Koofiyadaha Tensor
Tensor Cores waa unug qalab gaar ah oo ku dhex jira NVIDIA GPU-yada casriga ah kuwaas oo si xawli ah u qabta hawlo badan oo isku dhufanaya oo ururinaya.
Dulmar
They are the main reason a single GPU can train and run large neural networks orders of magnitude faster than general-purpose compute would allow.
quusid qoto dheer
Waxaa lagu soo bandhigay qaab-dhismeedka Volta ee 2017, Tensor Cores waa wareegyo u go'ay oo xisaabiya isku-dhufashada matrix yar oo lagu daro (D = A x B + C) hal hawlgal, halkii la samayn lahaa mid kasta hal mar ku dhufo koodhadhka CUDA ee caadiga ah. Sababtoo ah ku dhawaad lakab kasta oo ka mid ah shabakada neerfaha ayaa hoos u dhigta isku dhufashada matrixka, tani waxay la mid tahay xisaabta AI runtii u baahan tahay. Jiil kasta oo GPU ah ayaa ballaariyay waxa ay gacanta ku hayaan: Volta wuxuu sameeyay 4 × 4 FP16 tiles, halka markii dambe Ampere, Hopper, iyo Blackwell ay ku darsadeen qaabab hoose oo sax ah sida TF32, BF16, INT8, FP8, iyo FP4. Saxnaanta hoose waxay la macno tahay tirooyin badan oo la farsameeyay saacadiiba, si wayn kor loogu qaadayo wax soo saarka tababarka iyo ka-fiirsashada iyadoo la ilaalinayo saxnaanta la aqbali karo.
Aragtida Farsamada
Qalabka Tensor Core wuxuu ku dhuftaa laba qaybood oo yaryar wuxuuna ku ururiyaa natiijada hal tallaabo oo isku dhafan, isagoo ka faa'iidaysanaya xaqiiqda ah in isla qiyamka wax-soo-saarka dib loogu isticmaalo qaybo badan oo wax soo saar ah. Waxay caadi ahaan u akhridaa wax-soo-gelinta si sax ah oo la dhimay (FP16, BF16, ama FP8) laakiin waxay uruurisaa wadarta socodsiinta si sax ah oo sare (badanaa FP32) si loo xaddido khaladka wareegga. Maktabadaha softiweerka ah sida cuBLAS iyo cuDNN, iyo qaab-dhismeedka sida PyTorch, waxay si toos ah ugu dhejiyaan matrixyada waaweyn ee baloogyadan yar yar si ay moodooyinka u helaan xawaareyn la'aanteed.
Saamaynta Istiraatijiyadeed
Qiimaha iyo miisaaniyada
Go'aamada qaab-dhismeedku waxay horseedaan waxqabadka iyo kharashka hawlgalka sannadaha.
Go'aamo cad
Waxbarashada farsamada waxay ka caawisaa kooxaha inay doortaan xidhmo sax ah, ma aha oo kaliya kan ugu cusub.
Xakamaynta tayada
Doorashooyinka injineernimada ee wanaagsan waxay yareeyaan shilalka la isku halleyn karo ee wax soo saarka.
Mustaqbalka Kooreyada Tensor-ka
Kooreyada Tensor-ku waxay sii wadaan inay u socdaan dhinaca saxda ah ee weligood-hooseeya: Hopper wuxuu ku daray FP8 iyo Blackwell wuxuu soo bandhigay 4-bit FP4 oo leh qalab lagu maamulo qalabka, oo qiyaas ahaan labanlaabaya soo saarista tallaabo kasta ee culeyska shaqo ee culus. Filo taageero adag oo loogu talagalay yaraynta (ka boodka miisaanka eber), qaababka mikroskooliga ee ku lifaaqa qodobbada cabbirka tiro yar oo tirooyin ah, iyo is-dhexgalka qoto dheer ee nidaamyada xusuusta si ay udubyadu u sii quudiyaan. Sida moodooyinka u koraan, mishiinka matrixka, ee aan ahayn xawaaraha saacada ceyriinka ah, ayaa weli ah goobta dagaalka dhexe ee waxqabadka qalabka AI.
Dhaqangelinta Adduunka-dhabta ah
Tababarka moodooyinka luqadaha waaweyn sida transformers-yada qaabka GPT, halkaas oo balaayiin isku dhufasho matrix ah talaba kasta ay ku socdaan Tensor Cores ee BF16 ama FP8.
