course/pr_1189/ko/_app/immutable/nodes/9.008eaae2.js

import{s as J,n as K,o as V}from"../chunks/scheduler.505acc25.js";import{S as W,i as Z,e as m,s as a,c as v,h as tt,a as o,d as n,b as s,f as X,g as b,j as w,k as I,l as et,m as l,n as P,t as T,o as L,p as y}from"../chunks/index.821724d0.js";import{C as nt,H as lt,E as at}from"../chunks/MermaidChart.svelte_svelte_type_style_lang.7483318d.js";import{Y as st}from"../chunks/Youtube.c5effbdd.js";import{C as rt}from"../chunks/CourseFloatingBanner.a3154b9b.js";function mt(B){let r,M,d,E,i,H,f,k,p,S,u,q,$,D="디코더 모델(Decoder models)은 트랜스포머 모델의 디코더만 사용합니다. 각각의 단계마다, 어텐션 레이어는 주어진 단어에 대해 문장 내에서 해당 단어 앞에 위치한 단어들에 대해서만 액세스 할 수 있습니다. 이러한 모델을 <em>자동 회귀(auto-regressive) 모델</em>이라고 부릅니다.",z,c,F="디코더 모델의 사전 학습은 보통 문장 내 다음 단어 예측을 반복하는 방식으로 이루어집니다.",G,g,N="이러한 모델은 텍스트 생성에 특화되어 있습니다.",j,h,O="디코더 모델 계열의 대표 주자들은 다음과 같습니다:",R,_,Q='<li><a href="https://huggingface.co/transformers/model_doc/ctrl.html" rel="nofollow">CTRL</a></li> <li><a href="https://huggingface.co/docs/transformers/model_doc/openai-gpt" rel="nofollow">GPT</a></li> <li><a href="https://huggingface.co/transformers/model_doc/gpt2.html" rel="nofollow">GPT-2</a></li> <li><a href="https://huggingface.co/transformers/model_doc/transformerxl.html" rel="nofollow">Transformer XL</a></li>',U,x,Y,C,A;return i=new nt({props:{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"}}),f=new lt({props:{title:"디코더 모델",local:"디코더-모델",headingTag:"h1"}}),p=new rt({props:{chapter:1,classNames:"absolute z-10 right-0 top-0"}}),u=new st({props:{id:"d_ixlCubqQw"}}),x=new at({props:{source:"https://github.com/huggingface/course/blob/main/chapters/ko/chapter1/6.mdx"}}),{c(){r=m("meta"),M=a(),d=m("p"),E=a(),v(i.$$.fragment),H=a(),v(f.$$.fragment),k=a(),v(p.$$.fragment),S=a(),v(u.$$.fragment),q=a(),$=m("p"),$.innerHTML=D,z=a(),c=m("p"),c.textContent=F,G=a(),g=m("p"),g.textContent=N,j=a(),h=m("p"),h.textContent=O,R=a(),_=m("ul"),_.innerHTML=Q,U=a(),v(x.$$.fragment),Y=a(),C=m("p"),this.h()},l(t){const e=tt("svelte-u9bgzb",document.head);r=o(e,"META",{name:!0,content:!0}),e.forEach(n),M=s(t),d=o(t,"P",{}),X(d).forEach(n),E=s(t),b(i.$$.fragment,t),H=s(t),b(f.$$.fragment,t),k=s(t),b(p.$$.fragment,t),S=s(t),b(u.$$.fragment,t),q=s(t),$=o(t,"P",{"data-svelte-h":!0}),w($)!=="svelte-1p9xm7k"&&($.innerHTML=D),z=s(t),c=o(t,"P",{"data-svelte-h":!0}),w(c)!=="svelte-vpsok7"&&(c.textContent=F),G=s(t),g=o(t,"P",{"data-svelte-h":!0}),w(g)!=="svelte-2694ur"&&(g.textContent=N),j=s(t),h=o(t,"P",{"data-svelte-h":!0}),w(h)!=="svelte-ov0q2c"&&(h.textContent=O),R=s(t),_=o(t,"UL",{"data-svelte-h":!0}),w(_)!=="svelte-jabf5n"&&(_.innerHTML=Q),U=s(t),b(x.$$.fragment,t),Y=s(t),C=o(t,"P",{}),X(C).forEach(n),this.h()},h(){I(r,"name","hf:doc:metadata"),I(r,"content",ot)},m(t,e){et(document.head,r),l(t,M,e),l(t,d,e),l(t,E,e),P(i,t,e),l(t,H,e),P(f,t,e),l(t,k,e),P(p,t,e),l(t,S,e),P(u,t,e),l(t,q,e),l(t,$,e),l(t,z,e),l(t,c,e),l(t,G,e),l(t,g,e),l(t,j,e),l(t,h,e),l(t,R,e),l(t,_,e),l(t,U,e),P(x,t,e),l(t,Y,e),l(t,C,e),A=!0},p:K,i(t){A||(T(i.$$.fragment,t),T(f.$$.fragment,t),T(p.$$.fragment,t),T(u.$$.fragment,t),T(x.$$.fragment,t),A=!0)},o(t){L(i.$$.fragment,t),L(f.$$.fragment,t),L(p.$$.fragment,t),L(u.$$.fragment,t),L(x.$$.fragment,t),A=!1},d(t){t&&(n(M),n(d),n(E),n(H),n(k),n(S),n(q),n($),n(z),n(c),n(G),n(g),n(j),n(h),n(R),n(_),n(U),n(Y),n(C)),n(r),y(i,t),y(f,t),y(p,t),y(u,t),y(x,t)}}}const ot='{"title":"디코더 모델","local":"디코더-모델","sections":[],"depth":1}';function it(B){return V(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class gt extends W{constructor(r){super(),Z(this,r,it,mt,J,{})}}export{gt as component};