diffusers/pr_11739/ko/_app/immutable/nodes/20.9d181541.js

import{s as Yt,o as At,n as Dt}from"../chunks/scheduler.94020406.js";import{S as Pt,i as qt,g as a,s as i,r as M,E as Kt,h as p,f as t,c as n,j as Et,u as m,x as r,k as J,y as Ot,a as s,v as f,d as o,t as c,w as u}from"../chunks/index.a08c8d92.js";import{T as Ft}from"../chunks/Tip.3b0aeee8.js";import{C as y}from"../chunks/CodeBlock.f1fae7de.js";import{D as es}from"../chunks/DocNotebookDropdown.a1753374.js";import{H as W,E as ls}from"../chunks/getInferenceSnippets.3bf24426.js";function ts(Re){let d,h='💡 GPU에 액세스할 수 없는 경우 다음과 같은 GPU 제공업체에서 무료로 사용할 수 있습니다!. <a href="https://colab.research.google.com/" rel="nofollow">Colab</a>';return{c(){d=a("p"),d.innerHTML=h},l(b){d=p(b,"P",{"data-svelte-h":!0}),r(d)!=="svelte-vjsly4"&&(d.innerHTML=h)},m(b,w){s(b,d,w)},p:Dt,d(b){b&&t(d)}}}function ss(Re){let d,h="💡 파이프라인은 항상 <code>float16</code>에서 실행할 것을 강력히 권장하며, 지금까지 출력 품질이 저하되는 경우는 거의 없었습니다.";return{c(){d=a("p"),d.innerHTML=h},l(b){d=p(b,"P",{"data-svelte-h":!0}),r(d)!=="svelte-169sehu"&&(d.innerHTML=h)},m(b,w){s(b,d,w)},p:Dt,d(b){b&&t(d)}}}function is(Re){let d,h,b,w,C,Ne,I,Le,k,st="특정 스타일로 이미지를 생성하거나 원하는 내용을 포함하도록<code>DiffusionPipeline</code>을 설정하는 것은 까다로울 수 있습니다. 종종 만족스러운 이미지를 얻기까지 <code>DiffusionPipeline</code>을 여러 번 실행해야 하는 경우가 많습니다. 그러나 무에서 유를 창조하는 것은 특히 추론을 반복해서 실행하는 경우 계산 집약적인 프로세스입니다.",ze,B,it="그렇기 때문에 파이프라인에서 <em>계산</em>(속도) 및 <em>메모리</em>(GPU RAM) 효율성을 극대화하여 추론 주기 사이의 시간을 단축하여 더 빠르게 반복할 수 있도록 하는 것이 중요합니다.",Se,V,nt="이 튜토리얼에서는 <code>DiffusionPipeline</code>을 사용하여 더 빠르고 효과적으로 생성하는 방법을 안내합니다.",Qe,H,at='<a href="https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5" rel="nofollow"><code>stable-diffusion-v1-5/stable-diffusion-v1-5</code></a> 모델을 불러와서 시작합니다:',Ee,x,Fe,R,pt="예제 프롬프트는 “portrait of an old warrior chief” 이지만, 자유롭게 자신만의 프롬프트를 사용해도 됩니다:",De,X,Ye,N,Ae,T,Pe,L,rt="추론 속도를 높이는 가장 간단한 방법 중 하나는 Pytorch 모듈을 사용할 때와 같은 방식으로 GPU에 파이프라인을 배치하는 것입니다:",qe,z,Ke,S,Mt='동일한 이미지를 사용하고 개선할 수 있는지 확인하려면 <a href="https://pytorch.org/docs/stable/generated/torch.Generator.html" rel="nofollow"><code>Generator</code></a>를 사용하고 <a href="./using-diffusers/reusing_seeds">재현성</a>에 대한 시드를 설정하세요:',Oe,Q,el,E,mt="이제 이미지를 생성할 수 있습니다:",ll,F,tl,U,ft='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_1.png"/>',sl,D,ot="이 프로세스는 T4 GPU에서 약 30초가 소요되었습니다(할당된 GPU가 T4보다 나은 경우 더 빠를 수 있음). 기본적으로 <code>DiffusionPipeline</code>은 50개의 추론 단계에 대해 전체 <code>float32</code> 정밀도로 추론을 실행합니다. <code>float16</code>과 같은 더 낮은 정밀도로 전환하거나 추론 단계를 더 적게 실행하여 속도를 높일 수 있습니다.",il,Y,ct="<code>float16</code>으로 모델을 로드하고 이미지를 생성해 보겠습니다:",nl,A,al,Z,ut='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_2.png"/>',pl,P,dt="이번에는 이미지를 생성하는 데 약 11초밖에 걸리지 않아 이전보다 3배 가까이 빨라졌습니다!",rl,g,Ml,q,yt="또 다른 옵션은 추론 단계의 수를 줄이는 것입니다. 보다 효율적인 스케줄러를 선택하면 출력 품질 저하 없이 단계 수를 줄이는 데 도움이 될 수 있습니다. 현재 모델과 호환되는 스케줄러는 <code>compatibles</code> 메서드를 호출하여 <code>DiffusionPipeline</code>에서 찾을 수 있습니다:",ml,K,fl,O,bt="Stable Diffusion 모델은 일반적으로 약 50개의 추론 단계가 필요한 <code>PNDMScheduler</code>를 기본으로 사용하지만, <code>DPMSolverMultistepScheduler</code>와 같이 성능이 더 뛰어난 스케줄러는 약 20개 또는 25개의 추론 단계만 필요로 합니다. 새 스케줄러를 로드하려면 <code>ConfigMixin.from_config()</code> 메서드를 사용합니다:",ol,ee,cl,le,ht="<code>num_inference_steps</code>를 20으로 설정합니다:",ul,te,dl,j,Jt='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_3.png"/>',yl,se,wt="추론시간을 4초로 단축할 수 있었습니다! ⚡️",bl,ie,hl,ne,Tt="파이프라인 성능 향상의 또 다른 핵심은 메모리 사용량을 줄이는 것인데, 초당 생성되는 이미지 수를 최대화하려고 하는 경우가 많기 때문에 간접적으로 더 빠른 속도를 의미합니다. 한 번에 생성할 수 있는 이미지 수를 확인하는 가장 쉬운 방법은 <code>OutOfMemoryError</code>(OOM)이 발생할 때까지 다양한 배치 크기를 시도해 보는 것입니다.",Jl,ae,Ut="프롬프트 목록과 <code>Generators</code>에서 이미지 배치를 생성하는 함수를 만듭니다. 