Buckets:

hf-doc-build
/

doc

Files
xet
hf-doc-build/doc / diffusers /main /ko /training /t2i_adapters.html
mishig
2 days ago
download
raw
40.7 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;T2I-Adapter&quot;,&quot;local&quot;:&quot;t2i-adapter&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;스크립트 매개변수&quot;,&quot;local&quot;:&quot;script-parameters&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;학습 스크립트&quot;,&quot;local&quot;:&quot;training-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;스크립트 실행&quot;,&quot;local&quot;:&quot;launch-the-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;다음 단계&quot;,&quot;local&quot;:&quot;next-steps&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/>
		<link href="/docs/diffusers/main/en/_app/immutable/entry/start.CM-2NerT.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/CTWHa_kr.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/BLi2et3c.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/entry/app.DrYcqW7m.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/BeiZzCMo.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DnVIJfC5.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DsnmJJEf.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/t03vT2vk.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/nodes/0.DdeITYyU.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DAwShlVY.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/nodes/15.CnDb3cX_.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/8Bi71p98.js" rel="modulepreload">
		<!--1vsm9a6--><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;T2I-Adapter&quot;,&quot;local&quot;:&quot;t2i-adapter&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Script parameters&quot;,&quot;local&quot;:&quot;script-parameters&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Training script&quot;,&quot;local&quot;:&quot;training-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Launch the script&quot;,&quot;local&quot;:&quot;launch-the-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Next steps&quot;,&quot;local&quot;:&quot;next-steps&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/><!---->
		<link href="/docs/diffusers/main/en/_app/immutable/assets/0.tn0RQdqM.css" rel="modulepreload"> <!--[--><!--[0--><!--[--><!--[0--><!--[--><p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: inline-end; margin-inline-start: 10px; display: inline-flex; position: relative; z-index: 10;"><div dir="ltr" class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----></span> <span dir="auto">Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-7 max-sm:h-7 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible sm:size-3.5 size-3 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg><!----></button></div> <!--[-1--><!--]--></div><!----> <!--[0--><h1 class="relative group"><a id="t2i-adapter" 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" href="#t2i-adapter"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->T2I-Adapter<!----></span></h1><!--]--><!----> <p><a href="https://hf.co/papers/2302.08453" rel="nofollow">T2I-Adapter</a>는 이미지 생성을 더 잘 제어할 수 있도록 추가 조건 입력 이미지(라인 아트, canny, 스케치, 깊이, 포즈)를 제공하는 경량 어댑터 모델입니다. ControlNet과 유사하지만, UNet을 복사하여 학습하는 대신 가중치만 삽입하기 때문에 훨씬 작습니다(약 77M 파라미터, 약 300MB 파일 크기).</p> <p>T2I-Adapter는 Stable Diffusion XL (SDXL) 모델로만 학습할 수 있습니다.</p> <p>이 가이드에서는 <a href="https://github.com/huggingface/diffusers/blob/main/examples/t2i_adapter/train_t2i_adapter_sdxl.py" rel="nofollow">train_t2i_adapter_sdxl.py</a> 학습 스크립트를 살펴보고, 이를 이해한 후 자신의 사용 사례에 맞게 수정하는 방법을 안내합니다.</p> <p>…</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!---->git <span class="hljs-built_in">clone</span> https://github.com/huggingface/diffusers
<span class="hljs-built_in">cd</span> diffusers
pip install .<!----></pre></div><!----> <p>이후 학습 스크립트가 포함된 예제 폴더로 이동하여 사용하는 스크립트에 필요한 의존성을 설치합니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!----><span class="hljs-built_in">cd</span> examples/t2i_adapter
pip install -r requirements.txt<!----></pre></div><!----> <blockquote class="tip"><p>🤗 Accelerate는 여러 GPU/TPU 또는 혼합 정밀도로 학습하는 것을 돕는 라이브러리입니다. 하드웨어와 환경에 따라 학습 설정을 자동으로 구성합니다. 자세한 내용은 🤗 Accelerate <a href="https://huggingface.co/docs/accelerate/quicktour" rel="nofollow">Quick tour</a>를 참고하세요.</p></blockquote> <p>🤗 Accelerate 환경을 초기화합니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!---->accelerate config<!----></pre></div><!----> <p>어떤 구성도 선택하지 않고 기본 🤗 Accelerate 환경을 설정하려면:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!---->accelerate config default<!----></pre></div><!----> <p>또는 노트북과 같이 인터랙티브 셸을 지원하지 않는 환경에서는 다음을 사용할 수 있습니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-py "><!----><span class="hljs-keyword">from</span> accelerate.utils <span class="hljs-keyword">import</span> write_basic_config

