Buckets:

hf-doc-build
/

doc-dev

Files

xet

hf-doc-build/doc-dev / huggingface_hub /pr_4113 /ko /guides /inference_endpoints.html

HuggingFaceDocBuilder

4 months ago

download

raw

65.8 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"추론 엔드포인트","local":"inference-endpoints","sections":[{"title":"추론 엔드포인트 생성","local":"create-an-inference-endpoint","sections":[{"title":"사용자 정의 이미지 사용","local":"using-a-custom-image","sections":[],"depth":4},{"title":"기존 추론 엔드포인트 가져오기 또는 리스트 조회","local":"get-or-list-existing-inference-endpoints","sections":[],"depth":3}],"depth":2},{"title":"배포 상태 확인","local":"check-deployment-status","sections":[],"depth":2},{"title":"추론 실행","local":"run-inference","sections":[],"depth":2},{"title":"라이프사이클 관리","local":"manage-lifecycle","sections":[{"title":"일시 중지 또는 0으로 확장","local":"pause-or-scale-to-zero","sections":[],"depth":3},{"title":"모델 또는 하드웨어 요구 사항 업데이트","local":"update-model-or-hardware-requirements","sections":[],"depth":3},{"title":"엔드포인트 삭제","local":"delete-the-endpoint","sections":[],"depth":3}],"depth":2},{"title":"엔드 투 엔드 예제 an-end-to-end-example","local":"엔드-투-엔드-예제-an-end-to-end-example","sections":[],"depth":2}],"depth":1}">
	<link href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/entry/start.e4d70577.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/scheduler.0cf4ef2e.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/singletons.7ad9a7dd.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/paths.5be5312b.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/entry/app.3b0f3b68.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/preload-helper.2e746cb0.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/index.abc433bf.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/nodes/0.c6150a55.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/each.e59479a4.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/nodes/8.51fd3a23.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.b2531632.js">
	<link rel="modulepreload" href="/docs/huggingface_hub/pr_4113/ko/_app/immutable/chunks/CodeBlock.db16c7c8.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"추론 엔드포인트","local":"inference-endpoints","sections":[{"title":"추론 엔드포인트 생성","local":"create-an-inference-endpoint","sections":[{"title":"사용자 정의 이미지 사용","local":"using-a-custom-image","sections":[],"depth":4},{"title":"기존 추론 엔드포인트 가져오기 또는 리스트 조회","local":"get-or-list-existing-inference-endpoints","sections":[],"depth":3}],"depth":2},{"title":"배포 상태 확인","local":"check-deployment-status","sections":[],"depth":2},{"title":"추론 실행","local":"run-inference","sections":[],"depth":2},{"title":"라이프사이클 관리","local":"manage-lifecycle","sections":[{"title":"일시 중지 또는 0으로 확장","local":"pause-or-scale-to-zero","sections":[],"depth":3},{"title":"모델 또는 하드웨어 요구 사항 업데이트","local":"update-model-or-hardware-requirements","sections":[],"depth":3},{"title":"엔드포인트 삭제","local":"delete-the-endpoint","sections":[],"depth":3}],"depth":2},{"title":"엔드 투 엔드 예제 an-end-to-end-example","local":"엔드-투-엔드-예제-an-end-to-end-example","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div 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>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> </div> <h1 class="relative group"><a id="inference-endpoints" 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="#inference-endpoints"><span><svg class="" 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>추론 엔드포인트</span></h1> <p data-svelte-h="svelte-ah66md">추론 엔드포인트는 Hugging Face가 관리하는 전용 및 자동 확장 인프라에 <code>transformers</code>, <code>sentence-transformers</code> 및 <code>diffusers</code> 모델을 쉽게 배포할 수 있는 안전한 프로덕션 솔루션을 제공합니다. 추론 엔드포인트는 <a href="https://huggingface.co/models" rel="nofollow">Hub</a>의 모델로 구축됩니다.
	이 가이드에서는 <code>huggingface_hub</code>를 사용하여 프로그래밍 방식으로 추론 엔드포인트를 관리하는 방법을 배웁니다. 추론 엔드포인트 제품 자체에 대한 자세한 내용은 <a href="https://huggingface.co/docs/inference-endpoints/index" rel="nofollow">공식 문서</a>를 참조하세요.</p> <p data-svelte-h="svelte-1ql8zue">이 가이드에서는 <code>huggingface_hub</code>가 올바르게 설치 및 로그인되어 있다고 가정합니다. 아직 그렇지 않은 경우 <a href="https://huggingface.co/docs/huggingface_hub/quick-start#quickstart" rel="nofollow">빠른 시작 가이드</a>를 참조하세요. 추론 엔드포인트 API를 지원하는 최소 버전은 <code>v0.19.0</code>입니다.</p> <h2 class="relative group"><a id="create-an-inference-endpoint" 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="#create-an-inference-endpoint"><span><svg class="" 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>추론 엔드포인트 생성</span></h2> <p data-svelte-h="svelte-13xsd97">첫 번째 단계는 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.create_inference_endpoint">create_inference_endpoint()</a>를 사용하여 추론 엔드포인트를 생성하는 것입니다:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> create_inference_endpoint