Ku socodsiinta qiimaynta-waqtiga-dhabta ah ee chatbots-ka iyo soo-saareyaasha sawirka, iyada oo la adeegsanayo tirooyinka INT8 ama FP8 si loogu adeego isticmaaleyaal badan halkii GPU.
Dardar gelinta NVIDIA DLSS ee ciyaaraha fiidyaha, halkaas oo shabakada neural ay kor u qaaddo xayndaabyada xallinta hoose iyada oo la adeegsanayo Tensor Cores jir kasta.
Dadajinta xisaabinta sayniska sida borotiinka-lalaabista (AlphaFold) iyo moodooyinka cimilada ee dib loo habeeyey sidii culeysyo shaqo oo culus oo neerfaha ah.
Khatarta & Dariiqyada Ilaalada
Hagaajinta hal bartilmaameed waxay qarin kartaa daciifnimada nidaamka ballaaran.
Kaabayaasha dhaqaalaha iyo dayactirka inta badan waa la dhayalsadaa.
Nabadgelyada iyo daldaloolada u fiirsashada ayaa kori kara marka nidaamyadu noqdaan kuwo aad u adag.
Qorshe Hawleedka Dhaqangelinta
Qeex daahida, tayada, iyo bartilmaameedyada qiimaha ka hor inta aan la hirgelin.
Benchmark marka la eego culeyska dhabta ah iyo xaaladaha xogta.
La socodka qalabka khaladaadka, leexashada, iyo saamaynta isticmaalaha.
U diyaari dib-u-noqoshada iyo dariiqyada jawaab-celinta dhacdada ka hor inta aanad miisaan.
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 Tensor Cores 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
Isbarbardhigga Tensor-ka ee Modelyada Waaweyn
Su'aalaha soo noqnoqda
What is Tensor Cores?
Tensor Cores waa unug qalab gaar ah oo ku dhex jira NVIDIA GPU-yada casriga ah kuwaas oo si xawli ah u qabta hawlo badan oo isku dhufanaya oo ururinaya. Iyagu waa sababta ugu weyn ee hal GPU uu u tababari karo oo uu maamuli karo shabakadaha neerfaha ee waaweyn amarada baaxadda leh ee ka dhaqso badan xisaabinta ujeeddada guud ay oggolaan karto.
Waa maxay hawlgalka xisaabeed ee xudunta u ah Tensor Cores si gaar ah loogu talagalay in la dardargeliyo?
Kooreyada Tensor-yadu waxay qabtaan isku-dhufashada matrix-ka isku dhafan iyo ururinta hal tallaabo, taas oo dhab ahaan ah hawlgalka xukuma lakabyada shabakada neerfaha.
Waa kuwee NVIDIA GPU architecture markii ugu horeysay soo bandhigtay Tensor Cores?
Tensor Cores ayaa lagaga dooday qaab dhismeedka Volta sanadkii 2017, laga bilaabo hawlgalada FP16 4x4.
Maxay Tensor Cores inta badan u isticmaalaan saxnaanta la dhimay sida FP16 ama FP8?
Isticmaalka xoogaa yar lambarkiiba waxay ka dhigan tahay in qiyam badan ay ku dhex socdaan qalabka wareeg kasta, si aad ah u kordhiya xawaaraha iyada oo yar, inta badan loo dulqaadan karo, luminta saxnaanta.
Si loo ilaaliyo saxnaanta, waa maxay saxnaanta ay Tensor Cores caadi ahaan u isticmaalaan wadarta socodsiinta (urursiga)?
Wax-soo-galyadu waxay noqon karaan kuwo sax ah oo hooseeya, laakiin ururiyaha badanaa wuxuu ku shaqeeyaa si sax ah oo sarreeya sida FP32 si loo xaddido dhisidda khaladaadka wareegga.
Sidee qaab-dhismeedka AI ee maalinlaha ah sida PyTorch uga faa'iideystaan Kooreyada Tensor?
Maktabadaha hoose waxay si toos ah u jebiyaan hawlgallada matrix-yada waaweyn ee loo yaqaan 'Tensor-Core tiles' si toos ah, marka qaab-dhismeedyadu waxay helayaan xawaaraha iyada oo aan gacanta lagu qorin.