write_basic_config()<!----></pre></div><!----> <p>마지막으로, 자체 데이터셋으로 모델을 학습하려면 <a href="create_dataset">학습용 데이터셋 만들기</a> 가이드를 참고하여 학습 스크립트와 호환되는 데이터셋을 만드는 방법을 배워보세요.</p> <blockquote class="tip"><p>다음 섹션에서는 스크립트를 수정하는 방법을 이해하는 데 중요한 학습 스크립트의 일부 부분을 강조합니다. 그러나 스크립트의 모든 측면을 자세히 다루지는 않습니다. 더 자세히 알고 싶으시다면 <a href="https://github.com/huggingface/diffusers/blob/main/examples/t2i_adapter/train_t2i_adapter_sdxl.py" rel="nofollow">스크립트</a>를 읽어보시고, 질문이나 우려 사항이 있으시면 알려주시기 바랍니다.</p></blockquote> <!--[1--><h2 class="relative group"><a id="script-parameters" 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" href="#script-parameters"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->스크립트 매개변수<!----></span></h2><!--]--><!----> <p>학습 스크립트는 학습 실행을 사용자 정의하는 데 도움이 되는 많은 매개변수를 제공합니다. 모든 매개변수와 그 설명은 <a href="https://github.com/huggingface/diffusers/blob/aab6de22c33cc01fb7bc81c0807d6109e2c998c9/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L233" rel="nofollow"><code>parse_args()</code></a> 함수에서 찾을 수 있습니다. 학습 배치 크기나 학습률과 같이 각 매개변수에 대한 기본값을 제공하지만, 원하시는 경우 학습 명령에서 직접 값을 설정할 수도 있습니다.</p> <p>예를 들어, gradient accumulation을 활성화하려면 학습 명령에 <code>--gradient_accumulation_steps</code> 매개변수를 추가합니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!---->accelerate launch train_t2i_adapter_sdxl.py \
  ----gradient_accumulation_steps=4<!----></pre></div><!----> <p>기본적이고 중요한 대부분의 매개변수는 <a href="text2image#script-parameters">텍스트-이미지</a> 학습 가이드에 설명되어 있으므로, 이 가이드에서는 관련 T2I-Adapter 매개변수에만 초점을 맞춥니다:</p> <ul><li><code>--pretrained_vae_model_name_or_path</code>: 사전 학습된 VAE의 경로; SDXL VAE는 수치적 불안정성이 있는 것으로 알려져 있으므로, 이 매개변수를 통해 더 나은 <a href="https://huggingface.co/madebyollin/sdxl-vae-fp16-fix" rel="nofollow">VAE</a>를 지정할 수 있습니다</li> <li><code>--crops_coords_top_left_h</code>와 <code>--crops_coords_top_left_w</code>: SDXL의 crop 좌표 임베딩에 포함할 높이와 너비 좌표</li> <li><code>--conditioning_image_column</code>: 데이터셋에서 conditioning image가 포함된 열</li> <li><code>--proportion_empty_prompts</code>: 빈 문자열로 대체할 이미지 프롬프트의 비율</li></ul> <!--[1--><h2 class="relative group"><a id="training-script" 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" href="#training-script"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->학습 스크립트<!----></span></h2><!--]--><!----> <p>스크립트 매개변수와 마찬가지로, <a href="text2image#training-script">텍스트-이미지</a> 학습 가이드에서 학습 스크립트의 전체적인 흐름을 안내합니다. 대신 이 가이드에서는 스크립트 중 T2I-Adapter와 관련된 부분을 살펴봅니다.</p> <p>학습 스크립트는 데이터셋을 준비하는 것으로 시작합니다. 여기에는 프롬프트를 <a href="https://github.com/huggingface/diffusers/blob/aab6de22c33cc01fb7bc81c0807d6109e2c998c9/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L674" rel="nofollow">토큰화</a>하고 이미지와 조건 이미지에 <a href="https://github.com/huggingface/diffusers/blob/aab6de22c33cc01fb7bc81c0807d6109e2c998c9/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L714" rel="nofollow">변환을 적용</a>하는 작업이 포함됩니다.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-py "><!---->conditioning_image_transforms = transforms.Compose(
    [
        transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
        transforms.CenterCrop(args.resolution),
        transforms.ToTensor(),
    ]
)<!----></pre></div><!----> <p><a href="https://github.com/huggingface/diffusers/blob/aab6de22c33cc01fb7bc81c0807d6109e2c998c9/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L770" rel="nofollow"><code>main()</code></a> 함수 내에서 T2I-Adapter는 사전 학습된 어댑터에서 로드되거나 무작위로 초기화됩니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-py "><!----><span class="hljs-keyword">if</span> args.adapter_model_name_or_path:
    logger.info(<span class="hljs-string">&quot;Loading existing adapter weights.&quot;</span>)
    t2iadapter = T2IAdapter.from_pretrained(args.adapter_model_name_or_path)
<span class="hljs-keyword">else</span>:
    logger.info(<span class="hljs-string">&quot;Initializing t2iadapter weights.&quot;</span>)
    t2iadapter = T2IAdapter(
        in_channels=<span class="hljs-number">3</span>,
        channels=(<span class="hljs-number">320</span>, <span class="hljs-number">640</span>, <span class="hljs-number">1280</span>, <span class="hljs-number">1280</span>),
        num_res_blocks=<span class="hljs-number">2</span>,
        downscale_factor=<span class="hljs-number">16</span>,
        adapter_type=<span class="hljs-string">&quot;full_adapter_xl&quot;</span>,
    )<!----></pre></div><!----> <p><a href="https://github.com/huggingface/diffusers/blob/aab6de22c33cc01fb7bc81c0807d6109e2c998c9/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L952" rel="nofollow">optimizer</a>는 T2I-Adapter의 매개변수에 대해 초기화됩니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-py "><!---->params_to_optimize = t2iadapter.parameters()
optimizer = optimizer_class(
    params_to_optimize,
    lr=args.learning_rate,
    betas=(args.adam_beta1, args.adam_beta2),
    weight_decay=args.adam_weight_decay,
    eps=args.adam_epsilon,
)<!----></pre></div><!----> <p>마지막으로, <a href="https://github.com/huggingface/diffusers/blob/aab6de22c33cc01fb7bc81c0807d6109e2c998c9/examples/t2i_adapter/train_t2i_adapter_sdxl.py#L1086" rel="nofollow">training loop</a>에서는 어댑터 조건 이미지와 텍스트 임베딩을 UNet에 전달하여 노이즈 잔차를 예측합니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-py "><!---->t2iadapter_image = batch[<span class="hljs-string">&quot;conditioning_pixel_values&quot;</span>].to(dtype=weight_dtype)
down_block_additional_residuals = t2iadapter(t2iadapter_image)
down_block_additional_residuals = [
    sample.to(dtype=weight_dtype) <span class="hljs-keyword">for</span> sample <span class="hljs-keyword">in</span> down_block_additional_residuals
]

