Buckets:
| import{s as D,n as J,o as Q}from"../chunks/scheduler.37c15a92.js";import{S as V,i as W,g as s,s as o,r as A,A as X,h as r,f as a,c as l,j as N,u as k,x as v,k as O,y as Z,a as n,v as B,d as I,t as R,w as S}from"../chunks/index.2bf4358c.js";import{Y as ee}from"../chunks/Youtube.1e50a667.js";import{C as te}from"../chunks/CourseFloatingBanner.9ff4c771.js";import{H as ae,E as ne}from"../chunks/getInferenceSnippets.24b50994.js";function oe(U){let i,w,_,q,m,x,p,z,u,T,c,Y="I modelli encoder-decoder (detti anche modelli <em>sequence-to-sequence</em>) utilizzano entrambi i componenti dell’architettura Transformer. Ad ogni passaggio, gli attention layer dell’encoder hanno accesso a tutte le parole della frase iniziale, mentre gli attention layer del decoder possono solo accedere alle parole che precedono linearmente una data parola nell’input.",C,d,j='Il pre-addestramento di questi modelli può essere fatto utilizzando gli obiettivi dei modelli encoder o decoder, anche se solitamente include un livello di complessità maggiore. Ad esempio, <a href="https://huggingface.co/t5-base" rel="nofollow">T5</a> è pre-addestrato rimpiazzando porzioni random di testo (che possono contenere più di una parola) con una speciale mask word, con l’obiettivo di predirre il testo rimpiazzato dalla mask word stessa.',M,f,F="I modelli sequence-to-sequence sono più adatti ai compiti che hanno a che fare con la generazione di nuove frasi sulla base di un input preciso, come il riassunto, la traduzione, o la generazione di risposte a domande.",P,h,G="Tra i rappresentanti di questa famiglia di modelli ci sono:",y,g,K='<li><a href="https://huggingface.co/transformers/model_doc/bart.html" rel="nofollow">BART</a></li> <li><a href="https://huggingface.co/transformers/model_doc/mbart.html" rel="nofollow">mBART</a></li> <li><a href="https://huggingface.co/transformers/model_doc/marian.html" rel="nofollow">Marian</a></li> <li><a href="https://huggingface.co/transformers/model_doc/t5.html" rel="nofollow">T5</a></li>',E,$,H,b,L;return m=new ae({props:{title:"Modelli sequence-to-sequence",local:"modelli-sequence-to-sequence",headingTag:"h1"}}),p=new te({props:{chapter:1,classNames:"absolute z-10 right-0 top-0"}}),u=new ee({props:{id:"0_4KEb08xrE"}}),$=new ne({props:{source:"https://github.com/huggingface/course/blob/main/chapters/it/chapter1/7.mdx"}}),{c(){i=s("meta"),w=o(),_=s("p"),q=o(),A(m.$$.fragment),x=o(),A(p.$$.fragment),z=o(),A(u.$$.fragment),T=o(),c=s("p"),c.innerHTML=Y,C=o(),d=s("p"),d.innerHTML=j,M=o(),f=s("p"),f.textContent=F,P=o(),h=s("p"),h.textContent=G,y=o(),g=s("ul"),g.innerHTML=K,E=o(),A($.$$.fragment),H=o(),b=s("p"),this.h()},l(e){const t=X("svelte-u9bgzb",document.head);i=r(t,"META",{name:!0,content:!0}),t.forEach(a),w=l(e),_=r(e,"P",{}),N(_).forEach(a),q=l(e),k(m.$$.fragment,e),x=l(e),k(p.$$.fragment,e),z=l(e),k(u.$$.fragment,e),T=l(e),c=r(e,"P",{"data-svelte-h":!0}),v(c)!=="svelte-g17uha"&&(c.innerHTML=Y),C=l(e),d=r(e,"P",{"data-svelte-h":!0}),v(d)!=="svelte-nga904"&&(d.innerHTML=j),M=l(e),f=r(e,"P",{"data-svelte-h":!0}),v(f)!=="svelte-14dplnv"&&(f.textContent=F),P=l(e),h=r(e,"P",{"data-svelte-h":!0}),v(h)!=="svelte-fo4740"&&(h.textContent=G),y=l(e),g=r(e,"UL",{"data-svelte-h":!0}),v(g)!=="svelte-1lyffe"&&(g.innerHTML=K),E=l(e),k($.$$.fragment,e),H=l(e),b=r(e,"P",{}),N(b).forEach(a),this.h()},h(){O(i,"name","hf:doc:metadata"),O(i,"content",le)},m(e,t){Z(document.head,i),n(e,w,t),n(e,_,t),n(e,q,t),B(m,e,t),n(e,x,t),B(p,e,t),n(e,z,t),B(u,e,t),n(e,T,t),n(e,c,t),n(e,C,t),n(e,d,t),n(e,M,t),n(e,f,t),n(e,P,t),n(e,h,t),n(e,y,t),n(e,g,t),n(e,E,t),B($,e,t),n(e,H,t),n(e,b,t),L=!0},p:J,i(e){L||(I(m.$$.fragment,e),I(p.$$.fragment,e),I(u.$$.fragment,e),I($.$$.fragment,e),L=!0)},o(e){R(m.$$.fragment,e),R(p.$$.fragment,e),R(u.$$.fragment,e),R($.$$.fragment,e),L=!1},d(e){e&&(a(w),a(_),a(q),a(x),a(z),a(T),a(c),a(C),a(d),a(M),a(f),a(P),a(h),a(y),a(g),a(E),a(H),a(b)),a(i),S(m,e),S(p,e),S(u,e),S($,e)}}}const le='{"title":"Modelli sequence-to-sequence","local":"modelli-sequence-to-sequence","sections":[],"depth":1}';function ie(U){return Q(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class ce extends V{constructor(i){super(),W(this,i,ie,oe,D,{})}}export{ce as component}; | |
Xet Storage Details
- Size:
- 4.23 kB
- Xet hash:
- c27f7ffc8af3c911657a62e86d788317cbd61e2e6536d674befef7991dba6a47
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.