diffusers/pr_12403/ko/_app/immutable/nodes/44.347fedf2.js

import{s as zt,o as Pt,n as ze}from"../chunks/scheduler.94020406.js";import{S as Ot,i as Kt,g as o,s as a,r as c,E as es,h as M,f as t,c as i,j as qt,u as f,x as m,k as F,y as ls,a as s,v as d,d as J,t as T,w as h,m as ts,n as ss}from"../chunks/index.a08c8d92.js";import{T as De}from"../chunks/Tip.7c41dfd5.js";import{C as V}from"../chunks/CodeBlock.f1fae7de.js";import{D as as}from"../chunks/DocNotebookDropdown.a1753374.js";import{H as G,E as is}from"../chunks/getInferenceSnippets.8ef6721d.js";import{H as ns,a as Dt}from"../chunks/HfOption.df48b824.js";function ps(w){let n,C='사용할 체크포인트와 임베딩은 <a href="https://huggingface.co/spaces/sd-concepts-library/stable-diffusion-conceptualizer" rel="nofollow">Stable Diffusion Conceptualizer</a>, <a href="https://huggingface.co/spaces/multimodalart/LoraTheExplorer" rel="nofollow">LoRA the Explorer</a>, <a href="https://huggingface.co/spaces/huggingface-projects/diffusers-gallery" rel="nofollow">Diffusers Models Gallery</a>에서 찾아보시기 바랍니다.';return{c(){n=o("p"),n.innerHTML=C},l(r){n=M(r,"P",{"data-svelte-h":!0}),m(n)!=="svelte-1nuqq3"&&(n.innerHTML=C)},m(r,u){s(r,n,u)},p:ze,d(r){r&&t(n)}}}function os(w){let n,C='LoRA는 다른 학습 방법과 함께 사용할 수 있는 매우 일반적인 학습 기법입니다. 예를 들어, DreamBooth와 LoRA로 모델을 학습하는 것이 일반적입니다. 또한 새롭고 고유한 이미지를 생성하기 위해 여러 개의 LoRA를 불러오고 병합하는 것이 점점 더 일반화되고 있습니다. 병합은 이 불러오기 가이드의 범위를 벗어나므로 자세한 내용은 심층적인 <a href="merge_loras">LoRA 병합</a> 가이드에서 확인할 수 있습니다.';return{c(){n=o("p"),n.innerHTML=C},l(r){n=M(r,"P",{"data-svelte-h":!0}),m(n)!=="svelte-wtg3ww"&&(n.innerHTML=C)},m(r,u){s(r,n,u)},p:ze,d(r){r&&t(n)}}}function Ms(w){let n,C="현재 <code>set_adapters()</code>는 어텐션 가중치의 스케일링만 지원합니다. LoRA에 다른 부분(예: resnets or down-/upsamplers)이 있는 경우 1.0의 스케일을 유지합니다.";return{c(){n=o("p"),n.innerHTML=C},l(r){n=M(r,"P",{"data-svelte-h":!0}),m(n)!=="svelte-124qd55"&&(n.innerHTML=C)},m(r,u){s(r,n,u)},p:ze,d(r){r&&t(n)}}}function rs(w){let n,C="Kohya LoRA를 🤗 Diffusers와 함께 사용할 때 몇 가지 제한 사항이 있습니다:",r,u,U='<li><a href="https://github.com/huggingface/diffusers/pull/4287/#issuecomment-1655110736" rel="nofollow">여기</a>에 설명된 여러 가지 이유로 인해 이미지가 ComfyUI와 같은 UI에서 생성된 이미지와 다르게 보일 수 있습니다.</li> <li><a href="https://github.com/KohakuBlueleaf/LyCORIS" rel="nofollow">LyCORIS 체크포인트</a>가 완전히 지원되지 않습니다. <code>load_lora_weights()</code> 메서드는 LoRA 및 LoCon 모듈로 LyCORIS 체크포인트를 불러올 수 있지만, Hada 및 LoKR은 지원되지 않습니다.</li>';return{c(){n=o("p"),n.textContent=C,r=a(),u=o("ul"),u.innerHTML=U},l(y){n=M(y,"P",{"data-svelte-h":!0}),m(n)!=="svelte-dgiwss"&&(n.textContent=C),r=i(y),u=M(y,"UL",{"data-svelte-h":!0}),m(u)!=="svelte-ztryq"&&(u.innerHTML=U)},m(y,g){s(y,n,g),s(y,r,g),s(y,u,g)},p:ze,d(y){y&&(t(n),t(r),t(u))}}}function ms(w){let n,C='Kohya LoRA를 불러오기 위해, 예시로 <a href="https://civitai.com/" rel="nofollow">Civitai</a>에서 <a href="https://civitai.com/models/150986/blueprintify-sd-xl-10" rel="nofollow">Blueprintify SD XL 1.0</a> 체크포인트를 다운로드합니다:',r,u,U,y,g="LoRA 체크포인트를 <code>load_lora_weights()</code> 메서드로 불러오고 <code>weight_name</code> 파라미터에 파일명을 지정합니다:",Z,j,Y,I,W="이미지를 생성합니다:",q,B,Q,$,R;return u=new V({props:{code:"IXdnZXQlMjBodHRwcyUzQSUyRiUyRmNpdml0YWkuY29tJTJGYXBpJTJGZG93bmxvYWQlMkZtb2RlbHMlMkYxNjg3NzYlMjAtTyUyMGJsdWVwcmludGlmeS1zZC14bC0xMC5zYWZldGVuc29ycw==",highlighted:"!wget https://civitai.com/api/download/models/168776 -O blueprintify-sd-xl-10.safetensors",wrap:!1}}),j=new V({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9QaXBlbGluZUZvclRleHQySW1hZ2UlMEFpbXBvcnQlMjB0b3JjaCUwQSUwQXBpcGVsaW5lJTIwJTNEJTIwQXV0b1BpcGVsaW5lRm9yVGV4dDJJbWFnZS5mcm9tX3ByZXRyYWluZWQoJTIyc3RhYmlsaXR5YWklMkZzdGFibGUtZGlmZnVzaW9uLXhsLWJhc2UtMS4wJTIyJTJDJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDE2KS50byglMjJjdWRhJTIyKSUwQXBpcGVsaW5lLmxvYWRfbG9yYV93ZWlnaHRzKCUyMnBhdGglMkZ0byUyRndlaWdodHMlMjIlMkMlMjB3ZWlnaHRfbmFtZSUzRCUyMmJsdWVwcmludGlmeS1zZC14bC0xMC5zYWZldGVuc29ycyUyMik=",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)
pipeline.load_lora_weights(<span class="hljs-string">&quot;path/to/weights&quot;</span>, weight_name=<span class="hljs-string">&quot;blueprintify-sd-xl-10.safetensors&quot;</span>)`,wrap:!1}}),B=new V({props:{code:"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",highlighted:`<span class="hljs-comment"># LoRA를 트리거하기 위해 bl3uprint를 프롬프트에 사용</span>
prompt = <span class="hljs-string">&quot;bl3uprint, a highly detailed blueprint of the eiffel tower, explaining how to build all parts, many txt, blueprint grid backdrop&quot;</span>
