Parti Pathways Autoregressive Hoto
Parti (Pathways Autoregressive Rubutu-zuwa Hoto) yana haifar da hotuna yadda ƙirar harshe ke rubuta jimloli: alamar hoto ɗaya a lokaci guda, tsinkaya na gaba daga duk waɗanda suka zo a baya.
Dubawa
It matters because it showed that simply scaling a sequence model can produce strikingly detailed, prompt-faithful images.
Zurfafa nutsewa
Parti yana ɗaukar tsarar hoto azaman matsalar fassarar jeri-zuwa-jeri, kamar fassarar inji. Alamar alama ta ViT-VQGAN ta fara ɓoye hoto cikin jerin sahihan alamu waɗanda aka zana daga littafin lambar koyo. Mai canza canji yana karanta saƙon rubutu, sannan mai canza maɓalli sannan ya haifar da alamun hoton kai tsaye, kowanne yana da sharadi akan rubutun da kuma kan alamun da aka fitar a baya. Bayan an samar da duk alamu, mai gyara alamar tokenizer yana sake gina pixels. Google an daidaita Parti daga miliyan 350 har zuwa ma'auni biliyan 20, kuma ingancin hoto da daidaita rubutu ya inganta a hankali tare da girma. Samfurin 20B yana kula da dogon lokaci, faɗakarwa na ƙirƙira, ingantaccen rubutu, da cikakkun bayanai masu daraja. Parti ya kuma gabatar da maƙasudin PartiPrompts, saitin ƙalubale sama da 1,600 wanda ya ƙunshi nau'ikan nau'ikan da yawa da matakan wahala.
Fahimtar Fasaha
Siffar ma'anar ita ce tsantsar jujjuyawar kai akan sahihan alamu na gani: ƙirar ta ƙirƙira hoton azaman samfuri na yuwuwar alama ta gaba, iri ɗaya cikin ruhi ga tsarar rubutu irin na GPT. Wannan yana haɓaka hangen nesa da harshe a ƙarƙashin girke-girke na horarwa kuma yana ba shi damar gaji shekaru da yawa na dabaru na ƙirar ƙira. Farashin shine ƙididdige ƙididdiga, tun da dole ne a samar da alamu cikin tsari, wanda ke sa tsararru ya yi ƙasa da hankali fiye da hanyoyin layi ɗaya, amma yana da ƙima kuma yana amfana kai tsaye daga manyan samfura.
Dabarun Tasiri
Gudu da sikelin
Kayayyakin AI na iya sarrafa aiki da bincike, ganowa, da ayyuka masu alama a sikelin.
Gina zaɓuɓɓuka
Ƙungiyoyin ƙirƙira za su iya samar da ra'ayoyi cikin sauri tare da ƙarancin bita da hannu.
Ƙungiya da aikin aiki
Ayyuka na iya amfani da siginar hoto da bidiyo waɗanda a baya suke da wahalar aiwatarwa.
Makomar Parti Pathways Autoregressive Hoto
Hoto mai jujjuyawa yana jin daɗin farfaɗo saboda ƙashin baya ɗaya na iya ƙirar rubutu, hotuna, sauti, da bidiyo azaman rafi guda ɗaya, yana ba da damar haɗe-haɗen ƙirar multimodal da gaske. Bincike yana magance babban rauninsa, jinkirin samfurin jeri, tare da ƙididdige ƙididdiga, tsinkayar alama mai kama da juna, da mafi kyawun alamu. Yi tsammanin muryoyi masu ƙarfi a cikin mataimaka na gabaɗaya waɗanda ke barin karatu, tunani, da tsara hoto, da kuma ganin ƙa'idodin ƙira suna tura daidaiton ƙira da ingantaccen ma'anar rubutu cikin hoto har ma da gaba.
Aiwatar da Gaskiyar Duniya
Yin hadaddun al'amuran abubuwa da yawa daga dogon fa'ida, kamar takamaiman tsari na dabbobi, abubuwa, da asalinsu.
Ƙirƙirar hotuna waɗanda suka haɗa da rubutattun kalmomi ko alamomi, inda yin oda na kai-tsaye yana taimakawa rubuta rubutu daidai.
Ƙididdigar ƙididdiga da gwajin damuwa-nau'in rubutu-zuwa-hoto ta amfani da suite PartiPrompts a cikin nau'o'i kamar ilimin duniya da ra'ayoyi masu ban mamaki.
Samar da cikakkun bayanai don faɗakarwa da ke buƙatar madaidaicin ƙidayawa da alaƙar sararin samaniya tsakanin abubuwa da yawa.
Hatsari & Tsare-tsare
Haƙƙoƙin hoto da yarda na iya zama haxarin doka idan ba a fayyace ba.
Ayyukan samfuri na iya bambanta a ko'ina cikin haske, ƙididdiga, da mahalli.
Ƙarya tabbataccen ƙila ba za a iya lura da shi ba sai dai idan an kula da ƙofofin amincewa.
Taswirar Hanya
Ƙayyade ma'auni na karɓa don daidaito, tunowa, da farashi na kuskure.
Gwada tare da bayanan da suka dace da ainihin yanayin samarwa.
Ƙara bita na ɗan adam don ƙarancin amincewa ko tsinkaya mai tasiri.
Bi diddigin ƙirar ƙira kuma sake ingantawa bayan canje-canjen kamara ko saitin bayanai.
Ci gaba da Bincike
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 Parti Pathways Autoregressive Imaging 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
Jagora na gaba
Ƙarfin Hoto na Kai-da-kai
Tambayoyin da ake yawan yi
What is Parti Pathways Autoregressive Imaging?
Parti (Pathways Autoregressive Rubutu-zuwa Hoto) yana haifar da hotuna yadda ƙirar harshe ke rubuta jimloli: alamar hoto ɗaya a lokaci guda, tsinkaya na gaba daga duk waɗanda suka zo a baya. Yana da mahimmanci saboda ya nuna cewa kawai ƙirƙira samfurin jeri na iya samar da cikakkun hotuna masu saurin gaske.
Ta yaya Parti ke samar da hoto?
Parti ne autoregressive: yana samar da alamomin hoto a jere, kowane sharadi akan rubutu da kan alamun da aka ƙirƙira a baya.
Wane tsari gabaɗaya Parti ke amfani da shi don aikin rubutu-zuwa hoto?
Parti yana ɗaukar tsararraki kamar fassarar inji: mai rikodin rubutu yana karanta rubutu kuma mai yankewa yana fitar da alamun hoto, saitin jeri-zuwa-juyi na gargajiya.
Menene scaling Parti har zuwa mitoci biliyan 20 ya nuna?
Yayin da Parti ya girma daga sigogi na 350M zuwa 20B, duka aminci da aminci na gaggawa sun inganta, gami da ingantattun abubuwan faɗakarwa da rubutu mai iya karantawa.
Me ke canza hoto zuwa alamomi na musamman na Parti?
Parti yana amfani da ViT-VQGAN don ɓoye hotuna zuwa jeri-nau'i na alamomi masu hankali waɗanda mai canza canjin Transformer ya koyi tsinkaya.
Menene babban fage na yanke hukunci na autoregressive Parti?
Domin kowace alama ta dogara da waɗanda suka gabata, tsararru tana kan layi da hankali fiye da hanyoyin layi ɗaya, ko da yake tana da ƙima.