Ihe owuwu ihe nrụnye-Decoder
Encoder-decoder architectures kewara ihe atụ ụzọ abụọ: nke na-agụ ma na-atụgharị ntinye n'ime nnochite anya dị n'ime bara ụba, yana nke na-ewepụta ihe sitere na ya.
Nchịkọta
This design powers translation, summarization, and any task where the input and output are different sequences.
Ime miri emi
Ihe ngbanwe ihe ngbanwe na-ahazi nsogbu n'ụzọ abụọ. Ihe ngbanwe ahụ na-agụ usoro ntinye dum (sịnụ, ahịrịokwu bekee) wee tụgharịa ya ka ọ bụrụ usoro ntụgharị okwu na-ejide ihe ọ pụtara. Ihe ngbanwe ahụ na-ewepụta usoro mmepụta (sị, French) otu akara n'otu oge, na-eleghachi azụ na ntinye ya na mbụ yana ihe nnochite anya nke koodu. Ihe ntụgharị mbụ nke afọ 2017 bụ ihe ngbanwe nke e wuru maka ntụgharị asụsụ. Ụdị dị ka T5 na BART na-eji udi a ma kee ọrụ ọ bụla dị ka ederede-na, ederede-apụ. Nkewa ahụ dị ike n'ihi na ihe ngbanwe ahụ nwere ike ịhụ ntinye dum otu mgbe (bidirectional context), ebe ihe ngbanwe na-ewepụta aka ekpe gaa n'aka nri. Nke a na-eme ka nhazi ahụ bụrụ ihe dị mma maka nsogbu usoro-na-usoro ebe ogologo mmepụta na ọdịnaya dị iche na ntinye.
Nghọta nka nka
Ihe ngbanwe ahụ na-eji nlebara anya onwe onye bidirectional, yabụ ndenye ntinye ọ bụla na-aga akara ngosi ọ bụla n'otu oge. The decoder bụ autoregressive na-eji masked onwe-nleba anya, nke pụtara na onye ọ bụla na ọnọdụ nwere ike ịhụ na mbụ ọnọdụ iji chekwa ọgbọ kpatara. Ijikọ ha bụ nlebara anya n'ofe: decoder Layer na-ajụ ajụjụ nzuzo nzuzo ikpeazụ. Nkewa a na-ahapụ koodu ntinye ka ọ wuo nghọta zuru oke, nke na-adabereghị n'usoro ebe onye nrụpụta na-enye otu akara n'otu oge.
Mmetụta atụmatụ
Ọsọ na ọnụ ọgụgụ
Usoro ọrụ asụsụ nwere ike ịga ngwa ngwa n'achụghị nkwụsi ike.
Nweta na iru
Ọ na-agbasawanye ohere n'ofe asụsụ na ụdị nzikọrịta ozi.
Mkpebi doro anya
Ndị otu nwere ike itinyekwu oge na ikpe ebe akpaaka na-ejikwa nkwughachi.
Ọdịnihu nke ihe owuwu ihe nrụnye-Decoder
Ụdị decoder naanị dị ka GPT ugbu a na-achịkwa nkata izugbe n'ihi na otu ngwugwu na-arụ ọrụ dị mfe ma na-ejikwa ọtụtụ ọrụ site na mkpali. Mana ihe nrụnye-decoder na-aga n'ihu ebe nghọta ntinye na ọgbọ mmepụta dị iche: njirimara okwu (ịtụgharị), nchịkọta akwụkwọ, na sistemu multimodal na-ejikọta ihe ngbanwe ọhụụ na ihe ndozi ederede. Na-atụ anya nhazi ụlọ ngwakọ nke na-agbazinye nghota bidirectional nke koodu ahụ maka iweghachite na ntinye ala ma na-edobe mgbanwe ngbanwe, ọkachasị dị ka ụdị na-ejikọta ederede, ọdịyo na onyonyo.
