diffusers/main/en/_app/pages/optimization/opt_overview.mdx-hf-doc-builder.js

import{S as Q,i as V,s as W,e as n,k as S,w as Z,t as u,M as ee,c as i,d as t,m as N,a as s,x as te,h as p,b as r,G as o,g,y as oe,L as ae,q as re,o as ne,B as ie,v as se}from"../../chunks/vendor-hf-doc-builder.js";import{I as le}from"../../chunks/IconCopyLink-hf-doc-builder.js";function ce(R){let l,x,c,f,k,m,T,b,q,O,_,z,$,a,I,v,E,C,D,w,G,H,y,L,M,A;return m=new le({}),{c(){l=n("meta"),x=S(),c=n("h1"),f=n("a"),k=n("span"),Z(m.$$.fragment),T=S(),b=n("span"),q=u("Overview"),O=S(),_=n("p"),z=u("Generating high-quality outputs is computationally intensive, especially during each iterative step where you go from a noisy output to a less noisy output. One of \u{1F917} Diffuser\u2019s goal is to make this technology widely accessible to everyone, which includes enabling fast inference on consumer and specialized hardware."),$=S(),a=n("p"),I=u("This section will cover tips and tricks - like half-precision weights and sliced attention - for optimizing inference speed and reducing memory-consumption. You\u2019ll also learn how to speed up your PyTorch code with "),v=n("a"),E=n("code"),C=u("torch.compile"),D=u(" or "),w=n("a"),G=u("ONNX Runtime"),H=u(", and enable memory-efficient attention with "),y=n("a"),L=u("xFormers"),M=u(". There are also guides for running inference on specific hardware like Apple Silicon, and Intel or Habana processors."),this.h()},l(e){const h=ee('[data-svelte="svelte-1phssyn"]',document.head);l=i(h,"META",{name:!0,content:!0}),h.forEach(t),x=N(e),c=i(e,"H1",{class:!0});var P=s(c);f=i(P,"A",{id:!0,class:!0,href:!0});var F=s(f);k=i(F,"SPAN",{});var X=s(k);te(m.$$.fragment,X),X.forEach(t),F.forEach(t),T=N(P),b=i(P,"SPAN",{});var Y=s(b);q=p(Y,"Overview"),Y.forEach(t),P.forEach(t),O=N(e),_=i(e,"P",{});var B=s(_);z=p(B,"Generating high-quality outputs is computationally intensive, especially during each iterative step where you go from a noisy output to a less noisy output. One of \u{1F917} Diffuser\u2019s goal is to make this technology widely accessible to everyone, which includes enabling fast inference on consumer and specialized hardware."),B.forEach(t),$=N(e),a=i(e,"P",{});var d=s(a);I=p(d,"This section will cover tips and tricks - like half-precision weights and sliced attention - for optimizing inference speed and reducing memory-consumption. You\u2019ll also learn how to speed up your PyTorch code with "),v=i(d,"A",{href:!0,rel:!0});var J=s(v);E=i(J,"CODE",{});var U=s(E);C=p(U,"torch.compile"),U.forEach(t),J.forEach(t),D=p(d," or "),w=i(d,"A",{href:!0,rel:!0});var j=s(w);G=p(j,"ONNX Runtime"),j.forEach(t),H=p(d,", and enable memory-efficient attention with "),y=i(d,"A",{href:!0,rel:!0});var K=s(y);L=p(K,"xFormers"),K.forEach(t),M=p(d,". There are also guides for running inference on specific hardware like Apple Silicon, and Intel or Habana processors."),d.forEach(t),this.h()},h(){r(l,"name","hf:doc:metadata"),r(l,"content",JSON.stringify(he)),r(f,"id","overview"),r(f,"class","header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full"),r(f,"href","#overview"),r(c,"class","relative group"),r(v,"href","https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html"),r(v,"rel","nofollow"),r(w,"href","https://onnxruntime.ai/docs/"),r(w,"rel","nofollow"),r(y,"href","https://facebookresearch.github.io/xformers/"),r(y,"rel","nofollow")},m(e,h){o(document.head,l),g(e,x,h),g(e,c,h),o(c,f),o(f,k),oe(m,k,null),o(c,T),o(c,b),o(b,q),g(e,O,h),g(e,_,h),o(_,z),g(e,$,h),g(e,a,h),o(a,I),o(a,v),o(v,E),o(E,C),o(a,D),o(a,w),o(w,G),o(a,H),o(a,y),o(y,L),o(a,M),A=!0},p:ae,i(e){A||(re(m.$$.fragment,e),A=!0)},o(e){ne(m.$$.fragment,e),A=!1},d(e){t(l),e&&t(x),e&&t(c),ie(m),e&&t(O),e&&t(_),e&&t($),e&&t(a)}}}const he={local:"overview",title:"Overview"};function ue(R){return se(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class de extends Q{constructor(l){super();V(this,l,ue,ce,W,{})}}export{de as default,he as metadata};