image = pipeline(prompt).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),$=new De({props:{warning:!0,$$slots:{default:[rs]},$$scope:{ctx:w}}}),{c(){n=o("p"),n.innerHTML=C,r=a(),c(u.$$.fragment),U=a(),y=o("p"),y.innerHTML=g,Z=a(),c(j.$$.fragment),Y=a(),I=o("p"),I.textContent=W,q=a(),c(B.$$.fragment),Q=a(),c($.$$.fragment)},l(p){n=M(p,"P",{"data-svelte-h":!0}),m(n)!=="svelte-impx4k"&&(n.innerHTML=C),r=i(p),f(u.$$.fragment,p),U=i(p),y=M(p,"P",{"data-svelte-h":!0}),m(y)!=="svelte-zii06v"&&(y.innerHTML=g),Z=i(p),f(j.$$.fragment,p),Y=i(p),I=M(p,"P",{"data-svelte-h":!0}),m(I)!=="svelte-1ouwme5"&&(I.textContent=W),q=i(p),f(B.$$.fragment,p),Q=i(p),f($.$$.fragment,p)},m(p,b){s(p,n,b),s(p,r,b),d(u,p,b),s(p,U,b),s(p,y,b),s(p,Z,b),d(j,p,b),s(p,Y,b),s(p,I,b),s(p,q,b),d(B,p,b),s(p,Q,b),d($,p,b),R=!0},p(p,b){const Ye={};b&2&&(Ye.$$scope={dirty:b,ctx:p}),$.$set(Ye)},i(p){R||(J(u.$$.fragment,p),J(j.$$.fragment,p),J(B.$$.fragment,p),J($.$$.fragment,p),R=!0)},o(p){T(u.$$.fragment,p),T(j.$$.fragment,p),T(B.$$.fragment,p),T($.$$.fragment,p),R=!1},d(p){p&&(t(n),t(r),t(U),t(y),t(Z),t(Y),t(I),t(q),t(Q)),h(u,p),h(j,p),h(B,p),h($,p)}}}function ys(w){let n,C='TheLastBen에서 체크포인트를 불러오는 방법은 매우 유사합니다. 예를 들어, <a href="https://huggingface.co/TheLastBen/William_Eggleston_Style_SDXL" rel="nofollow">TheLastBen/William_Eggleston_Style_SDXL</a> 체크포인트를 불러오려면:',r,u,U;return u=new V({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)
pipeline.load_lora_weights(<span class="hljs-string">&quot;TheLastBen/William_Eggleston_Style_SDXL&quot;</span>, weight_name=<span class="hljs-string">&quot;wegg.safetensors&quot;</span>)

<span class="hljs-comment"># LoRA를 트리거하기 위해 william eggleston를 프롬프트에 사용</span>
prompt = <span class="hljs-string">&quot;a house by william eggleston, sunrays, beautiful, sunlight, sunrays, beautiful&quot;</span>
image = pipeline(prompt=prompt).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),{c(){n=o("p"),n.innerHTML=C,r=a(),c(u.$$.fragment)},l(y){n=M(y,"P",{"data-svelte-h":!0}),m(n)!=="svelte-fle9ti"&&(n.innerHTML=C),r=i(y),f(u.$$.fragment,y)},m(y,g){s(y,n,g),s(y,r,g),d(u,y,g),U=!0},p:ze,i(y){U||(J(u.$$.fragment,y),U=!0)},o(y){T(u.$$.fragment,y),U=!1},d(y){y&&(t(n),t(r)),h(u,y)}}}function Us(w){let n,C,r,u;return n=new Dt({props:{id:"other-trainers",option:"Kohya",$$slots:{default:[ms]},$$scope:{ctx:w}}}),r=new Dt({props:{id:"other-trainers",option:"TheLastBen",$$slots:{default:[ys]},$$scope:{ctx:w}}}),{c(){c(n.$$.fragment),C=a(),c(r.$$.fragment)},l(U){f(n.$$.fragment,U),C=i(U),f(r.$$.fragment,U)},m(U,y){d(n,U,y),s(U,C,y),d(r,U,y),u=!0},p(U,y){const g={};y&2&&(g.$$scope={dirty:y,ctx:U}),n.$set(g);const Z={};y&2&&(Z.$$scope={dirty:y,ctx:U}),r.$set(Z)},i(U){u||(J(n.$$.fragment,U),J(r.$$.fragment,U),u=!0)},o(U){T(n.$$.fragment,U),T(r.$$.fragment,U),u=!1},d(U){U&&t(C),h(n,U),h(r,U)}}}function us(w){let n;return{c(){n=ts("InsightFace 사전학습된 모델은 비상업적 연구 목적으로만 사용할 수 있으므로, IP-Adapter-FaceID 모델은 연구 목적으로만 릴리즈되었으며 상업적 용도로는 사용할 수 없습니다.")},l(C){n=ss(C,"InsightFace 사전학습된 모델은 비상업적 연구 목적으로만 사용할 수 있으므로, IP-Adapter-FaceID 모델은 연구 목적으로만 릴리즈되었으며 상업적 용도로는 사용할 수 없습니다.")},m(C,r){s(C,n,r)},d(C){C&&t(n)}}}function cs(w){let n,C,r,u,U,y,g,Z,j,Y='특정 물체의 이미지 또는 특정 스타일의 이미지를 생성하도록 diffusion 모델을 개인화하기 위한 몇 가지 <a href="../training/overview">학습</a> 기법이 있습니다. 이러한 학습 방법은 각각 다른 유형의 어댑터를 생성합니다. 일부 어댑터는 완전히 새로운 모델을 생성하는 반면, 다른 어댑터는 임베딩 또는 가중치의 작은 부분만 수정합니다. 이는 각 어댑터의 로딩 프로세스도 다르다는 것을 의미합니다.',I,W,q="이 가이드에서는 DreamBooth, textual inversion 및 LoRA 가중치를 불러오는 방법을 설명합니다.",B,Q,$,R,p,b,Ye='<a href="https://dreambooth.github.io/" rel="nofollow">DreamBooth</a>는 물체의 여러 이미지에 대한 <em>diffusion 모델 전체</em>를 미세 조정하여 새로운 스타일과 설정으로 해당 물체의 이미지를 생성합니다. 이 방법은 모델이 물체 이미지와 연관시키는 방법을 학습하는 프롬프트에 특수 단어를 사용하는 방식으로 작동합니다. 모든 학습 방법 중에서 드림부스는 전체 체크포인트 모델이기 때문에 파일 크기가 가장 큽니다(보통 몇 GB).',Pe,D,ot='Hergé가 그린 단 10개의 이미지로 학습된 <a href="https://huggingface.co/sd-dreambooth-library/herge-style" rel="nofollow">herge_style</a> 체크포인트를 불러와 해당 스타일의 이미지를 생성해 보겠습니다. 이 모델이 작동하려면 체크포인트를 트리거하는 프롬프트에 특수 단어 <code>herge_style</code>을 포함시켜야 합니다:',Oe,z,Ke,v,Mt='<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_dreambooth.png"/>',el,P,ll,O,rt='<a href="https://textual-inversion.github.io/" rel="nofollow">Textual inversion</a>은 DreamBooth와 매우 유사하며 몇 개의 이미지만으로 특정 개념(스타일, 개체)을 생성하는 diffusion 모델을 개인화할 수도 있습니다. 이 방법은 프롬프트에 특정 단어를 입력하면 해당 이미지를 나타내는 새로운 임베딩을 학습하고 찾아내는 방식으로 작동합니다. 결과적으로 diffusion 모델 가중치는 동일하게 유지되고 훈련 프로세스는 비교적 작은(수 KB) 파일을 생성합니다.',tl,K,mt="Textual inversion은 임베딩을 생성하기 때문에 DreamBooth처럼 단독으로 사용할 수 없으며 또 다른 모델이 필요합니다.",sl,ee,al,le,yt='이제 <code>load_textual_inversion()</code> 메서드를 사용하여 textual inversion 임베딩을 불러와 이미지를 생성할 수 있습니다. <a href="https://huggingface.co/sd-concepts-library/gta5-artwork" rel="nofollow">sd-concepts-library/gta5-artwork</a> 임베딩을 불러와 보겠습니다. 