	<span class="hljs-meta">>>> </span>endpoint = create_inference_endpoint(
	<span class="hljs-meta">... </span> <span class="hljs-string">"my-endpoint-name"</span>,
	<span class="hljs-meta">... </span> repository=<span class="hljs-string">"gpt2"</span>,
	<span class="hljs-meta">... </span> framework=<span class="hljs-string">"pytorch"</span>,
	<span class="hljs-meta">... </span> task=<span class="hljs-string">"text-generation"</span>,
	<span class="hljs-meta">... </span> accelerator=<span class="hljs-string">"cpu"</span>,
	<span class="hljs-meta">... </span> vendor=<span class="hljs-string">"aws"</span>,
	<span class="hljs-meta">... </span> region=<span class="hljs-string">"us-east-1"</span>,
	<span class="hljs-meta">... </span> <span class="hljs-built_in">type</span>=<span class="hljs-string">"protected"</span>,
	<span class="hljs-meta">... </span> instance_size=<span class="hljs-string">"x2"</span>,
	<span class="hljs-meta">... </span> instance_type=<span class="hljs-string">"intel-icl"</span>
	<span class="hljs-meta">... </span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1aqjns7">예시에서는 <code>"my-endpoint-name"</code>라는 <code>protected</code> 추론 엔드포인트를 생성하여 <code>text-generation</code>을 위한 <a href="https://huggingface.co/gpt2" rel="nofollow">gpt2</a>를 제공합니다. <code>protected</code> 추론 엔드포인트 API에 액세스하려면 토큰이 필요합니다. 또한 벤더, 지역, 액셀러레이터, 인스턴스 유형, 크기와 같은 하드웨어 요구 사항을 구성하기 위한 추가 정보를 제공해야 합니다. 사용 가능한 리소스 목록은 <a href="https://api.endpoints.huggingface.cloud/#/v2%3A%3Aprovider/list_vendors" rel="nofollow">여기</a>에서 확인할 수 있습니다. 또한 <a href="https://ui.endpoints.huggingface.co/new" rel="nofollow">웹 인터페이스</a>를 사용하여 편리하게 수동으로 추론 엔드포인트를 생성할 수 있습니다. 고급 설정 및 사용법에 대한 자세한 내용은 <a href="https://huggingface.co/docs/inference-endpoints/guides/advanced" rel="nofollow">이 가이드</a>를 참조하세요.</p> <p data-svelte-h="svelte-1qi1t71"><a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.create_inference_endpoint">create_inference_endpoint()</a>에서 반환된 값은 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint">InferenceEndpoint</a> 개체입니다:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span>endpoint
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-qy7mei">이것은 엔드포인트에 대한 정보를 저장하는 데이터클래스입니다. <code>name</code>, <code>repository</code>, <code>status</code>, <code>task</code>, <code>created_at</code>, <code>updated_at</code> 등과 같은 중요한 속성에 접근할 수 있습니다. 필요한 경우 <code>endpoint.raw</code>를 통해 서버로부터의 원시 응답에도 접근할 수 있습니다.</p> <p data-svelte-h="svelte-q97gnz">추론 엔드포인트가 생성되면 <a href="https://ui.endpoints.huggingface.co/" rel="nofollow">개인 대시보드</a>에서 확인할 수 있습니다.</p> <p data-svelte-h="svelte-1tajtyl"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/huggingface_hub/inference_endpoints_created.png"></p> <h4 class="relative group"><a id="using-a-custom-image" 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="#using-a-custom-image"><span><svg class="" 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>사용자 정의 이미지 사용</span></h4> <p data-svelte-h="svelte-52wlj1">기본적으로 추론 엔드포인트는 Hugging Face에서 제공하는 도커 이미지로 구축됩니다. 그러나 <code>custom_image</code> 매개변수를 사용하여 모든 도커 이미지를 지정할 수 있습니다. 일반적인 사용 사례는 <a href="https://github.com/huggingface/text-generation-inference" rel="nofollow">text-generation-inference</a> 프레임워크를 사용하여 LLM을 실행하는 것입니다. 다음과 같이 수행할 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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># TGI에서 Zephyr-7b-beta를 실행하는 추론 엔드포인트 시작하기</span>
