M\u00eb 24 janar, inxhinier\u00ebt e Intel publikuan nj\u00eb p\u00ebrdit\u00ebsim t\u00eb madh p\u00ebr paket\u00ebn udh\u00ebheq\u00ebse t\u00eb mjeteve open-source p\u00ebr inteligjenc\u00ebn artificiale, OpenVINO 2023.3. Kjo ofron mb\u00ebshtetje t\u00eb plot\u00eb p\u00ebr procesor\u00ebt e rinj Emerald Rapids dhe Meteor Lake, si dhe p\u00ebrmir\u00ebsime t\u00eb tjera t\u00eb harduerit Intel p\u00ebr inteligjenc\u00ebn artificiale gjenerative (GenAI) dhe modelet e m\u00ebdha gjuh\u00ebsore (LLM).<\/p>\n
OpenVINO 2023.3 prezanton depot OpenVINO Gen AI p\u00ebr t\u00eb demonstruar mostra t\u00eb tij p\u00ebr tubat C\/C++ t\u00eb modeleve t\u00eb m\u00ebdha gjuh\u00ebsore dhe verifikimin e modeleve t\u00eb tjera, duke p\u00ebrfshir\u00eb Mistral, Zephyr, ChatGLM3 dhe t\u00eb tjera. Torch.compile tani \u00ebsht\u00eb plot\u00ebsisht i integruar me OpenVINO.<\/p>\n
P\u00ebr t\u00eb zgjeruar mb\u00ebshtetje p\u00ebr modelet gjuh\u00ebsore t\u00eb m\u00ebdha, mb\u00ebshtetet formati i kompresionit t\u00eb modelit INT4 n\u00eb procesor\u00ebt Intel Xeon, si dhe n\u00eb Intel Core dhe Intel iGPU. Performanca e LLM-s\u00eb q\u00eb bazohen n\u00eb transformator\u00eb \u00ebsht\u00eb p\u00ebrmir\u00ebsuar si n\u00eb procesor\u00ebt qendror\u00eb ashtu edhe n\u00eb grafikat, dhe optimizimi p\u00ebr modelet Hugging Face \u00ebsht\u00eb b\u00ebr\u00eb m\u00eb i leht\u00eb dhe shum\u00eb m\u00eb tep\u00ebr.<\/p>\n
P\u00ebr informacion: OpenVINO \u00ebsht\u00eb nj\u00eb paket\u00eb e hapur dhe falas e mjeteve q\u00eb ndihmon zhvilluesit dhe analist\u00ebt e t\u00eb dh\u00ebnave t\u00eb \u0443\u0441\u043a\u043e\u0440ojn\u00eb zhvillimin e zgjidhjeve me performanc\u00eb t\u00eb lart\u00eb p\u00ebr p\u00ebrdorim n\u00eb sisteme t\u00eb ndryshme vizuale. Ky grup i plot\u00eb mjetesh mb\u00ebshtet nj\u00eb gam\u00eb t\u00eb gjer\u00eb zgjidhjesh p\u00ebr t\u00eb par\u00eb kompjuterike, optimizon shp\u00ebrndarjen e t\u00eb m\u00ebsuarit t\u00eb thell\u00eb dhe siguron ekzekutimin e leht\u00eb n\u00eb platforma t\u00eb ndryshme Intel. Po ashtu, OpenVINO zgjidh shum\u00ebllojshm\u00ebri problemesh, duke p\u00ebrfshir\u00eb detektimin e fytyr\u00ebs, njohjen automatike t\u00eb objekteve, teksteve dhe z\u00ebrit, p\u00ebrpunimin e imazheve dhe shum\u00eb m\u00eb tep\u00ebr.<\/p>\n
![\"OpenVINO](\"\/wp-content\/uploads\/2024\/01\/6a91550ef8c9ed32d2ec8e3d346da068.jpg\")
<\/p>\n