Ntuziaka Visual AI

MaskGIT Parallel Token Decoding

MaskGIT na-ewepụta onyonyo site n'ịkọ ọtụtụ akara n'otu oge wee mejupụta ndị nwere ntụkwasị obi na mbụ, jiri obere usoro ngwa ngwa dochie ọgbọ aka ekpe gaa n'aka nri.

2 nkeji na-agụEmelitere ikpeazụ

Ime miri emi

MaskGIT (Mkpuchi ihe ngbanwe ihe onyonyo nkpuchi), sitere na Google na 2022, na-echegharị ka ụdị onyonyo dabere na nbipụta. Ndị ngbanwe mbụ dị ka VQGAN na-emepụta ihe nrịbama na-akpaghị aka, otu n'otu oge n'usoro raster, nke na-adị nwayọọ na nke na-ekwekọghị n'okike maka ihe oyiyi 2D. MaskGIT kama na-azụ ya n'ebumnobi ihe nlegharị anya masked dị ka BERT: ezoro ezo nke obere ihe ngosi onyonyo na ihe nlereanya ahụ na-amụta ịkọ ha niile n'otu oge site na iji nlebara anya bidirectional. N'oge ọgbọ ọ na-amalite site na grid zuru oke ma depụta koodu n'ime ọnụ ọgụgụ a kapịrị ọnụ nke iterations (na-abụkarị 8 ruo 12). Nzọụkwụ ọ bụla ọ na-ebu amụma ihe mkpuchi ọ bụla kpuchiri ekpuchi, na-edobe amụma ntụkwasị obi kachasị elu, ma megharịa ihe ndị ọzọ maka agba na-esote. Nke a na-emepụta onyogho dị elu n'usoro n'usoro dị ntakịrị karịa ngbanwe nke autoregressive.

Nghọta nka nka

Akụkụ dị mkpa bụ usoro nkpuchi dabere na ntụkwasị obi. Usoro ihe omume cosine na-ekpebi ọnụọgụ ole ị ga-ekpughe iteration nke ọ bụla, na-amalite nwayọ na ngwa ngwa. N'ihi na nlebara anya bụ bidirectional, akara ọ bụla na-ahụ onyinyo ahụ dum, yabụ ibu amụma amụma kacha nwee ntụkwasị obi na-eme ka ọnọdụ emechaa mee ọnọdụ na ọnọdụ siri ike, dị ka idozi akụkụ dị mfe nke egwuregwu mgbagwoju anya n'ihu ndị na-enweghị isi.

Mmetụta atụmatụ

Ọsọ na ọnụ ọgụgụ

Visual AI nwere ike megharịa nyocha, nchọpụta na mkpado ọrụ n'ọtụtụ.

Mee nhọrọ

Otu ndị na-emepụta ihe nwere ike imepụta echiche ngwa ngwa site na ngbanwe akwụkwọ ntuziaka ole na ole.

Team na usoro ọrụ

Ọrụ nwere ike iji onyonyo na akara vidiyo siri ike ịhazi.

Ọdịnihu nke MaskGIT Parallel Token Decoding

Ndozi nzigharị nke MaskGIT na-akpalite ebili mmiri nke ndị na-emepụta ihe na-anaghị akwụ ụgwọ, gụnyere MUSE maka ederede gaa na onyonyo yana ụzọ mkpuchi maka vidiyo. Ụkpụrụ ahụ, na-ebu amụma n'otu n'otu ma na-emezigharị n'ime usoro ole na ole, na-anọdụ n'etiti GAN otu égbè na mgbasa ozi ọtụtụ nzọụkwụ, na-enye mgbanwe mgbanwe ngwa ngwa. Na-atụ anya ndozi ihe mkpuchi mkpuchi ga-anọgide na-apụta na igwe na-emepụta ngwa ngwa multimodal na sistemụ ndezi ebe eserese na ihe njuju dị mma.

Mmejuputa n'ezie n'ụwa

Na-emepụta onyonyo zuru oke n'ihe dị ka usoro 8 ruo 12 yiri ya kama ịbụ narị otu narị amụma token autoregressive.

Ịtinye mpaghara ihe mkpuchi nke foto site na ịkọgharị naanị akara ngosi zoro ezo na gburugburu gbara ya gburugburu

Ngwakọta onyonyo ọnọdụ klaasị na ImageNet na asọmpi dị mma yana ụdị nwayọ nwayọ

Na-eje ozi dị ka ọkpụkpụ ngbanwe maka sistemụ ederede gaa na onyonyo dị ka MUSEGoogle nke chọrọ ọgbọ ngwa ngwa.

Ihe ize ndụ & okporo ụzọ nche

Ikike onyonyo na nkwenye nwere ike bụrụ ihe egwu dị n'iwu ma ọ bụrụ na edoghị anya.

Ọrụ nlereanya nwere ike ịdịgasị iche n'ofe ọkụ, igwe mmadụ, na gburugburu.

Enwere ike ghara ịhụ ihe dị mma ma ọ bụrụ na enyochaghị oke ntụkwasị obi.

Map mmejuputa

1

Kọwaa ụkpụrụ nnabata maka nkenke, icheta, na ụgwọ njehie.

2

Nwalee na data dabara na ọnọdụ mmepụta n'ezie.

3

Tinye nyocha mmadụ maka obere obi ike ma ọ bụ amụma mmetụta dị elu.

4

Sochie ihe nlere anya wee megharịa ka emechara mgbanwe igwefoto ma ọ bụ dataset.

Nọgide na-eme nchọpụta

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 MaskGIT Parallel Token Decoding quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Malite ajụjụ

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ntuziaka na-esote

Ọkpụkpụ-nke-Echere Ndozi ndakọrịta

Ajụjụ a na-ajụkarị

What is MaskGIT Parallel Token Decoding?

MaskGIT na-ewepụta onyonyo site n'ịkọ ọtụtụ akara n'otu oge wee mejupụta ndị nwere ntụkwasị obi na mbụ, jiri obere usoro ngwa ngwa dochie ọgbọ aka ekpe gaa n'aka nri.

Kedu ka MaskGIT si ewepụta akara ngosi onyonyo dị iche na nke VQGAN's transformer?

MaskGIT na-ebu amụma akara niile kpuchiri ekpuchi n'otu oge na nlebara anya bidirectional, n'adịghị ka VQGAN na-eji nwayọ ngbanwe akpaaka aka ekpe gaa aka nri.

Kedu ebumnuche nkuzi MaskGIT na-agbaziri n'ụdị asụsụ?

MaskGIT na-ezobe obere ihe nrịbama na-enweghị usoro wee mụta ịkọ ha, ebumnobi imepụta ihe mkpuchi nke yiri BERT.

N'ime ọgbọ, kedu akara ngosi MaskGIT na-eme n'oge ọ bụla?

Nzọụkwụ ọ bụla na-edobe amụma kacha nwee obi ike ma na-ekpuchi nke ọzọ, yabụ na-eme usoro ọnọdụ na ọnọdụ a pụrụ ịdabere na ya.

Ihe dị ka ugboro ole MaskGIT na-achọkarị imepụta onyonyo?

MaskGIT na-ekpebi n'ime obere usoro a kapịrị ọnụ, na-abụkarị 8 ruo 12, dị obere karịa ụzọ mgbagha ma ọ bụ mgbasa ozi.

Kedu nhazi oge na-achịkwa akara ole MaskGIT na-ekpughe nzọụkwụ ọ bụla?

Usoro ihe omume cosine na-ekpughe akara ole na ole n'oge na karịa, na-edozi ogo na ọsọ n'ofe mmegharị.