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> create_inference_endpoint
	<span class="hljs-meta">>>> </span>endpoint = create_inference_endpoint(
	<span class="hljs-meta">... </span> <span class="hljs-string">"aws-zephyr-7b-beta-0486"</span>,
	<span class="hljs-meta">... </span> repository=<span class="hljs-string">"HuggingFaceH4/zephyr-7b-beta"</span>,
	<span class="hljs-meta">... </span> framework=<span class="hljs-string">"pytorch"</span>,
	<span class="hljs-meta">... </span> task=<span class="hljs-string">"text-generation"</span>,
	<span class="hljs-meta">... </span> accelerator=<span class="hljs-string">"gpu"</span>,
	<span class="hljs-meta">... </span> vendor=<span class="hljs-string">"aws"</span>,
	<span class="hljs-meta">... </span> region=<span class="hljs-string">"us-east-1"</span>,
	<span class="hljs-meta">... </span> <span class="hljs-built_in">type</span>=<span class="hljs-string">"protected"</span>,
	<span class="hljs-meta">... </span> instance_size=<span class="hljs-string">"x1"</span>,
	<span class="hljs-meta">... </span> instance_type=<span class="hljs-string">"nvidia-a10g"</span>,
	<span class="hljs-meta">... </span> custom_image={
	<span class="hljs-meta">... </span> <span class="hljs-string">"health_route"</span>: <span class="hljs-string">"/health"</span>,
	<span class="hljs-meta">... </span> <span class="hljs-string">"env"</span>: {
	<span class="hljs-meta">... </span> <span class="hljs-string">"MAX_BATCH_PREFILL_TOKENS"</span>: <span class="hljs-string">"2048"</span>,
	<span class="hljs-meta">... </span> <span class="hljs-string">"MAX_INPUT_LENGTH"</span>: <span class="hljs-string">"1024"</span>,
	<span class="hljs-meta">... </span> <span class="hljs-string">"MAX_TOTAL_TOKENS"</span>: <span class="hljs-string">"1512"</span>,
	<span class="hljs-meta">... </span> <span class="hljs-string">"MODEL_ID"</span>: <span class="hljs-string">"/repository"</span>
	<span class="hljs-meta">... </span> },
	<span class="hljs-meta">... </span> <span class="hljs-string">"url"</span>: <span class="hljs-string">"ghcr.io/huggingface/text-generation-inference:1.1.0"</span>,
	<span class="hljs-meta">... </span> },
	<span class="hljs-meta">... </span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-kvtuxg"><code>custom_image</code>에 전달할 값은 도커 컨테이너의 URL과 이를 실행하기 위한 구성이 포함된 딕셔너리입니다. 자세한 내용은 <a href="https://api.endpoints.huggingface.cloud/#/v2%3A%3Aendpoint/create_endpoint" rel="nofollow">Swagger 문서</a>를 참조하세요.</p> <h3 class="relative group"><a id="get-or-list-existing-inference-endpoints" 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="#get-or-list-existing-inference-endpoints"><span><svg class="" 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>기존 추론 엔드포인트 가져오기 또는 리스트 조회</span></h3> <p data-svelte-h="svelte-51j956">경우에 따라 이전에 생성한 추론 엔드포인트를 관리해야 할 수 있습니다. 이름을 알고 있는 경우 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.get_inference_endpoint">get_inference_endpoint()</a>를 사용하여 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint">InferenceEndpoint</a> 개체를 가져올 수 있습니다. 또는 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.list_inference_endpoints">list_inference_endpoints()</a>를 사용하여 모든 추론 엔드포인트 리스트를 검색할 수 있습니다. 두 메소드 모두 선택적 <code>namespace</code> 매개변수를 허용합니다. 속해 있는 조직의 <code>namespace</code>를 설정할 수 있습니다. 그렇지 않으면 기본적으로 사용자 이름이 사용됩니다.</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> get_inference_endpoint, list_inference_endpoints