이를 트리거하려면 프롬프트에 특수 단어 <code>&lt;gta5-artwork&gt;</code>를 포함시켜야 합니다:',il,te,nl,_,Ut='<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_txt_embed.png"/>',pl,se,ut="Textual inversion은 또한 바람직하지 않은 사물에 대해 <em>네거티브 임베딩</em>을 생성하여 모델이 흐릿한 이미지나 손의 추가 손가락과 같은 바람직하지 않은 사물이 포함된 이미지를 생성하지 못하도록 학습할 수도 있습니다. 이는 프롬프트를 빠르게 개선하는 것이 쉬운 방법이 될 수 있습니다. 이는 이전과 같이 임베딩을 <code>load_textual_inversion()</code>으로 불러오지만 이번에는 두 개의 매개변수가 더 필요합니다:",ol,ae,ct="<li><code>weight_name</code>: 파일이 특정 이름의 🤗 Diffusers 형식으로 저장된 경우이거나 파일이 A1111 형식으로 저장된 경우, 불러올 가중치 파일을 지정합니다.</li> <li><code>token</code>: 임베딩을 트리거하기 위해 프롬프트에서 사용할 특수 단어를 지정합니다.</li>",Ml,ie,ft='<a href="https://huggingface.co/sayakpaul/EasyNegative-test" rel="nofollow">sayakpaul/EasyNegative-test</a> 임베딩을 불러와 보겠습니다:',rl,ne,ml,pe,dt="이제 <code>token</code>을 사용해 네거티브 임베딩이 있는 이미지를 생성할 수 있습니다:",yl,oe,Ul,X,Jt='<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_neg_embed.png"/>',ul,Me,cl,re,Tt='<a href="https://huggingface.co/papers/2106.09685" rel="nofollow">Low-Rank Adaptation (LoRA)</a>은 속도가 빠르고 파일 크기가 (수백 MB로) 작기 때문에 널리 사용되는 학습 기법입니다. 이 가이드의 다른 방법과 마찬가지로, LoRA는 몇 장의 이미지만으로 새로운 스타일을 학습하도록 모델을 학습시킬 수 있습니다. 이는 diffusion 모델에 새로운 가중치를 삽입한 다음 전체 모델 대신 새로운 가중치만 학습시키는 방식으로 작동합니다. 따라서 LoRA를 더 빠르게 학습시키고 더 쉽게 저장할 수 있습니다.',fl,k,dl,me,ht="LoRA는 다른 모델과 함께 사용해야 합니다:",Jl,ye,Tl,Ue,Ct='그리고 <code>load_lora_weights()</code> 메서드를 사용하여 <a href="https://huggingface.co/ostris/super-cereal-sdxl-lora" rel="nofollow">ostris/super-cereal-sdxl-lora</a> 가중치를 불러오고 리포지토리에서 가중치 파일명을 지정합니다:',hl,ue,Cl,L,bt='<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_lora.png"/>',bl,ce,gt="<code>load_lora_weights()</code> 메서드는 LoRA 가중치를 UNet과 텍스트 인코더에 모두 불러옵니다. 이 메서드는 해당 케이스에서 LoRA를 불러오는 데 선호되는 방식입니다:",gl,fe,Vt="<li>LoRA 가중치에 UNet 및 텍스트 인코더에 대한 별도의 식별자가 없는 경우</li> <li>LoRA 가중치에 UNet과 텍스트 인코더에 대한 별도의 식별자가 있는 경우</li>",Vl,de,wt='하지만 LoRA 가중치만 UNet에 로드해야 하는 경우에는 <code>load_attn_procs()</code> 메서드를 사용할 수 있습니다. <a href="https://huggingface.co/jbilcke-hf/sdxl-cinematic-1" rel="nofollow">jbilcke-hf/sdxl-cinematic-1</a> LoRA를 불러와 보겠습니다:',wl,Je,jl,x,jt='<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_attn_proc.png"/>',Ql,Te,Qt="LoRA 가중치를 언로드하려면 <code>unload_lora_weights()</code> 메서드를 사용하여 LoRA 가중치를 삭제하고 모델을 원래 가중치로 복원합니다:",$l,he,Zl,Ce,Il,be,$t="<code>load_lora_weights()</code> 및 <code>load_attn_procs()</code> 모두 <code>cross_attention_kwargs={&quot;scale&quot;: 0.5}</code> 파라미터를 전달하여 얼마나 LoRA 가중치를 사용할지 조정할 수 있습니다. 값이 <code>0</code>이면 기본 모델 가중치만 사용하는 것과 같고, 값이 <code>1</code>이면 완전히 미세 조정된 LoRA를 사용하는 것과 같습니다.",Bl,ge,Zt="레이어당 사용되는 LoRA 가중치의 양을 보다 세밀하게 제어하려면 <code>set_adapters()</code>를 사용하여 각 레이어의 가중치를 얼마만큼 조정할지 지정하는 딕셔너리를 전달할 수 있습니다.",Rl,Ve,Wl,we,It='이는 여러 어댑터에서도 작동합니다. 방법은 <a href="https://huggingface.co/docs/diffusers/tutorials/using_peft_for_inference#customize-adapters-strength" rel="nofollow">이 가이드</a>를 참조하세요.',Fl,S,Gl,je,vl,Qe,Bt='커뮤니티에서 인기 있는 다른 LoRA trainer로는 <a href="https://github.com/kohya-ss/sd-scripts/" rel="nofollow">Kohya</a>와 <a href="https://github.com/TheLastBen/fast-stable-diffusion" rel="nofollow">TheLastBen</a>의 trainer가 있습니다. 이 trainer들은 🤗 Diffusers가 훈련한 것과는 다른 LoRA 체크포인트를 생성하지만, 같은 방식으로 불러올 수 있습니다.',_l,N,Xl,$e,kl,Ze,Rt='<a href="https://ip-adapter.github.io/" rel="nofollow">IP-Adapter</a>는 모든 diffusion 모델에 이미지 프롬프트를 사용할 수 있는 경량 어댑터입니다. 이 어댑터는 이미지와 텍스트 feature의 cross-attention 레이어를 분리하여 작동합니다. 다른 모든 모델 컴포넌트튼 freeze되고 UNet의 embedded 이미지 features만 학습됩니다. 따라서 IP-Adapter 파일은 일반적으로 최대 100MB에 불과합니다.',Ll,Ie,Wt='다양한 작업과 구체적인 사용 사례에 IP-Adapter를 사용하는 방법에 대한 자세한 내용은 <a href="../using-diffusers/ip_adapter">IP-Adapter</a> 가이드에서 확인할 수 있습니다.',xl,H,Ft=`<p>Diffusers는 현재 가장 많이 사용되는 일부 파이프라인에 대해서만 IP-Adapter를 지원합니다. 멋진 사용 사례가 있는 지원되지 않는 파이프라인에 IP-Adapter를 통합하고 싶다면 언제든지 기능 요청을 여세요!
공식 IP-Adapter 체크포인트는 <a href="https://huggingface.co/h94/IP-Adapter" rel="nofollow">h94/IP-Adapter</a>에서 확인할 수 있습니다.</p>`,Sl,Be,Gt="시작하려면 Stable Diffusion 체크포인트를 불러오세요.",Nl,Re,Hl,We,vt="그런 다음 IP-Adapter 가중치를 불러와 <code>load_ip_adapter()</code> 메서드를 사용하여 파이프라인에 추가합니다.",Al,Fe,El,Ge,_t="불러온 뒤, 이미지 및 텍스트 프롬프트가 있는 파이프라인을 사용하여 이미지 생성 프로세스를 가이드할 수 있습니다.",Yl,ve,ql,A,Xt='    <img src="https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/ip-bear.png"/>',Dl,_e,zl,Xe,kt="IP-Adapter는 이미지 인코더를 사용하여 이미지 feature를 생성합니다. IP-Adapter 리포지토리에 <code>image_encoder</code> 하위 폴더가 있는 경우, 이미지 인코더가 자동으로 불러와 파이프라인에 등록됩니다. 그렇지 않은 경우, <code>CLIPVisionModelWithProjection</code> 모델을 사용하여 이미지 인코더를 명시적으로 불러와 파이프라인에 전달해야 합니다.",Pl,ke,Lt="이는 ViT-H 이미지 인코더를 사용하는 <em>IP-Adapter Plus</em> 체크포인트에 해당하는 케이스입니다.",Ol,Le,Kl,xe,et,Se,xt=`IP-Adapter FaceID 모델은 CLIP 이미지 임베딩 대신 <code>insightface</code>에서 생성한 이미지 임베딩을 사용하는 실험적인 IP Adapter입니다. 이러한 모델 중 일부는 LoRA를 사용하여 ID 일관성을 개선하기도 합니다.