model_pred = unet(
    inp_noisy_latents,
    timesteps,
    encoder_hidden_states=batch[<span class="hljs-string">&quot;prompt_ids&quot;</span>],
    added_cond_kwargs=batch[<span class="hljs-string">&quot;unet_added_conditions&quot;</span>],
    down_block_additional_residuals=down_block_additional_residuals,
).sample<!----></pre></div><!----> <p>학습 루프가 어떻게 작동하는지 더 자세히 알고 싶다면, 디노이징 과정의 기본 패턴을 단계별로 설명하는 <a href="../using-diffusers/write_own_pipeline">Understanding pipelines, models and schedulers</a> 튜토리얼을 확인해 보세요.</p> <!--[1--><h2 class="relative group"><a id="launch-the-script" 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" href="#launch-the-script"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->스크립트 실행<!----></span></h2><!--]--><!----> <p>이제 학습 스크립트를 실행할 준비가 되었습니다! 🚀</p> <p>이 예제 학습에서는 <a href="https://huggingface.co/datasets/fusing/fill50k" rel="nofollow">fusing/fill50k</a> 데이터셋을 사용합니다. 원하시는 경우 직접 데이터셋을 생성하여 사용할 수도 있습니다(<a href="create_dataset">학습용 데이터셋 만들기</a> 가이드 참고).</p> <p>환경 변수 <code>MODEL_DIR</code>을 Hub의 모델 ID 또는 로컬 모델 경로로, <code>OUTPUT_DIR</code>을 모델을 저장할 위치로 설정합니다.</p> <p>다음 이미지들을 다운로드하여 학습을 조건화하세요:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!---->wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png
wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_2.png<!----></pre></div><!----> <blockquote class="tip"><p>Weights &amp; Biases로 학습 진행 상황을 모니터링하려면 학습 명령에 <code>--report_to=wandb</code> 매개변수를 추가하세요. 결과를 추적하려면 학습 명령에 <code>--validation_image</code>, <code>--validation_prompt</code>, <code>--validation_steps</code>도 추가해야 합니다. 이는 모델 디버깅 및 중간 결과 확인에 매우 유용합니다.</p></blockquote> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-bash "><!----><span class="hljs-built_in">export</span> MODEL_DIR=<span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>
<span class="hljs-built_in">export</span> OUTPUT_DIR=<span class="hljs-string">&quot;path to save model&quot;</span>