좋은 결과를 생성하는 경우 재사용할 수 있도록 각 <code>Generator</code>에 시드를 할당해야 합니다.",wl,pe,Tl,re,Zt="또한 각 이미지 배치를 보여주는 기능이 필요합니다:",Ul,Me,Zl,me,gt="<code>batch_size=4</code>부터 시작해 얼마나 많은 메모리를 소비했는지 확인합니다:",gl,fe,jl,oe,jt="RAM이 더 많은 GPU가 아니라면 위의 코드에서 <code>OOM</code> 오류가 반환되었을 것입니다! 대부분의 메모리는 cross-attention 레이어가 차지합니다. 이 작업을 배치로 실행하는 대신 순차적으로 실행하면 상당한 양의 메모리를 절약할 수 있습니다. 파이프라인을 구성하여 <code>enable_attention_slicing()</code> 함수를 사용하기만 하면 됩니다:",$l,ce,Gl,ue,$t="이제 <code>batch_size</code>를 8로 늘려보세요!",vl,de,_l,$,Gt='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_5.png"/>',Wl,ye,vt="이전에는 4개의 이미지를 배치로 생성할 수도 없었지만, 이제는 이미지당 약 3.5초 만에 8개의 이미지를 배치로 생성할 수 있습니다! 이는 아마도 품질 저하 없이 T4 GPU에서 가장 빠른 속도일 것입니다.",Cl,be,Il,he,_t="지난 두 섹션에서는 <code>fp16</code>을 사용하여 파이프라인의 속도를 최적화하고, 더 성능이 좋은 스케줄러를 사용하여 추론 단계의 수를 줄이고, attention slicing을 활성화하여 메모리 소비를 줄이는 방법을 배웠습니다. 이제 생성된 이미지의 품질을 개선하는 방법에 대해 집중적으로 알아보겠습니다.",kl,Je,Bl,we,Wt='가장 확실한 단계는 더 나은 체크포인트를 사용하는 것입니다. Stable Diffusion 모델은 좋은 출발점이며, 공식 출시 이후 몇 가지 개선된 버전도 출시되었습니다. 하지만 최신 버전을 사용한다고 해서 자동으로 더 나은 결과를 얻을 수 있는 것은 아닙니다. 여전히 다양한 체크포인트를 직접 실험해보고, <a href="https://minimaxir.com/2022/11/stable-diffusion-negative-prompt/" rel="nofollow">negative prompts</a> 사용 등 약간의 조사를 통해 최상의 결과를 얻어야 합니다.',Vl,Te,Ct='이 분야가 성장함에 따라 특정 스타일을 연출할 수 있도록 세밀하게 조정된 고품질 체크포인트가 점점 더 많아지고 있습니다. <a href="https://huggingface.co/models?library=diffusers&amp;sort=downloads" rel="nofollow">Hub</a>와 <a href="https://huggingface.co/spaces/huggingface-projects/diffusers-gallery" rel="nofollow">Diffusers Gallery</a>를 둘러보고 관심 있는 것을 찾아보세요!',Hl,Ue,xl,Ze,It='현재 파이프라인 구성 요소를 최신 버전으로 교체해 볼 수도 있습니다. Stability AI의 최신 <a href="https://huggingface.co/stabilityai/stable-diffusion-2-1/tree/main/vae" rel="nofollow">autodecoder</a>를 파이프라인에 로드하고 몇 가지 이미지를 생성해 보겠습니다:',Rl,ge,Xl,G,kt='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_6.png"/>',Nl,je,Ll,$e,Bt="이미지를 생성하는 데 사용하는 텍스트 프롬프트는 <em>prompt engineering</em>이라고 할 정도로 매우 중요합니다. 프롬프트 엔지니어링 시 고려해야 할 몇 가지 사항은 다음과 같습니다:",zl,Ge,Vt="<li>생성하려는 이미지 또는 유사한 이미지가 인터넷에 어떻게 저장되어 있는가?</li> <li>내가 원하는 스타일로 모델을 유도하기 위해 어떤 추가 세부 정보를 제공할 수 있는가?</li>",Sl,ve,Ht="이를 염두에 두고 색상과 더 높은 품질의 디테일을 포함하도록 프롬프트를 개선해 봅시다:",Ql,_e,El,We,xt="새로운 프롬프트로 이미지 배치를 생성합니다:",Fl,Ce,Dl,v,Rt='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_7.png"/>',Yl,Ie,Xt="꽤 인상적입니다! <code>1</code>의 시드를 가진 <code>Generator</code>에 해당하는 두 번째 이미지에 피사체의 나이에 대한 텍스트를 추가하여 조금 더 조정해 보겠습니다:",Al,ke,Pl,_,Nt='<img src="https://huggingface.co/datasets/diffusers/docs-images/resolve/main/stable_diffusion_101/sd_101_8.png"/>',ql,Be,Kl,Ve,Lt="이 튜토리얼에서는 계산 및 메모리 효율을 높이고 생성된 출력의 품질을 개선하기 위해 <code>DiffusionPipeline</code>을 최적화하는 방법을 배웠습니다. 파이프라인을 더 빠르게 만드는 데 관심이 있다면 다음 리소스를 살펴보세요:",Ol,He,zt='<li><a href="./optimization/torch2.0">PyTorch 2.0</a> 및 <a href="https://pytorch.org/docs/stable/generated/torch.compile.html" rel="nofollow"><code>torch.compile</code></a>이 어떻게 추론 속도를 5~300% 향상시킬 수 있는지 알아보세요. A100 GPU에서는 추론 속도가 최대 50%까지 빨라질 수 있습니다!</li> <li>PyTorch 2를 사용할 수 없는 경우, <a href="./optimization/xformers">xFormers</a>를 설치하는 것이 좋습니다. 메모리 효율적인 어텐션 메커니즘은 PyTorch 1.13.1과 함께 사용하면 속도가 빨라지고 메모리 소비가 줄어듭니다.</li> <li>모델 오프로딩과 같은 다른 최적화 기법은 <a href="./optimization/fp16">이 가이드</a>에서 다루고 있습니다.</li>',et,xe,lt,Xe,tt;return C=new W({props:{title:"효과적이고 효율적인 Diffusion",local:"효과적이고-효율적인-diffusion",headingTag:"h1"}}),I=new es({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/stable_diffusion.ipynb"},{label:"PyTorch",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/ko/pytorch/stable_diffusion.ipynb"},{label:"TensorFlow",value:"https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers_doc/ko/tensorflow/stable_diffusion.ipynb"},{label:"Mixed",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/ko/stable_diffusion.ipynb"},{label:"PyTorch",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/ko/pytorch/stable_diffusion.ipynb"},{label:"TensorFlow",value:"https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/main/diffusers_doc/ko/tensorflow/stable_diffusion.ipynb"}]}}),x=new y({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMERpZmZ1c2lvblBpcGVsaW5lJTBBJTBBbW9kZWxfaWQlMjAlM0QlMjAlMjJzdGFibGUtZGlmZnVzaW9uLXYxLTUlMkZzdGFibGUtZGlmZnVzaW9uLXYxLTUlMjIlMEFwaXBlbGluZSUyMCUzRCUyMERpZmZ1c2lvblBpcGVsaW5lLmZyb21fcHJldHJhaW5lZChtb2RlbF9pZCk=",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> DiffusionPipeline