이러한 모델을 사용하려면 <code>insightface</code>와 해당 요구 사항을 모두 설치해야 합니다.`,lt,E,tt,Ne,st,He,St="두 가지 IP 어댑터 FaceID Plus 모델 중 하나를 사용하려는 경우, 이 모델들은 더 나은 사실감을 얻기 위해 <code>insightface</code>와 CLIP 이미지 임베딩을 모두 사용하므로, CLIP 이미지 인코더도 불러와야 합니다.",at,Ae,it,Ee,nt,qe,pt;return U=new G({props:{title:"어댑터 불러오기",local:"어댑터-불러오기",headingTag:"h1"}}),g=new as({props:{classNames:"absolute z-10 right-0 top-0",options:[{label:"Mixed",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/ko/loading_adapters.ipynb"},{label:"PyTorch",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/ko/pytorch/loading_adapters.ipynb"},{label:"TensorFlow",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/ko/tensorflow/loading_adapters.ipynb"},{label:"Mixed",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/ko/loading_adapters.ipynb"},{label:"PyTorch",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/ko/pytorch/loading_adapters.ipynb"},{label:"TensorFlow",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/ko/tensorflow/loading_adapters.ipynb"}]}}),Q=new De({props:{$$slots:{default:[ps]},$$scope:{ctx:w}}}),R=new G({props:{title:"DreamBooth",local:"dreambooth",headingTag:"h2"}}),z=new V({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;sd-dreambooth-library/herge-style&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)
prompt = <span class="hljs-string">&quot;A cute herge_style brown bear eating a slice of pizza, stunning color scheme, masterpiece, illustration&quot;</span>
image = pipeline(prompt).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),P=new G({props:{title:"Textual inversion",local:"textual-inversion",headingTag:"h2"}}),ee=new V({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9QaXBlbGluZUZvclRleHQySW1hZ2UlMEFpbXBvcnQlMjB0b3JjaCUwQSUwQXBpcGVsaW5lJTIwJTNEJTIwQXV0b1BpcGVsaW5lRm9yVGV4dDJJbWFnZS5mcm9tX3ByZXRyYWluZWQoJTIyc3RhYmxlLWRpZmZ1c2lvbi12MS01JTJGc3RhYmxlLWRpZmZ1c2lvbi12MS01JTIyJTJDJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDE2KS50byglMjJjdWRhJTIyKQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stable-diffusion-v1-5/stable-diffusion-v1-5&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)`,wrap:!1}}),te=new V({props:{code:"cGlwZWxpbmUubG9hZF90ZXh0dWFsX2ludmVyc2lvbiglMjJzZC1jb25jZXB0cy1saWJyYXJ5JTJGZ3RhNS1hcnR3b3JrJTIyKSUwQXByb21wdCUyMCUzRCUyMCUyMkElMjBjdXRlJTIwYnJvd24lMjBiZWFyJTIwZWF0aW5nJTIwYSUyMHNsaWNlJTIwb2YlMjBwaXp6YSUyQyUyMHN0dW5uaW5nJTIwY29sb3IlMjBzY2hlbWUlMkMlMjBtYXN0ZXJwaWVjZSUyQyUyMGlsbHVzdHJhdGlvbiUyQyUyMCUzQ2d0YTUtYXJ0d29yayUzRSUyMHN0eWxlJTIyJTBBaW1hZ2UlMjAlM0QlMjBwaXBlbGluZShwcm9tcHQpLmltYWdlcyU1QjAlNUQlMEFpbWFnZQ==",highlighted:`pipeline.load_textual_inversion(<span class="hljs-string">&quot;sd-concepts-library/gta5-artwork&quot;</span>)
prompt = <span class="hljs-string">&quot;A cute brown bear eating a slice of pizza, stunning color scheme, masterpiece, illustration, &lt;gta5-artwork&gt; style&quot;</span>
image = pipeline(prompt).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),ne=new V({props:{code:"cGlwZWxpbmUubG9hZF90ZXh0dWFsX2ludmVyc2lvbiglMEElMjAlMjAlMjAlMjAlMjJzYXlha3BhdWwlMkZFYXN5TmVnYXRpdmUtdGVzdCUyMiUyQyUyMHdlaWdodF9uYW1lJTNEJTIyRWFzeU5lZ2F0aXZlLnNhZmV0ZW5zb3JzJTIyJTJDJTIwdG9rZW4lM0QlMjJFYXN5TmVnYXRpdmUlMjIlMEEp",highlighted:`pipeline.load_textual_inversion(
    <span class="hljs-string">&quot;sayakpaul/EasyNegative-test&quot;</span>, weight_name=<span class="hljs-string">&quot;EasyNegative.safetensors&quot;</span>, token=<span class="hljs-string">&quot;EasyNegative&quot;</span>
)`,wrap:!1}}),oe=new V({props:{code:"cHJvbXB0JTIwJTNEJTIwJTIyQSUyMGN1dGUlMjBicm93biUyMGJlYXIlMjBlYXRpbmclMjBhJTIwc2xpY2UlMjBvZiUyMHBpenphJTJDJTIwc3R1bm5pbmclMjBjb2xvciUyMHNjaGVtZSUyQyUyMG1hc3RlcnBpZWNlJTJDJTIwaWxsdXN0cmF0aW9uJTJDJTIwRWFzeU5lZ2F0aXZlJTIyJTBBbmVnYXRpdmVfcHJvbXB0JTIwJTNEJTIwJTIyRWFzeU5lZ2F0aXZlJTIyJTBBJTBBaW1hZ2UlMjAlM0QlMjBwaXBlbGluZShwcm9tcHQlMkMlMjBuZWdhdGl2ZV9wcm9tcHQlM0RuZWdhdGl2ZV9wcm9tcHQlMkMlMjBudW1faW5mZXJlbmNlX3N0ZXBzJTNENTApLmltYWdlcyU1QjAlNUQlMEFpbWFnZQ==",highlighted:`prompt = <span class="hljs-string">&quot;A cute brown bear eating a slice of pizza, stunning color scheme, masterpiece, illustration, EasyNegative&quot;</span>
negative_prompt = <span class="hljs-string">&quot;EasyNegative&quot;</span>

image = pipeline(prompt, negative_prompt=negative_prompt, num_inference_steps=<span class="hljs-number">50</span>).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),Me=new G({props:{title:"LoRA",local:"lora",headingTag:"h2"}}),k=new De({props:{$$slots:{default:[os]},$$scope:{ctx:w}}}),ye=new V({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9QaXBlbGluZUZvclRleHQySW1hZ2UlMEFpbXBvcnQlMjB0b3JjaCUwQSUwQXBpcGVsaW5lJTIwJTNEJTIwQXV0b1BpcGVsaW5lRm9yVGV4dDJJbWFnZS5mcm9tX3ByZXRyYWluZWQoJTIyc3RhYmlsaXR5YWklMkZzdGFibGUtZGlmZnVzaW9uLXhsLWJhc2UtMS4wJTIyJTJDJTIwdG9yY2hfZHR5cGUlM0R0b3JjaC5mbG9hdDE2KS50byglMjJjdWRhJTIyKQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)`,wrap:!1}}),ue=new V({props:{code:"cGlwZWxpbmUubG9hZF9sb3JhX3dlaWdodHMoJTIyb3N0cmlzJTJGc3VwZXItY2VyZWFsLXNkeGwtbG9yYSUyMiUyQyUyMHdlaWdodF9uYW1lJTNEJTIyY2VyZWFsX2JveF9zZHhsX3YxLnNhZmV0ZW5zb3JzJTIyKSUwQXByb21wdCUyMCUzRCUyMCUyMmJlYXJzJTJDJTIwcGl6emElMjBiaXRlcyUyMiUwQWltYWdlJTIwJTNEJTIwcGlwZWxpbmUocHJvbXB0KS5pbWFnZXMlNUIwJTVEJTBBaW1hZ2U=",highlighted:`pipeline.load_lora_weights(<span class="hljs-string">&quot;ostris/super-cereal-sdxl-lora&quot;</span>, weight_name=<span class="hljs-string">&quot;cereal_box_sdxl_v1.safetensors&quot;</span>)
prompt = <span class="hljs-string">&quot;bears, pizza bites&quot;</span>
image = pipeline(prompt).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),Je=new V({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)
pipeline.unet.load_attn_procs(<span class="hljs-string">&quot;jbilcke-hf/sdxl-cinematic-1&quot;</span>, weight_name=<span class="hljs-string">&quot;pytorch_lora_weights.safetensors&quot;</span>)

<span class="hljs-comment"># 프롬프트에서 cnmt를 사용하여 LoRA를 트리거합니다.</span>
prompt = <span class="hljs-string">&quot;A cute cnmt eating a slice of pizza, stunning color scheme, masterpiece, illustration&quot;</span>
image = pipeline(prompt).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),he=new V({props:{code:"cGlwZWxpbmUudW5sb2FkX2xvcmFfd2VpZ2h0cygp",highlighted:"pipeline.unload_lora_weights()",wrap:!1}}),Ce=new G({props:{title:"LoRA 가중치 스케일 조정하기",local:"lora-가중치-스케일-조정하기",headingTag:"h3"}}),Ve=new V({props:{code:"cGlwZSUyMCUzRCUyMC4uLiUyMCUyMyUyMCVFRCU4QyU4QyVFQyU5RCVCNCVFRCU5NCU4NCVFQiU5RCVCQyVFQyU5RCVCOCUyMCVFQyU4MyU5RCVFQyU4NCVCMSUwQXBpcGUubG9hZF9sb3JhX3dlaWdodHMoLi4uJTJDJTIwYWRhcHRlcl9uYW1lJTNEJTIybXlfYWRhcHRlciUyMiklMEFzY2FsZXMlMjAlM0QlMjAlN0IlMEElMjAlMjAlMjAlMjAlMjJ0ZXh0X2VuY29kZXIlMjIlM0ElMjAwLjUlMkMlMEElMjAlMjAlMjAlMjAlMjJ0ZXh0X2VuY29kZXJfMiUyMiUzQSUyMDAuNSUyQyUyMCUyMCUyMyUyMCVFRCU4QyU4QyVFQyU5RCVCNCVFRCU5NCU4NCVFQyU5NyU5MCUyMCVFQiU5MSU5MCUyMCVFQiVCMiU4OCVFQyVBNyVCOCUyMCVFRCU4NSU4RCVFQyU4QSVBNCVFRCU4QSVCOCUyMCVFQyU5RCVCOCVFQyVCRCU5NCVFQiU4RCU5NCVFQSVCMCU4MCUyMCVFQyU5RSU4OCVFQiU4QSU5NCUyMCVFQSVCMiVCRCVFQyU5QSVCMCVFQyU5NyU5MCVFQiVBNyU4QyUyMCVFQyU4MiVBQyVFQyU5QSVBOSUyMCVFQSVCMCU4MCVFQiU4QSVBNSUwQSUyMCUyMCUyMCUyMCUyMnVuZXQlMjIlM0ElMjAlN0IlMEElMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjJkb3duJTIyJTNBJTIwMC45JTJDJTIwJTIwJTIzJTIwZG93biUyMCVFQiVCNiU4MCVFQiVCNiU4NCVFQyU5RCU5OCUyMCVFQiVBQSVBOCVFQiU5MyVBMCUyMCVFRCU4QSVCOCVFQiU5RSU5QyVFQyU4QSVBNCVFRCU4RiVBQyVFQiVBOCVCOCVFQiU4QSU5NCUyMCVFQyU4QSVBNCVFQyVCQyU4MCVFQyU5RCVCQyUyMDAuOSVFQiVBNSVCQyUyMCVFQyU4MiVBQyVFQyU5QSVBOSUwQSUyMCUyMCUyMCUyMCUyMCUyMCUyMCUyMCUyMyUyMCUyMm1pZCUyMiUyMCUyMCUyMyUyMCVFQyU5RCVCNCUyMCVFQyU5OCU4OCVFQyVBMCU5QyVFQyU5NyU5MCVFQyU4NCU5QyVFQiU4QSU5NCUyMCUyMm1pZCUyMiVFQSVCMCU4MCUyMCVFQyVBNyU4MCVFQyVBMCU5NSVFQiU5MCU5OCVFQyVBNyU4MCUyMCVFQyU5NSU4QSVFQyU5NSU5OCVFQyU5QyVCQyVFQiVBRiU4MCVFQiVBMSU5QyUyMCVFQyVBNCU5MSVFQSVCMCU4NCUyMCVFQiVCNiU4MCVFQiVCNiU4NCVFQyU5RCU5OCUyMCVFQiVBQSVBOCVFQiU5MyVBMCUyMCVFRCU4QSVCOCVFQiU5RSU5QyVFQyU4QSVBNCVFRCU4RiVBQyVFQiVBOCVCOCVFQiU4QSU5NCUyMCVFQSVCOCVCMCVFQiVCMyVCOCUyMCVFQyU4QSVBNCVFQyVCQyU4MCVFQyU5RCVCQyUyMDEuMCVFQyU5RCU4NCUyMCVFQyU4MiVBQyVFQyU5QSVBOSUwQSUyMCUyMCUyMCUyMCUyMCUyMCUyMCUyMCUyMnVwJTIyJTNBJTIwJTdCJTBBJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIyYmxvY2tfMCUyMiUzQSUyMDAuNiUyQyUyMCUyMCUyMyUyMCUyMyUyMHVwJUVDJTlEJTk4JTIwMCVFQiVCMiU4OCVFQyVBNyVCOCUyMCVFQiVCOCU5NCVFQiVBMSU5RCVFQyU5NyU5MCUyMCVFQyU5RSU4OCVFQiU4QSU5NCUyMDMlRUElQjAlOUMlRUMlOUQlOTglMjAlRUQlOEElQjglRUIlOUUlOUMlRUMlOEElQTQlRUQlOEYlQUMlRUIlQTglQjglRUIlOEElOTQlMjAlRUIlQUElQTglRUIlOTElOTAlMjAlRUMlOEElQTQlRUMlQkMlODAlRUMlOUQlQkMlMjAwLjYlRUMlOUQlODQlMjAlRUMlODIlQUMlRUMlOUElQTklMEElMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjAlMjJibG9ja18xJTIyJTNBJTIwJTVCMC40JTJDJTIwMC44JTJDJTIwMS4wJTVEJTJDJTIwJTIwJTIzJTIwdXAlRUMlOUQlOTglMjAlRUMlQjIlQUIlMjAlRUIlQjIlODglRUMlQTclQjglMjAlRUIlQjglOTQlRUIlQTElOUQlRUMlOTclOTAlMjAlRUMlOUUlODglRUIlOEElOTQlMjAzJUVBJUIwJTlDJUVDJTlEJTk4JTIwJUVEJThBJUI4JUVCJTlFJTlDJUVDJThBJUE0JUVEJThGJUFDJUVCJUE4JUI4JUVCJThBJTk0JTIwJUVBJUIwJTgxJUVBJUIwJTgxJTIwJUVDJThBJUE0JUVDJUJDJTgwJUVDJTlEJUJDJTIwMC40JTJDJTIwMC44JTJDJTIwMS4wJUVDJTlEJTg0JTIwJUVDJTgyJUFDJUVDJTlBJUE5JTBBJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTIwJTdEJTBBJTIwJTIwJTIwJTIwJTdEJTBBJTdEJTBBcGlwZS5zZXRfYWRhcHRlcnMoJTIybXlfYWRhcHRlciUyMiUyQyUyMHNjYWxlcyk=",highlighted:`pipe = ... <span class="hljs-comment"># 파이프라인 생성</span>
pipe.load_lora_weights(..., adapter_name=<span class="hljs-string">&quot;my_adapter&quot;</span>)
scales = {
    <span class="hljs-string">&quot;text_encoder&quot;</span>: <span class="hljs-number">0.5</span>,
    <span class="hljs-string">&quot;text_encoder_2&quot;</span>: <span class="hljs-number">0.5</span>,  <span class="hljs-comment"># 파이프에 두 번째 텍스트 인코더가 있는 경우에만 사용 가능</span>
    <span class="hljs-string">&quot;unet&quot;</span>: {
        <span class="hljs-string">&quot;down&quot;</span>: <span class="hljs-number">0.9</span>,  <span class="hljs-comment"># down 부분의 모든 트랜스포머는 스케일 0.9를 사용</span>
        <span class="hljs-comment"># &quot;mid&quot;  # 이 예제에서는 &quot;mid&quot;가 지정되지 않았으므로 중간 부분의 모든 트랜스포머는 기본 스케일 1.0을 사용</span>
        <span class="hljs-string">&quot;up&quot;</span>: {
            <span class="hljs-string">&quot;block_0&quot;</span>: <span class="hljs-number">0.6</span>,  <span class="hljs-comment"># # up의 0번째 블록에 있는 3개의 트랜스포머는 모두 스케일 0.6을 사용</span>
            <span class="hljs-string">&quot;block_1&quot;</span>: [<span class="hljs-number">0.4</span>, <span class="hljs-number">0.8</span>, <span class="hljs-number">1.0</span>],  <span class="hljs-comment"># up의 첫 번째 블록에 있는 3개의 트랜스포머는 각각 스케일 0.4, 0.8, 1.0을 사용</span>
        }
    }
}
