo diido.
Khatarta & Dariiqyada Ilaalada
Xuquuqda sawirka iyo ogolaanshaha waxay noqon kartaa khataro sharci ah haddii caddayntu aanay caddayn.
Waxqabadka moodeelku wuu ku kala duwanaan karaa iftiinka, tirakoobka, iyo deegaanka.
Wanaagga beenta ah waxa laga yaabaa inaan la dareemin ilaa xadka kalsoonida aan la kormeerin.
Qorshe Hawleedka Dhaqangelinta
Qeex shuruudaha aqbalida ee saxnaanta, dib u celinta, iyo kharashyada khaladka.
Ku tijaabi xogta ku habboon xaaladaha wax soo saarka dhabta ah.
Ku dar dib u eegis bini'aadamka si aad u hesho kalsoonida hoose ama saameeynta sare.
Lasoco moodeel dhaqaaqa oo dib u cusboonaysii kamarada ama xogta kaydinta ka dib.
Sii wad Sahaminta
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Diffusion Transformers quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Hagaha xiga
Shabakadaha Transformer Space
Su'aalaha soo noqnoqda
What is Diffusion Transformers?
Beddelayaasha Fidinta (DiTs) waxay ku beddelaan U-Net-ka is-bedbeddelka leh ee udub-dhexaadka sawirka iyo soo-saareyaasha fiidyaha ee laf dhabarta Transformer. Nashqadani waxay awood u leedahay nidaamyada hogaaminta sida Stable Diffusion 3 iyo OpenAI's Sora, waxayna si cajiib ah u miisaamaysaa markaad ku darto xisaabinta.
Qaybkee ayuu beddelaa Transformer-ka fidinta marka la barbar dhigo moodooyinka fidinta dhaqameed?
DiT-yadu waxay ku beddelaan U-Net, shabakadda isku-dhafan ee samaysay been-abuur, oo leh laf dhabar Transformer.
Sidee buu DiT u beddelaa sawirka shay Transformer-ku socodsiin karo?
Sawirka dahsoon wuxuu u qaybsan yahay balastar yaryar, oo balastar kastaa wuxuu noqonayaa calaamad, oo la mid ah erayada qaabka luqadda.
Muxuu warqaddii asalka ahaa ee DiT ka helay miisaan?
Tayada DiT waxay u fiicnaanaysaa si nadiif ah, oo la saadaalin karo markaad kordhiso xisaabinta moodeelka oo aad isticmaasho balastar yar yar, adoo raacaya sharciyada miisaanka.
Sidee buu DiTs caadi ahaan u duraa qaboojiyaha sida wakhtiga iyo qoraalka qoraalka?
DiTs waxay isticmaalaan habaynta lakabka la qabsiga si ay u saadaaliyaan cabbirada cabbirka iyo beddelka lakabyada caadiga ah ee calaamadda qaboojinta.
Maxaa bilawga 'adaLN-eber' qabtaa?
adaLN-eber billow habaynta si block kasta marka hore u dhaqmo sidii aqoonsi, kaas oo dejinaya oo hagaajinaya tababarka.