model_id = <span class="hljs-string">&quot;stable-diffusion-v1-5/stable-diffusion-v1-5&quot;</span>
pipeline = DiffusionPipeline.from_pretrained(model_id)`,wrap:!1}}),X=new y({props:{code:"cHJvbXB0JTIwJTNEJTIwJTIycG9ydHJhaXQlMjBwaG90byUyMG9mJTIwYSUyMG9sZCUyMHdhcnJpb3IlMjBjaGllZiUyMg==",highlighted:'prompt = <span class="hljs-string">&quot;portrait photo of a old warrior chief&quot;</span>',wrap:!1}}),N=new W({props:{title:"속도",local:"속도",headingTag:"h2"}}),T=new Ft({props:{$$slots:{default:[ts]},$$scope:{ctx:Re}}}),z=new y({props:{code:"cGlwZWxpbmUlMjAlM0QlMjBwaXBlbGluZS50byglMjJjdWRhJTIyKQ==",highlighted:'pipeline = pipeline.to(<span class="hljs-string">&quot;cuda&quot;</span>)',wrap:!1}}),Q=new y({props:{code:"aW1wb3J0JTIwdG9yY2glMEElMEFnZW5lcmF0b3IlMjAlM0QlMjB0b3JjaC5HZW5lcmF0b3IoJTIyY3VkYSUyMikubWFudWFsX3NlZWQoMCk=",highlighted:`<span class="hljs-keyword">import</span> torch

generator = torch.Generator(<span class="hljs-string">&quot;cuda&quot;</span>).manual_seed(<span class="hljs-number">0</span>)`,wrap:!1}}),F=new y({props:{code:"aW1hZ2UlMjAlM0QlMjBwaXBlbGluZShwcm9tcHQlMkMlMjBnZW5lcmF0b3IlM0RnZW5lcmF0b3IpLmltYWdlcyU1QjAlNUQlMEFpbWFnZQ==",highlighted:`image = pipeline(prompt, generator=generator).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),A=new y({props:{code:"aW1wb3J0JTIwdG9yY2glMEElMEFwaXBlbGluZSUyMCUzRCUyMERpZmZ1c2lvblBpcGVsaW5lLmZyb21fcHJldHJhaW5lZChtb2RlbF9pZCUyQyUyMHRvcmNoX2R0eXBlJTNEdG9yY2guZmxvYXQxNiklMEFwaXBlbGluZSUyMCUzRCUyMHBpcGVsaW5lLnRvKCUyMmN1ZGElMjIpJTBBZ2VuZXJhdG9yJTIwJTNEJTIwdG9yY2guR2VuZXJhdG9yKCUyMmN1ZGElMjIpLm1hbnVhbF9zZWVkKDApJTBBaW1hZ2UlMjAlM0QlMjBwaXBlbGluZShwcm9tcHQlMkMlMjBnZW5lcmF0b3IlM0RnZW5lcmF0b3IpLmltYWdlcyU1QjAlNUQlMEFpbWFnZQ==",highlighted:`<span class="hljs-keyword">import</span> torch

pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipeline = pipeline.to(<span class="hljs-string">&quot;cuda&quot;</span>)
generator = torch.Generator(<span class="hljs-string">&quot;cuda&quot;</span>).manual_seed(<span class="hljs-number">0</span>)
image = pipeline(prompt, generator=generator).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),g=new Ft({props:{$$slots:{default:[ss]},$$scope:{ctx:Re}}}),K=new y({props:{code:"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",highlighted:`pipeline.scheduler.compatibles
[
    diffusers.schedulers.scheduling_lms_discrete.LMSDiscreteScheduler,
    diffusers.schedulers.scheduling_unipc_multistep.UniPCMultistepScheduler,
    diffusers.schedulers.scheduling_k_dpm_2_discrete.KDPM2DiscreteScheduler,
    diffusers.schedulers.scheduling_deis_multistep.DEISMultistepScheduler,
    diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler,
    diffusers.schedulers.scheduling_dpmsolver_multistep.DPMSolverMultistepScheduler,
    diffusers.schedulers.scheduling_ddpm.DDPMScheduler,
    diffusers.schedulers.scheduling_dpmsolver_singlestep.DPMSolverSinglestepScheduler,
    diffusers.schedulers.scheduling_k_dpm_2_ancestral_discrete.KDPM2AncestralDiscreteScheduler,
    diffusers.schedulers.scheduling_heun_discrete.HeunDiscreteScheduler,
    diffusers.schedulers.scheduling_pndm.PNDMScheduler,
    diffusers.schedulers.scheduling_euler_ancestral_discrete.EulerAncestralDiscreteScheduler,
    diffusers.schedulers.scheduling_ddim.DDIMScheduler,
