InoteveraGaidhi rinotevera
Core ML for On-Device Apple Deployment
Tekinoroji
Nhungamiro yehunyanzvi
Tensor Cores
Tensor Cores inyanzvi dzehurdware mukati memazuva ano NVIDIA GPUs anoita matrix kuwanda-uye-kuunganidza mashandiro nekukurumidza.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
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.
Kudzika Kwakadzika
Yakaunzwa neVolta architecture muna 2017, Tensor Cores masekete akatsaurirwa anokokorodza diki matrix kuwanza pamwe nekuwedzera (D = A x B + C) mukushanda kumwe chete, pane kuita imwe neimwe kuwanda imwe panguva pane yakajairwa CUDA cores. Nekuti ingangoita yese layer yeneural network inodzikisira kuita matrix kuwanda, izvi zvinoenderana nemasvomhu AI anotoda. Chizvarwa chega chega cheGPU chakawedzera zvavanobata: Volta akaita 4x4 FP16 mataira, ukuwo Ampere, Hopper, uye Blackwell zvivakwa zvakawedzera akaderera-chaiyo mafomati seTF32, BF16, INT8, FP8, uye FP4. Kunyatsodzika kwakadzika kunoreva nhamba dzakawanda dzakagadziriswa pawachi imwe neimwe, zvichiwedzera kuwedzera kwekudzidzira uye kufungidzira uku uchichengeta chokwadi chinogamuchirwa.
Kuitwa muNyika yeChokwadi
Kudzidzira mamodheru emitauro mikuru seGPT-maitiro ekushandura, uko mabhiriyoni ematrix kuwanda padanho rinomhanya paTensor Cores muBF16 kana FP8.
Kumhanyisa chaiyo-nguva inference yechatbots uye majenareta emifananidzo, uchishandisa INT8 kana FP8 quantization yekushandira vashandisi vazhinji paGPU.
Kumhanyisa NVIDIA DLSS mumitambo yemavhidhiyo, uko neural network inokwidza yakaderera-resolution mafuremu uchishandisa Tensor Cores furemu yega yega.
Kumhanyisa komputa yesainzi senge protein-kupeta (AlphaFold) uye mamiriro ekunze akagadziridzwa sematrix-inorema neural workloads.
Ramangwana reTensor Cores
Tensor Cores anoramba achifamba akananga kunogara-pasi-chaiyo: Hopper yakawedzera FP8 uye Blackwell yakaunza 4-bit FP4 ine hardware-inogadziriswa scaling, ingangoita kaviri kubuda nhanho imwe neimwe yekufunga-inorema basa rekuita. Tarisira kusimba kwerutsigiro rwe sparsity (kusvetuka zero huremu), microscaling mafomati anonamira zvikero kune zvidiki zvidhinha zvenhamba, uye yakadzama yekubatanidza nemamemory system kuitira kuti macores arambe akadyiswa. Sezvo mamodheru achikura, iyo matrix injini, kwete mbishi wachi yekumhanya, inoramba iri yepakati nhandare yeAI hardware kuita.
Kuzvishandisa
Nei zvichikosha
- Mutengo uye bhajeti. Zvisarudzo zvezvivakwa zvinotyaira kuita uye mutengo wekushandisa kwemakore.
- Sarudzo dzakajeka. Dzidzo yehunyanzvi inobatsira zvikwata kusarudza murwi wakakodzera, kwete iwo mutsva chete.
- Kudzora kwemhando yepamusoro. Sarudzo dzeinjiniya dziri nani dzinoderedza zviitiko zvekuvimbika mukugadzira.
Chenjera
- Kugadzirisa imwe bhenji kunogona kuvanza yakafara system kushaya simba.
- Infrastructure uye mari yekugadzirisa inowanzotarisirwa pasi.
- Chengetedzo uye kucherechedzwa mapundu anogona kukura sezvo masisitimu anowedzera kuoma.
Matanho anotevera
- Tsanangura latency, mhando, uye mutengo zvinangwa usati waitwa.
- Benchmark pasi pechokwadi mutoro uye data mamiriro.
- Chishandiso chekutarisa zvikanganiso, kudonha, uye mushandisi maitiro.
- Gadzirira nzira dzekudzosera kumashure uye dzezviitiko usati wawedzera.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Take the Tensor Cores quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
What is Tensor Cores?
Tensor Cores inyanzvi dzehurdware mukati memazuva ano NVIDIA GPUs anoita matrix kuwanda-uye-kuunganidza mashandiro nekukurumidza. Ndivo chikonzero chikuru chekuti GPU imwe chete inogona kudzidzisa uye kumhanya hombe neural network maodha ehukuru nekukurumidza kupfuura general-chinangwa compute ingabvumira.
Ndeupi musimboti wekushanda kwemasvomhu ndiwo Tensor Cores akagadzirirwa kuti akurumidze?
Tensor Cores anoita yakasanganiswa matrix kuwanza uye kuunganidza mudanho rimwe, ndiko kunoshanda chaiko kunotonga neural network layer.
Ndeipi NVIDIA GPU yekuvaka yakatanga kuunza Tensor Cores?
Tensor Cores yakatanga neVolta architecture muna 2017, kutanga neFP16 4x4 matrix mashandiro.
Sei Tensor Cores achiwanzo shandisa yakaderedzwa chaiyo seFP16 kana FP8?
Kushandisa mabhiti mashoma panhamba kunoreva kuwanda kwehukoshi kuyerera kuburikidza ne Hardware kutenderera kwega kwega, zvakanyanya kuwedzera kumhanya nediki chete, kazhinji inoshiviririka, kurasikirwa kwechokwadi.
Kuti uchengetedze huchokwadi, ndeipi chaiyo iyo Tensor Cores anowanzo shandisa kune inomhanya sum (kuunganidza)?
Zvinopinza zvinogona kunge zvakaderera, asi accumulator inowanzo mhanya nepamusoro-soro seFP32 kudzikamisa kuunganidzwa kwekukanganisa kwekutenderera.
Mazuva ese AI masimusi sePyTorch anoshandisa sei Tensor Cores?
Maraibhurari ari pasi anopwanya mashandiro makuru ematrix muTensor-Core-saizi mataira otomatiki, saka masisitimu anowana kukurumidza pasina manyorero ekunyora.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi