ZML: Uphononongo — Ukusebenza Okuphezulu Nakweyiphi na Itshiphu (NVIDIA, AMD, TPU Nezinye)

#Quasa #QUA #zml

I-ZML yinkampani yeziseko zophuhliso lwe-AI eseParis, eyakha iseti yeenkqubo zoqikelelo ezikumgangatho wokusetyenziswa kwiinkqubo zemveliso, eyahlula imisebenzi ye-AI kwizixhobo zekhompyutha eziphantsi kolawulo olukhethekileyo lwabavelisi bazo. Ingcamango yabo esisiseko — “Ukusuka Kwimodeli Ukuya Kwizixhobo Zekhompyutha” — ivumela nayiphi na imodeli ukuba isebenze kwizikhawulezisi ezininzi (NVIDIA, AMD, TPU, Trainium nezinye) isebenzisa isiseko esinye sekhowudi, ngoxa ifikelela kowona mgangatho uphezulu wokusebenza kwezixhobo zekhompyutha.

Ngokungafaniyo nezakhelo eziqhelekileyo ezixhomekeke kwiimeko zokuqhuba iinkqubo ze-Python ezifuna izixhobo ezininzi nakumaleko afihla iinkcukacha zokusebenza, i-ZML iguqulela iimodeli ngqo kwikhowudi yezixhobo zekhompyutha isebenzisa i-Zig ne-MLIR. Oku kunciphisa kakhulu umsebenzi owongezelelweyo wenkqubo, kwenza kube lula ukuqikelela indlela eza kusebenza ngayo, kwandisa isantya, kuze kunike abaphuhlisi amava okusebenza alula nacoceke ngakumbi. Inkampani ikugatya ngokucacileyo ukufihlwa kwemeko yenkqubo, iindlela ezifihla iinkcukacha zokusebenza ngokungathi ngumlingo, neemeko zokuqhuba iinkqubo ezixhomekeke kakhulu kwi-Python, ikhetha ukucaca, ukudityaniswa ngokulula kwamacandelo, nokusebenza kakuhle.

I-ZML isandul’ ukukhupha i-ZML/LLMD, iseva enamandla yoqikelelo lwe-LLM, yaye sele ifumene ukuqatshelwa luluntu lwe-AI, kuquka nenkxaso evakaliswe esidlangalaleni nguYann LeCun, ophumelele i-Turing Award. Le seti yeenkqubo yakhelwe ngokukhethekileyo iimeko zemveliso apho ukusebenza kakuhle nokukwazi ukusebenzisa izixhobo zekhompyutha ezahlukeneyo kubaluleke kakhulu.

I-ZML ifaneleke kakhulu kwiinjineli ze-AI, amaqela e-ML, iinkampani ezisakhasayo, namashishini asebenzisa iimodeli ngomlinganiselo omkhulu nafuna ukuphepha ukubotshelelwa kumboneleli omnye ngoxa efikelela kwesona santya siphezulu sinokufumaneka soqikelelo kwiitshiphu ezahlukeneyo.

Amanqaku Abalulekileyo

• Uqikelelo olungaxhomekekanga ngokwenene kuhlobo lwezixhobo zekhompyutha ngesiseko esinye sekhowudi.
• Ukuguqulelwa ngokuthe ngqo kwekhowudi kwizikhawulezisi ezininzi ukuze kufikelelwe kowona mgangatho uphezulu wokusebenza.
• Umsebenzi owongezelelweyo wenkqubo omncinane kakhulu namava abalaseleyo kubaphuhlisi.
• Ugxininiso olukhulu ekulungeleni ukusetyenziswa kwimveliso nasekuqikeleleni indlela inkqubo eza kusebenza ngayo.

Izinto Ezinokufuna Ukuqwalaselwa

• Njengesisombululo esisesitsha nesinzulu ngokobugcisa, sinokufuna ixesha lokufunda kumaqela aqhelene nezakhelo ezisekelwe ikakhulu kwi-Python.
• Sifaneleke kakhulu kubasebenzisi ababeka phambili amandla okusebenza angundoqo nolawulo.

Isigqibo Ngokubanzi: 4.5/5 yeenkwenkwezi

I-ZML ibalasele njengenye yeendlela ezinamabhongo amakhulu nezingqongqo ngokobugcisa zoqikelelo lwe-AI ngo-2026. Ngokuhamba “ukusuka kwimodeli ukuya kwizixhobo zekhompyutha,” inkampani isombulula iingxaki zokwenene zemveliso ezinxulumene nokusebenza, ukubhetyebhetye, nokwaneliseka kwabaphuhlisi. Sisisombululo esithembisayo kuye nabani na ozimiseleyo ekusebenziseni i-AI esebenza kumgangatho ophezulu nenokusebenza kwizixhobo zekhompyutha ezahlukeneyo.

Nawe fumana umvuzo we-QUA ngokubhala uphononongo kwi-Quasa.io!

Qalisa: https://quasa.io/projects/zml

#Quasa #QUA #zml #QuasaRewards