]`,wrap:!1}}),ee=new y({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMERQTVNvbHZlck11bHRpc3RlcFNjaGVkdWxlciUwQSUwQXBpcGVsaW5lLnNjaGVkdWxlciUyMCUzRCUyMERQTVNvbHZlck11bHRpc3RlcFNjaGVkdWxlci5mcm9tX2NvbmZpZyhwaXBlbGluZS5zY2hlZHVsZXIuY29uZmlnKQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> DPMSolverMultistepScheduler

pipeline.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config)`,wrap:!1}}),te=new y({props:{code:"Z2VuZXJhdG9yJTIwJTNEJTIwdG9yY2guR2VuZXJhdG9yKCUyMmN1ZGElMjIpLm1hbnVhbF9zZWVkKDApJTBBaW1hZ2UlMjAlM0QlMjBwaXBlbGluZShwcm9tcHQlMkMlMjBnZW5lcmF0b3IlM0RnZW5lcmF0b3IlMkMlMjBudW1faW5mZXJlbmNlX3N0ZXBzJTNEMjApLmltYWdlcyU1QjAlNUQlMEFpbWFnZQ==",highlighted:`generator = torch.Generator(<span class="hljs-string">&quot;cuda&quot;</span>).manual_seed(<span class="hljs-number">0</span>)
image = pipeline(prompt, generator=generator, num_inference_steps=<span class="hljs-number">20</span>).images[<span class="hljs-number">0</span>]
image`,wrap:!1}}),ie=new W({props:{title:"메모리",local:"메모리",headingTag:"h2"}}),pe=new y({props:{code:"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",highlighted:`<span class="hljs-keyword">def</span> <span class="hljs-title function_">get_inputs</span>(<span class="hljs-params">batch_size=<span class="hljs-number">1</span></span>):
    generator = [torch.Generator(<span class="hljs-string">&quot;cuda&quot;</span>).manual_seed(i) <span class="hljs-keyword">for</span> i <span class="hljs-keyword">in</span> <span class="hljs-built_in">range</span>(batch_size)]
    prompts = batch_size * [prompt]
    num_inference_steps = <span class="hljs-number">20</span>

    <span class="hljs-keyword">return</span> {<span class="hljs-string">&quot;prompt&quot;</span>: prompts, <span class="hljs-string">&quot;generator&quot;</span>: generator, <span class="hljs-string">&quot;num_inference_steps&quot;</span>: num_inference_steps}`,wrap:!1}}),Me=new y({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> PIL <span class="hljs-keyword">import</span> Image


<span class="hljs-keyword">def</span> <span class="hljs-title function_">image_grid</span>(<span class="hljs-params">imgs, rows=<span class="hljs-number">2</span>, cols=<span class="hljs-number">2</span></span>):
    w, h = imgs[<span class="hljs-number">0</span>].size
    grid = Image.new(<span class="hljs-string">&quot;RGB&quot;</span>, size=(cols * w, rows * h))

    <span class="hljs-keyword">for</span> i, img <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(imgs):
        grid.paste(img, box=(i % cols * w, i // cols * h))
    <span class="hljs-keyword">return</span> grid`,wrap:!1}}),fe=new y({props:{code:"aW1hZ2VzJTIwJTNEJTIwcGlwZWxpbmUoKipnZXRfaW5wdXRzKGJhdGNoX3NpemUlM0Q0KSkuaW1hZ2VzJTBBaW1hZ2VfZ3JpZChpbWFnZXMp",highlighted:`images = pipeline(**get_inputs(batch_size=<span class="hljs-number">4</span>)).images
image_grid(images)`,wrap:!1}}),ce=new y({props:{code:"cGlwZWxpbmUuZW5hYmxlX2F0dGVudGlvbl9zbGljaW5nKCk=",highlighted:"pipeline.enable_attention_slicing()",wrap:!1}}),de=new y({props:{code:"aW1hZ2VzJTIwJTNEJTIwcGlwZWxpbmUoKipnZXRfaW5wdXRzKGJhdGNoX3NpemUlM0Q4KSkuaW1hZ2VzJTBBaW1hZ2VfZ3JpZChpbWFnZXMlMkMlMjByb3dzJTNEMiUyQyUyMGNvbHMlM0Q0KQ==",highlighted:`images = pipeline(**get_inputs(batch_size=<span class="hljs-number">8</span>)).images
image_grid(images, rows=<span class="hljs-number">2</span>, cols=<span class="hljs-number">4</span>)`,wrap:!1}}),be=new W({props:{title:"품질",local:"품질",headingTag:"h2"}}),Je=new W({props:{title:"더 나은 체크포인트",local:"더-나은-체크포인트",headingTag:"h3"}}),Ue=new W({props:{title:"더 나은 파이프라인 구성 요소",local:"더-나은-파이프라인-구성-요소",headingTag:"h3"}}),ge=new y({props:{code:"ZnJvbSUyMGRpZmZ1c2VycyUyMGltcG9ydCUyMEF1dG9lbmNvZGVyS0wlMEElMEF2YWUlMjAlM0QlMjBBdXRvZW5jb2RlcktMLmZyb21fcHJldHJhaW5lZCglMjJzdGFiaWxpdHlhaSUyRnNkLXZhZS1mdC1tc2UlMjIlMkMlMjB0b3JjaF9kdHlwZSUzRHRvcmNoLmZsb2F0MTYpLnRvKCUyMmN1ZGElMjIpJTBBcGlwZWxpbmUudmFlJTIwJTNEJTIwdmFlJTBBaW1hZ2VzJTIwJTNEJTIwcGlwZWxpbmUoKipnZXRfaW5wdXRzKGJhdGNoX3NpemUlM0Q4KSkuaW1hZ2VzJTBBaW1hZ2VfZ3JpZChpbWFnZXMlMkMlMjByb3dzJTNEMiUyQyUyMGNvbHMlM0Q0KQ==",highlighted:`<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoencoderKL

vae = AutoencoderKL.from_pretrained(<span class="hljs-string">&quot;stabilityai/sd-vae-ft-mse&quot;</span>, torch_dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)
pipeline.vae = vae
images = pipeline(**get_inputs(batch_size=<span class="hljs-number">8</span>)).images
image_grid(images, rows=<span class="hljs-number">2</span>, cols=<span class="hljs-number">4</span>)`,wrap:!1}}),je=new W({props:{title:"더 나은 프롬프트 엔지니어링",local:"더-나은-프롬프트-엔지니어링",headingTag:"h3"}}),_e=new y({props:{code:"cHJvbXB0JTIwJTJCJTNEJTIwJTIyJTJDJTIwdHJpYmFsJTIwcGFudGhlciUyMG1ha2UlMjB1cCUyQyUyMGJsdWUlMjBvbiUyMHJlZCUyQyUyMHNpZGUlMjBwcm9maWxlJTJDJTIwbG9va2luZyUyMGF3YXklMkMlMjBzZXJpb3VzJTIwZXllcyUyMiUwQXByb21wdCUyMCUyQiUzRCUyMCUyMiUyMDUwbW0lMjBwb3J0cmFpdCUyMHBob3RvZ3JhcGh5JTJDJTIwaGFyZCUyMHJpbSUyMGxpZ2h0aW5nJTIwcGhvdG9ncmFwaHktLWJldGElMjAtLWFyJTIwMiUzQTMlMjAlMjAtLWJldGElMjAtLXVwYmV0YSUyMg==",highlighted:`prompt += <span class="hljs-string">&quot;, tribal panther make up, blue on red, side profile, looking away, serious eyes&quot;</span>
prompt += <span class="hljs-string">&quot; 50mm portrait photography, hard rim lighting photography--beta --ar 2:3  --beta --upbeta&quot;</span>`,wrap:!1}}),Ce=new y({props:{code:"aW1hZ2VzJTIwJTNEJTIwcGlwZWxpbmUoKipnZXRfaW5wdXRzKGJhdGNoX3NpemUlM0Q4KSkuaW1hZ2VzJTBBaW1hZ2VfZ3JpZChpbWFnZXMlMkMlMjByb3dzJTNEMiUyQyUyMGNvbHMlM0Q0KQ==",highlighted:`images = pipeline(**get_inputs(batch_size=<span class="hljs-number">8</span>)).images
image_grid(images, rows=<span class="hljs-number">2</span>, cols=<span class="hljs-number">4</span>)`,wrap:!1}}),ke=new y({props:{code:"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",highlighted:`prompts = [
    <span class="hljs-string">&quot;portrait photo of the oldest warrior chief, tribal panther make up, blue on red, side profile, looking away, serious eyes 50mm portrait photography, hard rim lighting photography--beta --ar 2:3  --beta --upbeta&quot;</span>,
    <span class="hljs-string">&quot;portrait photo of a old warrior chief, tribal panther make up, blue on red, side profile, looking away, serious eyes 50mm portrait photography, hard rim lighting photography--beta --ar 2:3  --beta --upbeta&quot;</span>,
    <span class="hljs-string">&quot;portrait photo of a warrior chief, tribal panther make up, blue on red, side profile, looking away, serious eyes 50mm portrait photography, hard rim lighting photography--beta --ar 2:3  --beta --upbeta&quot;</span>,
    <span class="hljs-string">&quot;portrait photo of a young warrior chief, tribal panther make up, blue on red, side profile, looking away, serious eyes 50mm portrait photography, hard rim lighting photography--beta --ar 2:3  --beta --upbeta&quot;</span>,
]

generator = [torch.Generator(<span class="hljs-string">&quot;cuda&quot;</span>).manual_seed(<span class="hljs-number">1</span>) <span class="hljs-keyword">for</span> _ <span class="hljs-keyword">in</span> <span class="hljs-built_in">range</span>(<span class="hljs-built_in">len</span>(prompts))]
images = pipeline(prompt=prompts, generator=generator, num_inference_steps=<span class="hljs-number">25</span>).images
image_grid(images)`,wrap:!1}}),Be=new W({props:{title:"다음 단계",local:"다음-단계",headingTag:"h2"}}),xe=new ls({props:{source:"https://github.com/huggingface/diffusers/blob/main/docs/source/ko/stable_diffusion.md"}}),{c(){d=a("meta"),h=i(),b=a("p"),w=i(),M(C.$$.fragment),Ne=i(),M(I.$$.fragment),Le=i(),k=a("p"),k.innerHTML=st,ze=i(),B=a("p"),B.innerHTML=it,Se=i(),V=a("p"),V.innerHTML=nt,Qe=i(),H=a("p"),H.innerHTML=at,Ee=i(),M(x.$$.fragment),Fe=i(),R=a("p"),R.textContent=pt,De=i(),M(X.$$.fragment),Ye=i(),M(N.$$.fragment),Ae=i(),M(T.$$.fragment),Pe=i(),L=a("p"),L.textContent=rt,qe=i(),M(z.$$.fragment),Ke=i(),S=a("p"),S.innerHTML=Mt,Oe=i(),M(Q.$$.fragment),el=i(),E=a("p"),E.textContent=mt,ll=i(),M(F.$$.fragment),tl=i(),U=a("div"),U.innerHTML=ft,sl=i(),D=a("p"),D.innerHTML=ot,il=i(),Y=a("p"),Y.innerHTML=ct,nl=i(),M(A.$$.fragment),al=i(),Z=a("div"),Z.innerHTML=ut,pl=i(),P=a("p"),P.textContent=dt,rl=i(),M(g.$$.fragment),Ml=i(),q=a("p"),q.innerHTML=yt,ml=i(),M(K.$$.fragment),fl=i(),O=a("p"),O.innerHTML=bt,ol=i(),M(ee.$$.fragment),cl=i(),le=a("p"),le.innerHTML=ht,ul=i(),M(te.$$.fragment),dl=i(),j=a("div"),j.innerHTML=Jt,yl=i(),se=a("p"),se.textContent=wt,bl=i(),M(ie.$$.fragment),hl=i(),ne=a("p"),ne.innerHTML=Tt,Jl=i(),ae=a("p"),ae.innerHTML=Ut,wl=i(),M(pe.$$.fragment),Tl=i(),re=a("p"),re.textContent=Zt,Ul=i(),M(Me.$$.fragment),Zl=i(),me=a("p"),me.innerHTML=gt,gl=i(),M(fe.$$.fragment),jl=i(),oe=a("p"),oe.innerHTML=jt,$l=i(),M(ce.$$.fragment),Gl=i(),ue=a("p"),ue.innerHTML=$t,vl=i(),M(de.$$.fragment),_l=i(),$=a("div"),$.innerHTML=Gt,Wl=i(),ye=a("p"),ye.textContent=vt,Cl=i(),M(be.$$.fragment),Il=i(),he=a("p"),he.innerHTML=_t,kl=i(),M(Je.$$.fragment),Bl=i(),we=a("p"),we.innerHTML=Wt,Vl=i(),Te=a("p"),Te.innerHTML=Ct,Hl=i(),M(Ue.$$.fragment),xl=i(),Ze=a("p"),Ze.innerHTML=It,Rl=i(),M(ge.$$.fragment),Xl=i(),G=a("div"),G.innerHTML=kt,Nl=i(),M(je.$$.fragment),Ll=i(),$e=a("p"),$e.innerHTML=Bt,zl=i(),Ge=a("ul"),Ge.innerHTML=Vt,Sl=i(),ve=a("p"),ve.textContent=Ht,Ql=i(),M(_e.$$.fragment),El=i(),We=a("p"),We.textContent=xt,Fl=i(),M(Ce.$$.fragment),Dl=i(),v=a("div"),v.innerHTML=Rt,Yl=i(),Ie=a("p"),Ie.innerHTML=Xt,Al=i(),M(ke.$$.fragment),Pl=i(),_=a("div"),_.innerHTML=Nt,ql=i(),M(Be.$$.fragment),Kl=i(),Ve=a("p"),Ve.innerHTML=Lt,Ol=i(),He=a("ul"),He.innerHTML=zt,et=i(),M(xe.$$.fragment),lt=i(),Xe=a("p"),this.h()},l(e){const l=Kt("svelte-u9bgzb",document.head);d=p(l,"META",{name:!0,content:!0}),l.forEach(t),h=n(e),b=p(e,"P",{}),Et(b).forEach(t),w=n(e),m(C.$$.fragment,e),Ne=n(e),m(I.$$.fragment,e),Le=n(e),k=p(e,"P",{"data-svelte-h":!0}),r(k)!=="svelte-18ayv9y"&&(k.innerHTML=st),ze=n(e),B=p(e,"P",{"data-svelte-h":!0}),r(B)!=="svelte-oyyvon"&&(B.innerHTML=it),Se=n(e),V=p(e,"P",{"data-svelte-h":!0}),r(V)!=="svelte-1enrctx"&&(V.innerHTML=nt),Qe=n(e),H=p(e,"P",{"data-svelte-h":!0}),r(H)!=="svelte-1oj4hh8"&&(H.innerHTML=at),Ee=n(e),m(x.$$.fragment,e),Fe=n(e),R=p(e,"P",{"data-svelte-h":!0}),r(R)!=="svelte-ssflff"&&(R.textContent=pt),De=n(e),m(X.$$.fragment,e),Ye=n(e),m(N.$$.fragment,e),Ae=n(e),m(T.$$.fragment,e),Pe=n(e),L=p(e,"P",{"data-svelte-h":!0}),r(L)!=="svelte-1rfznlg"&&(L.textContent=rt),qe=n(e),m(z.$$.fragment,e),Ke=n(e),S=p(e,"P",{"data-svelte-h":!0}),r(S)!=="svelte-qirl6n"&&(S.innerHTML=Mt),Oe=n(e),m(Q.$$.fragment,e),el=n(e),E=p(e,"P",{"data-svelte-h":!0}),r(E)!=="svelte-8nqpod"&&(E.textContent=mt),ll=n(e),m(F.$$.fragment,e),tl=n(e),U=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r(U)!=="svelte-1of5nwm"&&(U.innerHTML=ft),sl=n(e),D=p(e,"P",{"data-svelte-h":!0}),r(D)!=="svelte-dt1sqn"&&(D.innerHTML=ot),il=n(e),Y=p(e,"P",{"data-svelte-h":!0}),r(Y)!=="svelte-1jr1trd"&&(Y.innerHTML=ct),nl=n(e),m(A.$$.fragment,e),al=n(e),Z=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r(Z)!=="svelte-1why3l7"&&(Z.innerHTML=ut),pl=n(e),P=p(e,"P",{"data-svelte-h":!0}),r(P)!=="svelte-1gv5jsc"&&(P.textContent=dt),rl=n(e),m(g.$$.fragment,e),Ml=n(e),q=p(e,"P",{"data-svelte-h":!0}),r(q)!=="svelte-a5gi7u"&&(q.innerHTML=yt),ml=n(e),m(K.$$.fragment,e),fl=n(e),O=p(e,"P",{"data-svelte-h":!0}),r(O)!=="svelte-gjxj1i"&&(O.innerHTML=bt),ol=n(e),m(ee.$$.fragment,e),cl=n(e),le=p(e,"P",{"data-svelte-h":!0}),r(le)!=="svelte-f76kxz"&&(le.innerHTML=ht),ul=n(e),m(te.$$.fragment,e),dl=n(e),j=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r(j)!=="svelte-19w49w4"&&(j.innerHTML=Jt),yl=n(e),se=p(e,"P",{"data-svelte-h":!0}),r(se)!=="svelte-gcc2pb"&&(se.textContent=wt),bl=n(e),m(ie.$$.fragment,e),hl=n(e),ne=p(e,"P",{"data-svelte-h":!0}),r(ne)!=="svelte-nqgs99"&&(ne.innerHTML=Tt),Jl=n(e),ae=p(e,"P",{"data-svelte-h":!0}),r(ae)!=="svelte-j2noup"&&(ae.innerHTML=Ut),wl=n(e),m(pe.$$.fragment,e),Tl=n(e),re=p(e,"P",{"data-svelte-h":!0}),r(re)!=="svelte-wyajyh"&&(re.textContent=Zt),Ul=n(e),m(Me.$$.fragment,e),Zl=n(e),me=p(e,"P",{"data-svelte-h":!0}),r(me)!=="svelte-ubrl9r"&&(me.innerHTML=gt),gl=n(e),m(fe.$$.fragment,e),jl=n(e),oe=p(e,"P",{"data-svelte-h":!0}),r(oe)!=="svelte-i1t4xo"&&(oe.innerHTML=jt),$l=n(e),m(ce.$$.fragment,e),Gl=n(e),ue=p(e,"P",{"data-svelte-h":!0}),r(ue)!=="svelte-u177c0"&&(ue.innerHTML=$t),vl=n(e),m(de.$$.fragment,e),_l=n(e),$=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r($)!=="svelte-vxa9bu"&&($.innerHTML=Gt),Wl=n(e),ye=p(e,"P",{"data-svelte-h":!0}),r(ye)!=="svelte-1nhne3o"&&(ye.textContent=vt),Cl=n(e),m(be.$$.fragment,e),Il=n(e),he=p(e,"P",{"data-svelte-h":!0}),r(he)!=="svelte-yzzx1p"&&(he.innerHTML=_t),kl=n(e),m(Je.$$.fragment,e),Bl=n(e),we=p(e,"P",{"data-svelte-h":!0}),r(we)!=="svelte-97fik6"&&(we.innerHTML=Wt),Vl=n(e),Te=p(e,"P",{"data-svelte-h":!0}),r(Te)!=="svelte-1pk1tvb"&&(Te.innerHTML=Ct),Hl=n(e),m(Ue.$$.fragment,e),xl=n(e),Ze=p(e,"P",{"data-svelte-h":!0}),r(Ze)!=="svelte-3m63vd"&&(Ze.innerHTML=It),Rl=n(e),m(ge.$$.fragment,e),Xl=n(e),G=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r(G)!=="svelte-610rhb"&&(G.innerHTML=kt),Nl=n(e),m(je.$$.fragment,e),Ll=n(e),$e=p(e,"P",{"data-svelte-h":!0}),r($e)!=="svelte-15owmr0"&&($e.innerHTML=Bt),zl=n(e),Ge=p(e,"UL",{"data-svelte-h":!0}),r(Ge)!=="svelte-1y16yq3"&&(Ge.innerHTML=Vt),Sl=n(e),ve=p(e,"P",{"data-svelte-h":!0}),r(ve)!=="svelte-17wtiaq"&&(ve.textContent=Ht),Ql=n(e),m(_e.$$.fragment,e),El=n(e),We=p(e,"P",{"data-svelte-h":!0}),r(We)!=="svelte-8cwtf"&&(We.textContent=xt),Fl=n(e),m(Ce.$$.fragment,e),Dl=n(e),v=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r(v)!=="svelte-n1o5lk"&&(v.innerHTML=Rt),Yl=n(e),Ie=p(e,"P",{"data-svelte-h":!0}),r(Ie)!=="svelte-1a7wb1e"&&(Ie.innerHTML=Xt),Al=n(e),m(ke.$$.fragment,e),Pl=n(e),_=p(e,"DIV",{class:!0,"data-svelte-h":!0}),r(_)!=="svelte-1lkw2bx"&&(_.innerHTML=Nt),ql=n(e),m(Be.$$.fragment,e),Kl=n(e),Ve=p(e,"P",{"data-svelte-h":!0}),r(Ve)!=="svelte-10t9jtg"&&(Ve.innerHTML=Lt),Ol=n(e),He=p(e,"UL",{"data-svelte-h":!0}),r(He)!=="svelte-1xpkcqm"&&(He.innerHTML=zt),et=n(e),m(xe.$$.fragment,e),lt=n(e),Xe=p(e,"P",{}),Et(Xe).forEach(t),this.h()},h(){J(d,"name","hf:doc:metadata"),J(d,"content",ns),J(U,"class","flex justify-center"),J(Z,"class","flex justify-center"),J(j,"class","flex justify-center"),J($,"class","flex justify-center"),J(G,"class","flex justify-center"),J(v,"class","flex justify-center"),J(_,"class","flex justify-center")},m(e,l){Ot(document.head,d),s(e,h,l),s(e,b,l),s(e,w,l),f(C,e,l),s(e,Ne,l),f(I,e,l),s(e,Le,l),s(e,k,l),s(e,ze,l),s(e,B,l),s(e,Se,l),s(e,V,l),s(e,Qe,l),s(e,H,l),s(e,Ee,l),f(x,e,l),s(e,Fe,l),s(e,R,l),s(e,De,l),f(X,e,l),s(e,Ye,l),f(N,e,l),s(e,Ae,l),f(T,e,l),s(e,Pe,l),s(e,L,l),s(e,qe,l),f(z,e,l),s(e,Ke,l),s(e,S,l),s(e,Oe,l),f(Q,e,l),s(e,el,l),s(e,E,l),s(e,ll,l),f(F,e,l),s(e,tl,l),s(e,U,l),s(e,sl,l),s(e,D,l),s(e,il,l),s(e,Y,l),s(e,nl,l),f(A,e,l),s(e,al,l),s(e,Z,l),s(e,pl,l),s(e,P,l),s(e,rl,l),f(g,e,l),s(e,Ml,l),s(e,q,l),s(e,ml,l),f(K,e,l),s(e,fl,l),s(e,O,l),s(e,ol,l),f(ee,e,l),s(e,cl,l),s(e,le,l),s(e,ul,l),f(te,e,l),s(e,dl,l),s(e,j,l),