	<span class="hljs-comment"># 엔드포인트 개체 가져오기</span>
	<span class="hljs-meta">>>> </span>get_inference_endpoint(<span class="hljs-string">"my-endpoint-name"</span>)
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)

	<span class="hljs-comment"># 조직의 모든 추론 엔드포인트 나열</span>
	<span class="hljs-meta">>>> </span>list_inference_endpoints(namespace=<span class="hljs-string">"huggingface"</span>)
	[InferenceEndpoint(name=<span class="hljs-string">'aws-starchat-beta'</span>, namespace=<span class="hljs-string">'huggingface'</span>, repository=<span class="hljs-string">'HuggingFaceH4/starchat-beta'</span>, status=<span class="hljs-string">'paused'</span>, url=<span class="hljs-literal">None</span>), ...]

	<span class="hljs-comment"># 사용자가 속해있는 모든 조직의 엔드포인트 나열</span>
	<span class="hljs-meta">>>> </span>list_inference_endpoints(namespace=<span class="hljs-string">"*"</span>)
	[InferenceEndpoint(name=<span class="hljs-string">'aws-starchat-beta'</span>, namespace=<span class="hljs-string">'huggingface'</span>, repository=<span class="hljs-string">'HuggingFaceH4/starchat-beta'</span>, status=<span class="hljs-string">'paused'</span>, url=<span class="hljs-literal">None</span>), ...]<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="check-deployment-status" 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="#check-deployment-status"><span><svg class="" 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>배포 상태 확인</span></h2> <p data-svelte-h="svelte-9go20c">이 가이드의 나머지 부분에서는 <code>endpoint</code>라는 이름의 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint">InferenceEndpoint</a> 객체를 가지고 있다고 가정합니다. 엔드포인트에 <code>status</code> 속성이 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpointStatus">InferenceEndpointStatus</a> 유형이라는 것을 알 수 있었습니다. 추론 엔드포인트가 배포되고 접근 가능하면 상태가 <code>"running"</code>이 되고 <code>url</code> 속성이 설정됩니다:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span>endpoint
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'running'</span>, url=<span class="hljs-string">'https://jpj7k2q4j805b727.us-east-1.aws.endpoints.huggingface.cloud'</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-eadeby"><code>추론 엔드포인트가 "running"</code> 상태에 도달하기 전에 일반적으로 <code>"initializing"</code> 또는 <code>"pending"</code> 단계를 거칩니다. <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.fetch">fetch()</a>를 실행하여 엔드포인트의 새로운 상태를 가져올 수 있습니다. <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint">InferenceEndpoint</a>의 다른 메소드와 마찬가지로 이 메소드는 서버에 요청을 하며, <code>endpoint</code>의 내부 속성이 변경됩니다:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span>endpoint.fetch()
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-n0ojtq">추론 엔드포인트가 실행될 때까지 기다리면서 상태를 가져오는 대신 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.wait">wait()</a>를 직접 호출할 수 있습니다. 이 헬퍼는 <code>timeout</code>과 <code>fetch_every</code> 매개변수를 입력으로 받아 (초 단위) 추론 엔드포인트가 배포될 때까지 스레드를 차단합니다. 기본값은 각각 <code>None</code>(제한 시간 없음)과 <code>5</code>초입니다.</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># 엔드포인트 보류</span>
	<span class="hljs-meta">>>> </span>endpoint
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)