pipe.set_adapters(<span class="hljs-string">&quot;my_adapter&quot;</span>, scales)`,wrap:!1}}),S=new De({props:{warning:!0,$$slots:{default:[Ms]},$$scope:{ctx:w}}}),je=new G({props:{title:"Kohya와 TheLastBen",local:"kohya와-thelastben",headingTag:"h3"}}),N=new ns({props:{id:"other-trainers",options:["Kohya","TheLastBen"],$$slots:{default:[Us]},$$scope:{ctx:w}}}),$e=new G({props:{title:"IP-Adapter",local:"ip-adapter",headingTag:"h2"}}),Re=new V({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9QaXBlbGluZUZvclRleHQySW1hZ2UlMEFpbXBvcnQlMjB0b3JjaCUwQWZyb20lMjBkaWZmdXNlcnMudXRpbHMlMjBpbXBvcnQlMjBsb2FkX2ltYWdlJTBBJTBBcGlwZWxpbmUlMjAlM0QlMjBBdXRvUGlwZWxpbmVGb3JUZXh0MkltYWdlLmZyb21fcHJldHJhaW5lZCglMjJzdGFibGUtZGlmZnVzaW9uLXYxLTUlMkZzdGFibGUtZGlmZnVzaW9uLXYxLTUlMjIlMkMlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmZsb2F0MTYpLnRvKCUyMmN1ZGElMjIp",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> load_image

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stable-diffusion-v1-5/stable-diffusion-v1-5&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)`,wrap:!1}}),Fe=new V({props:{code:"cGlwZWxpbmUubG9hZF9pcF9hZGFwdGVyKCUyMmg5NCUyRklQLUFkYXB0ZXIlMjIlMkMlMjBzdWJmb2xkZXIlM0QlMjJtb2RlbHMlMjIlMkMlMjB3ZWlnaHRfbmFtZSUzRCUyMmlwLWFkYXB0ZXJfc2QxNS5iaW4lMjIp",highlighted:'pipeline.load_ip_adapter(<span class="hljs-string">&quot;h94/IP-Adapter&quot;</span>, subfolder=<span class="hljs-string">&quot;models&quot;</span>, weight_name=<span class="hljs-string">&quot;ip-adapter_sd15.bin&quot;</span>)',wrap:!1}}),ve=new V({props:{code:"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",highlighted:`image = load_image(<span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/load_neg_embed.png&quot;</span>)
generator = torch.Generator(device=<span class="hljs-string">&quot;cpu&quot;</span>).manual_seed(<span class="hljs-number">33</span>)
images = pipeline(
    prompt=<span class="hljs-string">&#x27;best quality, high quality, wearing sunglasses&#x27;</span>,
    ip_adapter_image=image,
    negative_prompt=<span class="hljs-string">&quot;monochrome, lowres, bad anatomy, worst quality, low quality&quot;</span>,
    num_inference_steps=<span class="hljs-number">50</span>,
    generator=generator,
).images[<span class="hljs-number">0</span>]
images`,wrap:!1}}),_e=new G({props:{title:"IP-Adapter Plus",local:"ip-adapter-plus",headingTag:"h3"}}),Le=new V({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> CLIPVisionModelWithProjection

image_encoder = CLIPVisionModelWithProjection.from_pretrained(
    <span class="hljs-string">&quot;h94/IP-Adapter&quot;</span>,
    subfolder=<span class="hljs-string">&quot;models/image_encoder&quot;</span>,
    torch_dtype=torch.float16
)

pipeline = AutoPipelineForText2Image.from_pretrained(
    <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>,
    image_encoder=image_encoder,
    torch_dtype=torch.float16
).to(<span class="hljs-string">&quot;cuda&quot;</span>)

pipeline.load_ip_adapter(<span class="hljs-string">&quot;h94/IP-Adapter&quot;</span>, subfolder=<span class="hljs-string">&quot;sdxl_models&quot;</span>, weight_name=<span class="hljs-string">&quot;ip-adapter-plus_sdxl_vit-h.safetensors&quot;</span>)`,wrap:!1}}),xe=new G({props:{title:"IP-Adapter Face ID 모델",local:"ip-adapter-face-id-모델",headingTag:"h3"}}),E=new De({props:{warning:!0,$$slots:{default:[us]},$$scope:{ctx:w}}}),Ne=new V({props:{code:"cGlwZWxpbmUlMjAlM0QlMjBBdXRvUGlwZWxpbmVGb3JUZXh0MkltYWdlLmZyb21fcHJldHJhaW5lZCglMEElMjAlMjAlMjAlMjAlMjJzdGFiaWxpdHlhaSUyRnN0YWJsZS1kaWZmdXNpb24teGwtYmFzZS0xLjAlMjIlMkMlMEElMjAlMjAlMjAlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmZsb2F0MTYlMEEpLnRvKCUyMmN1ZGElMjIpJTBBJTBBcGlwZWxpbmUubG9hZF9pcF9hZGFwdGVyKCUyMmg5NCUyRklQLUFkYXB0ZXItRmFjZUlEJTIyJTJDJTIwc3ViZm9sZGVyJTNETm9uZSUyQyUyMHdlaWdodF9uYW1lJTNEJTIyaXAtYWRhcHRlci1mYWNlaWRfc2R4bC5iaW4lMjIlMkMlMjBpbWFnZV9lbmNvZGVyX2ZvbGRlciUzRE5vbmUp",highlighted:`pipeline = AutoPipelineForText2Image.from_pretrained(
    <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>,
    torch_dtype=torch.float16
).to(<span class="hljs-string">&quot;cuda&quot;</span>)

pipeline.load_ip_adapter(<span class="hljs-string">&quot;h94/IP-Adapter-FaceID&quot;</span>, subfolder=<span class="hljs-literal">None</span>, weight_name=<span class="hljs-string">&quot;ip-adapter-faceid_sdxl.bin&quot;</span>, image_encoder_folder=<span class="hljs-literal">None</span>)`,wrap:!1}}),Ae=new V({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> CLIPVisionModelWithProjection

image_encoder = CLIPVisionModelWithProjection.from_pretrained(
    <span class="hljs-string">&quot;laion/CLIP-ViT-H-14-laion2B-s32B-b79K&quot;</span>,
    torch_dtype=torch.float16,
)

pipeline = AutoPipelineForText2Image.from_pretrained(
    <span class="hljs-string">&quot;stable-diffusion-v1-5/stable-diffusion-v1-5&quot;</span>,
    image_encoder=image_encoder,
    torch_dtype=torch.float16
).to(<span class="hljs-string">&quot;cuda&quot;</span>)

pipeline.load_ip_adapter(<span class="hljs-string">&quot;h94/IP-Adapter-FaceID&quot;</span>, subfolder=<span class="hljs-literal">None</span>, weight_name=<span class="hljs-string">&quot;ip-adapter-faceid-plus_sd15.bin&quot;</span>)`,wrap:!1}}),Ee=new is({props:{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/ko/using-diffusers/loading_adapters.md"}}),{c(){n=o("meta"),C=a(),r=o("p"),u=a(),c(U.$$.fragment),y=a(),c(g.$$.fragment),Z=a(),j=o("p"),j.innerHTML=Y,I=a(),W=o("p"),W.textContent=q,B=a(),c(Q.$$.fragment),$=a(),c(R.$$.fragment),p=a(),b=o("p"),b.innerHTML=Ye,Pe=a(),D=o("p"),D.innerHTML=ot,Oe=a(),c(z.$$.fragment),Ke=a(),v=o("div"),v.innerHTML=Mt,el=a(),c(P.$$.fragment),ll=a(),O=o("p"),O.innerHTML=rt,tl=a(),K=o("p"),K.textContent=mt,sl=a(),c(ee.$$.fragment),al=a(),le=o("p"),le.innerHTML=yt,il=a(),c(te.$$.fragment),nl=a(),_=o("div"),_.innerHTML=Ut,pl=a(),se=o("p"),se.innerHTML=ut,ol=a(),ae=o("ul"),ae.innerHTML=ct,Ml=a(),ie=o("p"),ie.innerHTML=ft,rl=a(),c(ne.