s(e,yl,l),s(e,se,l),s(e,bl,l),f(ie,e,l),s(e,hl,l),s(e,ne,l),s(e,Jl,l),s(e,ae,l),s(e,wl,l),f(pe,e,l),s(e,Tl,l),s(e,re,l),s(e,Ul,l),f(Me,e,l),s(e,Zl,l),s(e,me,l),s(e,gl,l),f(fe,e,l),s(e,jl,l),s(e,oe,l),s(e,$l,l),f(ce,e,l),s(e,Gl,l),s(e,ue,l),s(e,vl,l),f(de,e,l),s(e,_l,l),s(e,$,l),s(e,Wl,l),s(e,ye,l),s(e,Cl,l),f(be,e,l),s(e,Il,l),s(e,he,l),s(e,kl,l),f(Je,e,l),s(e,Bl,l),s(e,we,l),s(e,Vl,l),s(e,Te,l),s(e,Hl,l),f(Ue,e,l),s(e,xl,l),s(e,Ze,l),s(e,Rl,l),f(ge,e,l),s(e,Xl,l),s(e,G,l),s(e,Nl,l),f(je,e,l),s(e,Ll,l),s(e,$e,l),s(e,zl,l),s(e,Ge,l),s(e,Sl,l),s(e,ve,l),s(e,Ql,l),f(_e,e,l),s(e,El,l),s(e,We,l),s(e,Fl,l),f(Ce,e,l),s(e,Dl,l),s(e,v,l),s(e,Yl,l),s(e,Ie,l),s(e,Al,l),f(ke,e,l),s(e,Pl,l),s(e,_,l),s(e,ql,l),f(Be,e,l),s(e,Kl,l),s(e,Ve,l),s(e,Ol,l),s(e,He,l),s(e,et,l),f(xe,e,l),s(e,lt,l),s(e,Xe,l),tt=!0},p(e,[l]){const St={};l&2&&(St.$$scope={dirty:l,ctx:e}),T.$set(St);const Qt={};l&2&&(Qt.$$scope={dirty:l,ctx:e}),g.$set(Qt)},i(e){tt||(o(C.$$.fragment,e),o(I.$$.fragment,e),o(x.$$.fragment,e),o(X.$$.fragment,e),o(N.$$.fragment,e),o(T.$$.fragment,e),o(z.$$.fragment,e),o(Q.$$.fragment,e),o(F.$$.fragment,e),o(A.$$.fragment,e),o(g.$$.fragment,e),o(K.$$.fragment,e),o(ee.$$.fragment,e),o(te.$$.fragment,e),o(ie.$$.fragment,e),o(pe.$$.fragment,e),o(Me.$$.fragment,e),o(fe.$$.fragment,e),o(ce.$$.fragment,e),o(de.$$.fragment,e),o(be.$$.fragment,e),o(Je.$$.fragment,e),o(Ue.$$.fragment,e),o(ge.$$.fragment,e),o(je.$$.fragment,e),o(_e.$$.fragment,e),o(Ce.$$.fragment,e),o(ke.$$.fragment,e),o(Be.$$.fragment,e),o(xe.$$.fragment,e),tt=!0)},o(e){c(C.$$.fragment,e),c(I.$$.fragment,e),c(x.$$.fragment,e),c(X.$$.fragment,e),c(N.$$.fragment,e),c(T.$$.fragment,e),c(z.$$.fragment,e),c(Q.$$.fragment,e),c(F.$$.fragment,e),c(A.$$.fragment,e),c(g.$$.fragment,e),c(K.$$.fragment,e),c(ee.$$.fragment,e),c(te.$$.fragment,e),c(ie.$$.fragment,e),c(pe.$$.fragment,e),c(Me.$$.fragment,e),c(fe.$$.fragment,e),c(ce.$$.fragment,e),c(de.$$.fragment,e),c(be.$$.fragment,e),c(Je.$$.fragment,e),c(Ue.$$.fragment,e),c(ge.$$.fragment,e),c(je.$$.fragment,e),c(_e.$$.fragment,e),c(Ce.$$.fragment,e),c(ke.$$.fragment,e),c(Be.$$.fragment,e),c(xe.$$.fragment,e),tt=!1},d(e){e&&(t(h),t(b),t(w),t(Ne),t(Le),t(k),t(ze),t(B),t(Se),t(V),t(Qe),t(H),t(Ee),t(Fe),t(R),t(De),t(Ye),t(Ae),t(Pe),t(L),t(qe),t(Ke),t(S),t(Oe),t(el),t(E),t(ll),t(tl),t(U),t(sl),t(D),t(il),t(Y),t(nl),t(al),t(Z),t(pl),t(P),t(rl),t(Ml),t(q),t(ml),t(fl),t(O),t(ol),t(cl),t(le),t(ul),t(dl),t(j),t(yl),t(se),t(bl),t(hl),t(ne),t(Jl),t(ae),t(wl),t(Tl),t(re),t(Ul),t(Zl),t(me),t(gl),t(jl),t(oe),t($l),t(Gl),t(ue),t(vl),t(_l),t($),t(Wl),t(ye),t(Cl),t(Il),t(he),t(kl),t(Bl),t(we),t(Vl),t(Te),t(Hl),t(xl),t(Ze),t(Rl),t(Xl),t(G),t(Nl),t(Ll),t($e),t(zl),t(Ge),t(Sl),t(ve),t(Ql),t(El),t(We),t(Fl),t(Dl),t(v),t(Yl),t(Ie),t(Al),t(Pl),t(_),t(ql),t(Kl),t(Ve),t(Ol),t(He),t(et),t(lt),t(Xe)),t(d),u(C,e),u(I,e),u(x,e),u(X,e),u(N,e),u(T,e),u(z,e),u(Q,e),u(F,e),u(A,e),u(g,e),u(K,e),u(ee,e),u(te,e),u(ie,e),u(pe,e),u(Me,e),u(fe,e),u(ce,e),u(de,e),u(be,e),u(Je,e),u(Ue,e),u(ge,e),u(je,e),u(_e,e),u(Ce,e),u(ke,e),u(Be,e),u(xe,e)}}}const ns='{"title":"효과적이고 효율적인 Diffusion","local":"효과적이고-효율적인-diffusion","sections":[{"title":"속도","local":"속도","sections":[],"depth":2},{"title":"메모리","local":"메모리","sections":[],"depth":2},{"title":"품질","local":"품질","sections":[{"title":"더 나은 체크포인트","local":"더-나은-체크포인트","sections":[],"depth":3},{"title":"더 나은 파이프라인 구성 요소","local":"더-나은-파이프라인-구성-요소","sections":[],"depth":3},{"title":"더 나은 프롬프트 엔지니어링","local":"더-나은-프롬프트-엔지니어링","sections":[],"depth":3}],"depth":2},{"title":"다음 단계","local":"다음-단계","sections":[],"depth":2}],"depth":1}';function as(Re){return At(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class cs extends Pt{constructor(d){super(),qt(this,d,as,is,Yt,{})}}export{cs as component};