	<span class="hljs-comment"># 10초 대기 => InferenceEndpointTimeoutError 발생</span>
	<span class="hljs-meta">>>> </span>endpoint.wait(timeout=<span class="hljs-number">10</span>)
	<span class="hljs-keyword">raise</span> InferenceEndpointTimeoutError(<span class="hljs-string">"Timeout while waiting for Inference Endpoint to be deployed."</span>)
	huggingface_hub._inference_endpoints.InferenceEndpointTimeoutError: Timeout <span class="hljs-keyword">while</span> waiting <span class="hljs-keyword">for</span> Inference Endpoint to be deployed.

	<span class="hljs-comment"># 추가 대기</span>
	<span class="hljs-meta">>>> </span>endpoint.wait()
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'running'</span>, url=<span class="hljs-string">'https://jpj7k2q4j805b727.us-east-1.aws.endpoints.huggingface.cloud'</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1su15dj"><code>timeout</code>이 설정되어 있고 추론 엔드포인트를 불러오는 데 너무 오래 걸리면, <code>InferenceEndpointTimeoutError</code> 제한 시간 초과 오류가 발생합니다.</p> <h2 class="relative group"><a id="run-inference" 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="#run-inference"><span><svg class="" 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>추론 실행</span></h2> <p data-svelte-h="svelte-mbyh0u">추론 엔드포인트가 실행되면, 마침내 추론을 실행할 수 있습니다!</p> <p data-svelte-h="svelte-1cbasi5"><a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint">InferenceEndpoint</a>에는 각각 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_client#huggingface_hub.InferenceClient">InferenceClient</a>와 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_client#huggingface_hub.AsyncInferenceClient">AsyncInferenceClient</a>를 반환하는 <code>client</code>와 <code>async_client</code> 속성이 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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># 텍스트 생성 작업 실행:</span>
	<span class="hljs-meta">>>> </span>endpoint.client.text_generation(<span class="hljs-string">"I am"</span>)
	<span class="hljs-string">' not a fan of the idea of a "big-budget" movie. I think it\'s a'</span>