$$.fragment),ml=a(),pe=o("p"),pe.innerHTML=dt,yl=a(),c(oe.$$.fragment),Ul=a(),X=o("div"),X.innerHTML=Jt,ul=a(),c(Me.$$.fragment),cl=a(),re=o("p"),re.innerHTML=Tt,fl=a(),c(k.$$.fragment),dl=a(),me=o("p"),me.textContent=ht,Jl=a(),c(ye.$$.fragment),Tl=a(),Ue=o("p"),Ue.innerHTML=Ct,hl=a(),c(ue.$$.fragment),Cl=a(),L=o("div"),L.innerHTML=bt,bl=a(),ce=o("p"),ce.innerHTML=gt,gl=a(),fe=o("ul"),fe.innerHTML=Vt,Vl=a(),de=o("p"),de.innerHTML=wt,wl=a(),c(Je.$$.fragment),jl=a(),x=o("div"),x.innerHTML=jt,Ql=a(),Te=o("p"),Te.innerHTML=Qt,$l=a(),c(he.$$.fragment),Zl=a(),c(Ce.$$.fragment),Il=a(),be=o("p"),be.innerHTML=$t,Bl=a(),ge=o("p"),ge.innerHTML=Zt,Rl=a(),c(Ve.$$.fragment),Wl=a(),we=o("p"),we.innerHTML=It,Fl=a(),c(S.$$.fragment),Gl=a(),c(je.$$.fragment),vl=a(),Qe=o("p"),Qe.innerHTML=Bt,_l=a(),c(N.$$.fragment),Xl=a(),c($e.$$.fragment),kl=a(),Ze=o("p"),Ze.innerHTML=Rt,Ll=a(),Ie=o("p"),Ie.innerHTML=Wt,xl=a(),H=o("blockquote"),H.innerHTML=Ft,Sl=a(),Be=o("p"),Be.textContent=Gt,Nl=a(),c(Re.$$.fragment),Hl=a(),We=o("p"),We.innerHTML=vt,Al=a(),c(Fe.$$.fragment),El=a(),Ge=o("p"),Ge.textContent=_t,Yl=a(),c(ve.$$.fragment),ql=a(),A=o("div"),A.innerHTML=Xt,Dl=a(),c(_e.$$.fragment),zl=a(),Xe=o("p"),Xe.innerHTML=kt,Pl=a(),ke=o("p"),ke.innerHTML=Lt,Ol=a(),c(Le.$$.fragment),Kl=a(),c(xe.$$.fragment),et=a(),Se=o("p"),Se.innerHTML=xt,lt=a(),c(E.$$.fragment),tt=a(),c(Ne.$$.fragment),st=a(),He=o("p"),He.innerHTML=St,at=a(),c(Ae.$$.fragment),it=a(),c(Ee.$$.fragment),nt=a(),qe=o("p"),this.h()},l(e){const l=es("svelte-u9bgzb",document.head);n=M(l,"META",{name:!0,content:!0}),l.forEach(t),C=i(e),r=M(e,"P",{}),qt(r).forEach(t),u=i(e),f(U.$$.fragment,e),y=i(e),f(g.$$.fragment,e),Z=i(e),j=M(e,"P",{"data-svelte-h":!0}),m(j)!=="svelte-16bf5k2"&&(j.innerHTML=Y),I=i(e),W=M(e,"P",{"data-svelte-h":!0}),m(W)!=="svelte-1vi87ff"&&(W.textContent=q),B=i(e),f(Q.$$.fragment,e),$=i(e),f(R.$$.fragment,e),p=i(e),b=M(e,"P",{"data-svelte-h":!0}),m(b)!=="svelte-or9czk"&&(b.innerHTML=Ye),Pe=i(e),D=M(e,"P",{"data-svelte-h":!0}),m(D)!=="svelte-emmkw5"&&(D.innerHTML=ot),Oe=i(e),f(z.$$.fragment,e),Ke=i(e),v=M(e,"DIV",{class:!0,"data-svelte-h":!0}),m(v)!=="svelte-1qqjocv"&&(v.innerHTML=Mt),el=i(e),f(P.$$.fragment,e),ll=i(e),O=M(e,"P",{"data-svelte-h":!0}),m(O)!=="svelte-117ykpw"&&(O.innerHTML=rt),tl=i(e),K=M(e,"P",{"data-svelte-h":!0}),m(K)!=="svelte-6w30tt"&&(K.textContent=mt),sl=i(e),f(ee.$$.fragment,e),al=i(e),le=M(e,"P",{"data-svelte-h":!0}),m(le)!=="svelte-8n0ka2"&&(le.innerHTML=yt),il=i(e),f(te.$$.fragment,e),nl=i(e),_=M(e,"DIV",{class:!0,"data-svelte-h":!0}),m(_)!=="svelte-vwb4li"&&(_.innerHTML=Ut),pl=i(e),se=M(e,"P",{"data-svelte-h":!0}),m(se)!=="svelte-xhnj7d"&&(se.innerHTML=ut),ol=i(e),ae=M(e,"UL",{"data-svelte-h":!0}),m(ae)!=="svelte-hrjisu"&&(ae.innerHTML=ct),Ml=i(e),ie=M(e,"P",{"data-svelte-h":!0}),m(ie)!=="svelte-1l1rgk1"&&(ie.innerHTML=ft),rl=i(e),f(ne.$$.fragment,e),ml=i(e),pe=M(e,"P",{"data-svelte-h":!0}),m(pe)!=="svelte-crnf53"&&(pe.innerHTML=dt),yl=i(e),f(oe.$$.fragment,e),Ul=i(e),X=M(e,"DIV",{class:!0,"data-svelte-h":!0}),m(X)!=="svelte-j6euo"&&(X.innerHTML=Jt),ul=i(e),f(Me.$$.fragment,e),cl=i(e),re=M(e,"P",{"data-svelte-h":!0}),m(re)!=="svelte-ujyh26"&&(re.innerHTML=Tt),fl=i(e),f(k.$$.fragment,e),dl=i(e),me=M(e,"P",{"data-svelte-h":!0}),m(me)!=="svelte-1dvocne"&&(me.textContent=ht),Jl=i(e),f(ye.$$.fragment,e),Tl=i(e),Ue=M(e,"P",{"data-svelte-h":!0}),m(Ue)!=="svelte-18gn0k8"&&(Ue.innerHTML=Ct),hl=i(e),f(ue.$$.fragment,e),Cl=i(e),L=M(e,"DIV",{class:!0,"data-svelte-h":!0}),m(L)!=="svelte-848z5s"&&(L.innerHTML=bt),bl=i(e),ce=M(e,"P",{"data-svelte-h":!0}),m(ce)!=="svelte-1qz3wem"&&(ce.innerHTML=gt),gl=i(e),fe=M(e,"UL",{"data-svelte-h":!0}),m(fe)!=="svelte-dlnea7"&&(fe.innerHTML=Vt),Vl=i(e),de=M(e,"P",{"data-svelte-h":!0}),m(de)!=="svelte-1fsp6a2"&&(de.innerHTML=wt),wl=i(e),f(Je.$$.fragment,e),jl=i(e),x=M(e,"DIV",{class:!0,"data-svelte-h":!0}),m(x)!=="svelte-1etahws"&&(x.innerHTML=jt),Ql=i(e),Te=M(e,"P",{"data-svelte-h":!0}),m(Te)!=="svelte-1e49s07"&&(Te.innerHTML=Qt),$l=i(e),f(he.$$.fragment,e),Zl=i(e),f(Ce.$$.fragment,e),Il=i(e),be=M(e,"P",{"data-svelte-h":!0}),m(be)!=="svelte-6nr24m"&&(be.innerHTML=$t),Bl=i(e),ge=M(e,"P",{"data-svelte-h":!0}),m(ge)!=="svelte-ytof64"&&(ge.innerHTML=Zt),Rl=i(e),f(Ve.$$.fragment,e),Wl=i(e),we=M(e,"P",{"data-svelte-h":!0}),m(we)!=="svelte-1disx0g"&&(we.innerHTML=It),Fl=i(e),f(S.$$.fragment,e),Gl=i(e),f(je.$$.fragment,e),vl=i(e),Qe=M(e,"P",{"data-svelte-h":!0}),m(Qe)!=="svelte-l3jlx7"&&(Qe.innerHTML=Bt),_l=i(e),f(N.$$.fragment,e),Xl=i(e),f($e.$$.fragment,e),kl=i(e),Ze=M(e,"P",{"data-svelte-h":!0}),m(Ze)!=="svelte-jqqmto"&&(Ze.innerHTML=Rt),Ll=i(e),Ie=M(e,"P",{"data-svelte-h":!0}),m(Ie)!=="svelte-1aw44ax"&&(Ie.innerHTML=Wt),xl=i(e),H=M(e,"BLOCKQUOTE",{class:!0,"data-svelte-h":!0}),m(H)!=="svelte-1mdjx1c"&&(H.innerHTML=Ft),Sl=i(e),Be=M(e,"P",{"data-svelte-h":!0}),m(Be)!=="svelte-13chlgn"&&(Be.textContent=Gt),Nl=i(e),f(Re.$$.fragment,e),Hl=i(e),We=M(e,"P",{"data-svelte-h":!0}),m(We)!=="svelte-63lykc"&&(We.innerHTML=vt),Al=i(e),f(Fe.$$.fragment,e),El=i(e),Ge=M(e,"P",{"data-svelte-h":!0}),m(Ge)!=="svelte-1hljpgw"&&(Ge.textContent=_t),Yl=i(e),f(ve.$$.fragment,e),ql=i(e),A=M(e,"DIV",{class:!0,"data-svelte-h":!0}),m(A)!=="svelte-1vum0wo"&&(A.innerHTML=Xt),Dl=i(e),f(_e.$$.fragment,e),zl=i(e),Xe=M(e,"P",{"data-svelte-h":!0}),m(Xe)!=="svelte-9vlttf"&&(Xe.innerHTML=kt),Pl=i(e),ke=M(e,"P",{"data-svelte-h":!0}),m(ke)!