Mmejuputa n'ezie n'ụwa
Google Tụgharịa asụsụ na DeepL na-eji transformers-decoder na-edepụta ahịrịokwu n'otu asụsụ gaa na nke ọzọ.
Whisper OpenAI na-etinye koodu spectrogram ọdịyo wee depụta ya ka ọ bụrụ ederede ederede ma ọ bụ atụgharịrị.
T5 na BART ike abstractive nchịkọta, na-achịkọta ogologo akụkọ n'ime nkenke nchịkọta.
Sistemụ ntinye aha onyonyo jikọtara ihe ngbanwe ọhụụ yana ihe ndozi ederede iji kọwaa foto n'okwu.
Ihe ize ndụ & okporo ụzọ nche
Eziokwu ndị e chepụtara echepụta nwere ike jiri nwayọ tinye akụkọ, nkwado nkwado, ma ọ bụ nsonaazụ nyocha.
Mmetụta ngwa ngwa nwere ike ịmepụta nsonaazụ na-ekwekọghị ekwekọ n'ofe arịrịọ ndị yiri ya.
Enwere ike ikpughe data ederede nwere mmetụta ma ọ bụrụ na njikwa ohere adịghị ike.
Map mmejuputa
Kọwaa usoro mmepụta, ụda, na ụkpụrụ ịdịmma tupu ibugharị.
Weghachite nzaghachi site na isi mmalite ntụkwasị obi mgbe ọ bụla izi ezi dị mkpa.
Debe ebe nleba anya mmadụ maka mpụta dị elu.
Sochie ụkpụrụ ọdịda ma na-azụghachi mkpali ma ọ bụ usoro ọrụ mgbe niile.
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 Encoder-Decoder Architectures 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
Ntuziaka na-esote
Cross-Encoders vs Bi-Encoders
Ajụjụ a na-ajụkarị
What is Encoder-Decoder Architectures?
Encoder-decoder architectures kewara ihe atụ ụzọ abụọ: nke na-agụ ma na-atụgharị ntinye n'ime nnochite anya dị n'ime bara ụba, yana nke na-ewepụta ihe sitere na ya. Nhazi a na-enye ike ntụgharị asụsụ, nchịkọta, na ọrụ ọ bụla ebe ntinye na mmepụta bụ usoro dị iche iche.
Kedu ọrụ bụ isi nke ihe ngbanwe na ihe ngbanwe koodu?
Ihe ngbanwe ahụ na-eri usoro ntinye dum wee wepụta vectors gburugburu na-ewere ihe ọ pụtara, nke onye ndoputa na-ejizi.
Kedu ihe kpatara onye nrụpụta ji eji nkpuchi (ihe kpatara) nlebara anya onwe ya?
Ikpuchi ihu na-eme ka ọnọdụ mmepụta ọ bụla na-aga naanị na ọnọdụ mbụ, na-echekwa usoro ọgbara ọhụrụ nke aka ekpe gaa aka nri.
Kedu usoro jikọtara ihe ngbanwe na nnochite anya ihe ngbanwe?
Nleba anya nlebara anya na-eme ka oyi akwa ọ bụla decoder jụọ ajụjụ ọnọdụ zoro ezo nke koodu ahụ, na-ejikọta nghọta nke ntinye na ọgbọ nke mmepụta.
Kedu n'ime ụdị ndị a bụ ihe nrụpụta ihe ngbanwe (usoro-na-usoro) architecture?
T5 (na BART) bụ ụdị ihe ngbanwe-edochaghị anya nke na-ahazi ọrụ dị ka ederede-na-ederede. BERT bụ koodu ntinye naanị yana GPT-3 bụ naanị ihe ngbanwe.
Kedu ihe kpatara imepụta koodu-decoder ji bụrụ ihe dabara n'okike maka ntụgharị asụsụ?
Ntụgharị maapụ otu usoro gaa na nke dị iche, yabụ na-agụ ebe niile (encoder) na imepụta ebumnuche ọhụrụ (decoder) dabara nke ọma.