	<span class="hljs-comment"># 비동기 컨텍스트에서도 마찬가지로 실행:</span>
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">await</span> endpoint.async_client.text_generation(<span class="hljs-string">"I am"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-kx3rup">추론 엔드포인트가 실행 중이 아니면 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpointError">InferenceEndpointError</a> 오류가 발생합니다:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span>endpoint.client
	huggingface_hub._inference_endpoints.InferenceEndpointError: Cannot create a client <span class="hljs-keyword">for</span> this Inference Endpoint <span class="hljs-keyword">as</span> it <span class="hljs-keyword">is</span> <span class="hljs-keyword">not</span> yet deployed. Please wait <span class="hljs-keyword">for</span> the Inference Endpoint to be deployed using `endpoint.wait()` <span class="hljs-keyword">and</span> <span class="hljs-keyword">try</span> again.<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1iw98ah"><a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_client#huggingface_hub.InferenceClient">InferenceClient</a>를 사용하는 방법에 대한 자세한 내용은 <a href="../guides/inference">추론 가이드</a>를 참조하세요.</p> <h2 class="relative group"><a id="manage-lifecycle" 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="#manage-lifecycle"><span><svg class="" 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>라이프사이클 관리</span></h2> <p data-svelte-h="svelte-d4vepm">이제 추론 엔드포인트를 생성하고 추론을 실행하는 방법을 살펴보았으니, 라이프사이클을 관리하는 방법을 살펴봅시다.</p> <blockquote class="tip" data-svelte-h="svelte-3ofzf"><p>이 섹션에서는 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.pause">pause()</a>, <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.resume">resume()</a>, <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.scale_to_zero">scale_to_zero()</a>, <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.update">update()</a> 및 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.delete">delete()</a> 등의 메소드를 살펴볼 것입니다. 모든 메소드는 편의를 위해 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint">InferenceEndpoint</a>에 추가된 별칭입니다. 원한다면 <code>HfApi</code>에 정의된 일반 메소드 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.pause_inference_endpoint">pause_inference_endpoint()</a>, <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.resume_inference_endpoint">resume_inference_endpoint()</a>, <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.scale_to_zero_inference_endpoint">scale_to_zero_inference_endpoint()</a>, <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.update_inference_endpoint">update_inference_endpoint()</a> 및 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/hf_api#huggingface_hub.HfApi.delete_inference_endpoint">delete_inference_endpoint()</a>를 사용할 수도 있습니다.</p></blockquote> <h3 class="relative group"><a id="pause-or-scale-to-zero" 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="#pause-or-scale-to-zero"><span><svg class="" 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>일시 중지 또는 0으로 확장</span></h3> <p data-svelte-h="svelte-gzqooj">추론 엔드포인트를 사용하지 않을 때 비용을 절감하기 위해 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.pause">pause()</a>를 사용하여 일시 중지하거나 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.scale_to_zero">scale_to_zero()</a>를 사용하여 0으로 스케일링할 수 있습니다.</p> <blockquote class="tip" data-svelte-h="svelte-1b726qd"><p><em>일시 중지</em> 또는 <em>0으로 스케일링</em>된 추론 엔드포인트는 비용이 들지 않습니다. 이 두 가지의 차이점은 <em>일시 중지</em> 엔드포인트는 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.resume">resume()</a>를 사용하여 명시적으로 <em>재개</em>해야 한다는 것입니다. 반대로 <em>0으로 스케일링</em>된 엔드포인트는 추론 호출이 있으면 추가 콜드 스타트 지연과 함께 자동으로 시작됩니다. 추론 엔드포인트는 일정 기간 비활성화된 후 자동으로 0으로 스케일링되도록 구성할 수도 있습니다.</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># 엔드포인트 일시중지 및 재시작</span>
	<span class="hljs-meta">>>> </span>endpoint.pause()
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'paused'</span>, url=<span class="hljs-literal">None</span>)
	<span class="hljs-meta">>>> </span>endpoint.resume()
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)
	<span class="hljs-meta">>>> </span>endpoint.wait().client.text_generation(...)
	...

	<span class="hljs-comment"># 0으로 스케일링</span>
	<span class="hljs-meta">>>> </span>endpoint.scale_to_zero()
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2'</span>, status=<span class="hljs-string">'scaledToZero'</span>, url=<span class="hljs-string">'https://jpj7k2q4j805b727.us-east-1.aws.endpoints.huggingface.cloud'</span>)
	<span class="hljs-comment"># 엔드포인트는 'running'은 아니지만 URL�을 가지고 있으며 첫 번째 호출 시 다시 시작됩니다.</span><!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="update-model-or-hardware-requirements" 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="#update-model-or-hardware-requirements"><span><svg class="" 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>모델 또는 하드웨어 요구 사항 업데이트</span></h3> <p data-svelte-h="svelte-e7c6jb">경우에 따라 새로운 엔드포인트를 생성하지 않고 추론 엔드포인트를 업데이트하고 싶을 수 있습니다. 호스팅된 모델이나 모델 실행에 필요한 하드웨어 요구 사항을 업데이트할 수 있습니다. 이렇게 하려면 <a href="/docs/huggingface_hub/pr_4113/ko/package_reference/inference_endpoints#huggingface_hub.InferenceEndpoint.update">update()</a>를 사용합니다:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># 타겟 모델 변경</span>
	<span class="hljs-meta">>>> </span>endpoint.update(repository=<span class="hljs-string">"gpt2-large"</span>)
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2-large'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)