=="svelte-gjj407"&&(ke.innerHTML=Lt),Ol=i(e),f(Le.$$.fragment,e),Kl=i(e),f(xe.$$.fragment,e),et=i(e),Se=M(e,"P",{"data-svelte-h":!0}),m(Se)!=="svelte-1dpf8s3"&&(Se.innerHTML=xt),lt=i(e),f(E.$$.fragment,e),tt=i(e),f(Ne.$$.fragment,e),st=i(e),He=M(e,"P",{"data-svelte-h":!0}),m(He)!=="svelte-1b45tv3"&&(He.innerHTML=St),at=i(e),f(Ae.$$.fragment,e),it=i(e),f(Ee.$$.fragment,e),nt=i(e),qe=M(e,"P",{}),qt(qe).forEach(t),this.h()},h(){F(n,"name","hf:doc:metadata"),F(n,"content",fs),F(v,"class","flex justify-center"),F(_,"class","flex justify-center"),F(X,"class","flex justify-center"),F(L,"class","flex justify-center"),F(x,"class","flex justify-center"),F(H,"class","tip"),F(A,"class","flex justify-center")},m(e,l){ls(document.head,n),s(e,C,l),s(e,r,l),s(e,u,l),d(U,e,l),s(e,y,l),d(g,e,l),s(e,Z,l),s(e,j,l),s(e,I,l),s(e,W,l),s(e,B,l),d(Q,e,l),s(e,$,l),d(R,e,l),s(e,p,l),s(e,b,l),s(e,Pe,l),s(e,D,l),s(e,Oe,l),d(z,e,l),s(e,Ke,l),s(e,v,l),s(e,el,l),d(P,e,l),s(e,ll,l),s(e,O,l),s(e,tl,l),s(e,K,l),s(e,sl,l),d(ee,e,l),s(e,al,l),s(e,le,l),s(e,il,l),d(te,e,l),s(e,nl,l),s(e,_,l),s(e,pl,l),s(e,se,l),s(e,ol,l),s(e,ae,l),s(e,Ml,l),s(e,ie,l),s(e,rl,l),d(ne,e,l),s(e,ml,l),s(e,pe,l),s(e,yl,l),d(oe,e,l),s(e,Ul,l),s(e,X,l),s(e,ul,l),d(Me,e,l),s(e,cl,l),s(e,re,l),s(e,fl,l),d(k,e,l),s(e,dl,l),s(e,me,l),s(e,Jl,l),d(ye,e,l),s(e,Tl,l),s(e,Ue,l),s(e,hl,l),d(ue,e,l),s(e,Cl,l),s(e,L,l),s(e,bl,l),s(e,ce,l),s(e,gl,l),s(e,fe,l),s(e,Vl,l),s(e,de,l),s(e,wl,l),d(Je,e,l),s(e,jl,l),s(e,x,l),s(e,Ql,l),s(e,Te,l),s(e,$l,l),d(he,e,l),s(e,Zl,l),d(Ce,e,l),s(e,Il,l),s(e,be,l),s(e,Bl,l),s(e,ge,l),s(e,Rl,l),d(Ve,e,l),s(e,Wl,l),s(e,we,l),s(e,Fl,l),d(S,e,l),s(e,Gl,l),d(je,e,l),s(e,vl,l),s(e,Qe,l),s(e,_l,l),d(N,e,l),s(e,Xl,l),d($e,e,l),s(e,kl,l),s(e,Ze,l),s(e,Ll,l),s(e,Ie,l),s(e,xl,l),s(e,H,l),s(e,Sl,l),s(e,Be,l),s(e,Nl,l),d(Re,e,l),s(e,Hl,l),s(e,We,l),s(e,Al,l),d(Fe,e,l),s(e,El,l),s(e,Ge,l),s(e,Yl,l),d(ve,e,l),s(e,ql,l),s(e,A,l),s(e,Dl,l),d(_e,e,l),s(e,zl,l),s(e,Xe,l),s(e,Pl,l),s(e,ke,l),s(e,Ol,l),d(Le,e,l),s(e,Kl,l),d(xe,e,l),s(e,et,l),s(e,Se,l),s(e,lt,l),d(E,e,l),s(e,tt,l),d(Ne,e,l),s(e,st,l),s(e,He,l),s(e,at,l),d(Ae,e,l),s(e,it,l),d(Ee,e,l),s(e,nt,l),s(e,qe,l),pt=!0},p(e,[l]){const Nt={};l&2&&(Nt.$$scope={dirty:l,ctx:e}),Q.$set(Nt);const Ht={};l&2&&(Ht.$$scope={dirty:l,ctx:e}),k.$set(Ht);const At={};l&2&&(At.$$scope={dirty:l,ctx:e}),S.$set(At);const Et={};l&2&&(Et.$$scope={dirty:l,ctx:e}),N.$set(Et);const Yt={};l&2&&(Yt.$$scope={dirty:l,ctx:e}),E.$set(Yt)},i(e){pt||(J(U.$$.fragment,e),J(g.$$.fragment,e),J(Q.$$.fragment,e),J(R.$$.fragment,e),J(z.$$.fragment,e),J(P.$$.fragment,e),J(ee.$$.fragment,e),J(te.$$.fragment,e),J(ne.$$.fragment,e),J(oe.$$.fragment,e),J(Me.$$.fragment,e),J(k.$$.fragment,e),J(ye.$$.fragment,e),J(ue.$$.fragment,e),J(Je.$$.fragment,e),J(he.$$.fragment,e),J(Ce.$$.fragment,e),J(Ve.$$.fragment,e),J(S.$$.fragment,e),J(je.$$.fragment,e),J(N.$$.fragment,e),J($e.$$.fragment,e),J(Re.$$.fragment,e),J(Fe.$$.fragment,e),J(ve.$$.fragment,e),J(_e.$$.fragment,e),J(Le.$$.fragment,e),J(xe.$$.fragment,e),J(E.$$.fragment,e),J(Ne.$$.fragment,e),J(Ae.$$.fragment,e),J(Ee.$$.fragment,e),pt=!0)},o(e){T(U.$$.fragment,e),T(g.$$.fragment,e),T(Q.$$.fragment,e),T(R.$$.fragment,e),T(z.$$.fragment,e),T(P.$$.fragment,e),T(ee.$$.fragment,e),T(te.$$.fragment,e),T(ne.$$.fragment,e),T(oe.$$.fragment,e),T(Me.$$.fragment,e),T(k.$$.fragment,e),T(ye.$$.fragment,e),T(ue.$$.fragment,e),T(Je.$$.fragment,e),T(he.$$.fragment,e),T(Ce.$$.fragment,e),T(Ve.$$.fragment,e),T(S.$$.fragment,e),T(je.$$.fragment,e),T(N.$$.fragment,e),T($e.$$.fragment,e),T(Re.$$.fragment,e),T(Fe.$$.fragment,e),T(ve.$$.fragment,e),T(_e.$$.fragment,e),T(Le.$$.fragment,e),T(xe.$$.fragment,e),T(E.$$.fragment,e),T(Ne.$$.fragment,e),T(Ae.$$.fragment,e),T(Ee.$$.fragment,e),pt=!1},d(e){e&&(t(C),t(r),t(u),t(y),t(Z),t(j),t(I),t(W),t(B),t($),t(p),t(b),t(Pe),t(D),t(Oe),t(Ke),t(v),t(el),t(ll),t(O),t(tl),t(K),t(sl),t(al),t(le),t(il),t(nl),t(_),t(pl),t(se),t(ol),t(ae),t(Ml),t(ie),t(rl),t(ml),t(pe),t(yl),t(Ul),t(X),t(ul),t(cl),t(re),t(fl),t(dl),t(me),t(Jl),t(Tl),t(Ue),t(hl),t(Cl),t(L),t(bl),t(ce),t(gl),t(fe),t(Vl),t(de),t(wl),t(jl),t(x),t(Ql),t(Te),t($l),t(Zl),t(Il),t(be),t(Bl),t(ge),t(Rl),t(Wl),t(we),t(Fl),t(Gl),t(vl),t(Qe),t(_l),t(Xl),t(kl),t(Ze),t(Ll),t(Ie),t(xl),t(H),t(Sl),t(Be),t(Nl),t(Hl),t(We),t(Al),t(El),t(Ge),t(Yl),t(ql),t(A),t(Dl),t(zl),t(Xe),t(Pl),t(ke),t(Ol),t(Kl),t(et),t(Se),t(lt),t(tt),t(st),t(He),t(at),t(it),t(nt),t(qe)),t(n),h(U,e),h(g,e),h(Q,e),h(R,e),h(z,e),h(P,e),h(ee,e),h(te,e),h(ne,e),h(oe,e),h(Me,e),h(k,e),h(ye,e),h(ue,e),h(Je,e),h(he,e),h(Ce,e),h(Ve,e),h(S,e),h(je,e),h(N,e),h($e,e),h(Re,e),h(Fe,e),h(ve,e),h(_e,e),h(Le,e),h(xe,e),h(E,e),h(Ne,e),h(Ae,e),h(Ee,e)}}}const fs='{"title":"어댑터 불러오기","local":"어댑터-불러오기","sections":[{"title":"DreamBooth","local":"dreambooth","sections":[],"depth":2},{"title":"Textual inversion","local":"textual-inversion","sections":[],"depth":2},{"title":"LoRA","local":"lora","sections":[{"title":"LoRA 가중치 스케일 조정하기","local":"lora-가중치-스케일-조정하기","sections":[],"depth":3},{"title":"Kohya와 TheLastBen","local":"kohya와-thelastben","sections":[],"depth":3}],"depth":2},{"title":"IP-Adapter","local":"ip-adapter","sections":[{"title":"IP-Adapter Plus","local":"ip-adapter-plus","sections":[],"depth":3},{"title":"IP-Adapter Face ID 모델","local":"ip-adapter-face-id-모델","sections":[],"depth":3}],"depth":2}],"depth":1}';function ds(w){return Pt(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class ws extends Ot{constructor(n){super(),Kt(this,n,ds,cs,zt,{})}}export{ws as component};