accelerate launch train_t2i_adapter_sdxl.py \
 --pretrained_model_name_or_path=<span class="hljs-variable">$MODEL_DIR</span> \
 --output_dir=<span class="hljs-variable">$OUTPUT_DIR</span> \
 --dataset_name=fusing/fill50k \
 --mixed_precision=<span class="hljs-string">&quot;fp16&quot;</span> \
 --resolution=1024 \
 --learning_rate=1e-5 \
 --max_train_steps=15000 \
 --validation_image <span class="hljs-string">&quot;./conditioning_image_1.png&quot;</span> <span class="hljs-string">&quot;./conditioning_image_2.png&quot;</span> \
 --validation_prompt <span class="hljs-string">&quot;red circle with blue background&quot;</span> <span class="hljs-string">&quot;cyan circle with brown floral background&quot;</span> \
 --validation_steps=100 \
 --train_batch_size=1 \
 --gradient_accumulation_steps=4 \
 --report_to=<span class="hljs-string">&quot;wandb&quot;</span> \
 --seed=42 \
 --push_to_hub<!----></pre></div><!----> <p>학습이 완료되면 추론에 T2I-Adapter를 사용할 수 있습니다:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600  " title="code excerpt" type="button"><svg xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg><!----> <div class=" absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0 "><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent;"></div> 복사됨</div><!----></button><!----></div> <pre class="language-py "><!----><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> StableDiffusionXLAdapterPipeline, T2IAdapter, EulerAncestralDiscreteSchedulerTest
<span class="hljs-keyword">from</span> diffusers.utils <span class="hljs-keyword">import</span> load_image
<span class="hljs-keyword">import</span> torch

adapter = T2IAdapter.from_pretrained(<span class="hljs-string">&quot;path/to/adapter&quot;</span>, dtype=torch.float16)
pipeline = StableDiffusionXLAdapterPipeline.from_pretrained(
    <span class="hljs-string">&quot;stabilityai/stable-diffusion-xl-base-1.0&quot;</span>, adapter=adapter, dtype=torch.float16
)

pipeline.scheduler = EulerAncestralDiscreteSchedulerTest.from_config(pipe.scheduler.config)
pipeline.enable_xformers_memory_efficient_attention()
pipeline.enable_model_cpu_offload()

control_image = load_image(<span class="hljs-string">&quot;./conditioning_image_1.png&quot;</span>)
prompt = <span class="hljs-string">&quot;pale golden rod circle with old lace background&quot;</span>

generator = torch.manual_seed(<span class="hljs-number">0</span>)
image = pipeline(
    prompt, image=control_image, generator=generator
).images[<span class="hljs-number">0</span>]
image.save(<span class="hljs-string">&quot;./output.png&quot;</span>)<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="next-steps" 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" href="#next-steps"><span><svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg><!----></span></a> <span><!---->다음 단계<!----></span></h2><!--]--><!----> <p>T2I-Adapter 모델 학습을 완료하신 것을 축하드립니다! 🎉 더 자세히 알아보려면:</p> <ul><li>T2I-Adapter 팀의 실험 결과에 대한 자세한 내용은 <a href="https://huggingface.co/blog/t2i-sdxl-adapters" rel="nofollow">SDXL에서 T2I-Adapters를 활용한 효율적이고 제어 가능한 생성</a> 블로그 포스트를 참고하세요.</li></ul> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/diffusers/blob/main/docs/source/en/training/t2i_adapters.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg><!----> <span><span class="underline">Update</span> on GitHub</span></a><!----> <p></p><!--]--><!----><!--]--><!--]--><!--]--> <!--[-1--><!--]--><!--]-->
			
			<script>
				{
					__sveltekit_1u9qigp = {
						base: "/docs/diffusers/main/en",
						assets: "/docs/diffusers/main/en"
					};
					const element = document.currentScript.parentElement;
					Promise.all([
						import("/docs/diffusers/main/en/_app/immutable/entry/start.CM-2NerT.js"),
						import("/docs/diffusers/main/en/_app/immutable/entry/app.DrYcqW7m.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 15],
							data: [null,null],
							form: null,
							error: null
						});
					});
				}
			</script>
		
Xet Storage Details

Size:: 40.7 kB
Xet hash:: a203464f57e0d68a687788bebaf186b2529fa51634509108770bdc734e6dd2e0
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.