	<span class="hljs-comment"># 복제본 갯수 업데이트</span>
	<span class="hljs-meta">>>> </span>endpoint.update(min_replica=<span class="hljs-number">2</span>, max_replica=<span class="hljs-number">6</span>)
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2-large'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)

	<span class="hljs-comment"># 더 큰 인스턴스로 업데이트</span>
	<span class="hljs-meta">>>> </span>endpoint.update(accelerator=<span class="hljs-string">"cpu"</span>, instance_size=<span class="hljs-string">"x4"</span>, instance_type=<span class="hljs-string">"intel-icl"</span>)
	InferenceEndpoint(name=<span class="hljs-string">'my-endpoint-name'</span>, namespace=<span class="hljs-string">'Wauplin'</span>, repository=<span class="hljs-string">'gpt2-large'</span>, status=<span class="hljs-string">'pending'</span>, url=<span class="hljs-literal">None</span>)<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="delete-the-endpoint" 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="#delete-the-endpoint"><span><svg class="" 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>엔드포인트 삭제</span></h3> <p data-svelte-h="svelte-16aox72">마지막으로 더 이상 추론 엔드포인트를 사용하지 않을 경우, <code>~InferenceEndpoint.delete()</code>를 호출하기만 하면 됩니다.</p> <blockquote class="warning" data-svelte-h="svelte-1i16qbl"><p>이것은 돌이킬 수 없는 작업이며, 구성, 로그 및 사용 메트릭을 포함한 엔드포인트를 완전히 제거합니다. 삭제된 추론 엔드포인트는 복원할 수 없습니다.</p></blockquote> <h2 class="relative group"><a id="엔드-투-엔드-예제-an-end-to-end-example" 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="#엔드-투-엔드-예제-an-end-to-end-example"><span><svg class="" 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>엔드 투 엔드 예제 an-end-to-end-example</span></h2> <p data-svelte-h="svelte-4lg1fr">추론 엔드포인트의 일반적인 사용 사례는 한 번에 여러 개의 작업을 처리하여 인프라 비용을 제한하는 것입니다. 이 가이드에서 본 것을 사용하여 이 프로세스를 자동화할 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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> asyncio
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> create_inference_endpoint

	<span class="hljs-comment"># 엔드포인트 시작 + 초기화될 때까지 대기</span>
	<span class="hljs-meta">>>> </span>endpoint = create_inference_endpoint(name=<span class="hljs-string">"batch-endpoint"</span>,...).wait()

	<span class="hljs-comment"># 추론 실행</span>
	<span class="hljs-meta">>>> </span>client = endpoint.client
	<span class="hljs-meta">>>> </span>results = [client.text_generation(...) <span class="hljs-keyword">for</span> job <span class="hljs-keyword">in</span> jobs]

	<span class="hljs-comment"># 비동기 추론 실행</span>
	<span class="hljs-meta">>>> </span>async_client = endpoint.async_client
	<span class="hljs-meta">>>> </span>results = asyncio.gather(*[async_client.text_generation(...) <span class="hljs-keyword">for</span> job <span class="hljs-keyword">in</span> jobs])

	<span class="hljs-comment"># 엔드포인트 중지</span>
	<span class="hljs-meta">>>> </span>endpoint.pause()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-5upsam">또는 추론 엔드포인트가 이미 존재하고 일시 중지된 경우:</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 class="" 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> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> asyncio
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> get_inference_endpoint

	<span class="hljs-comment"># 엔드포인트 가져오기 + 초기화될 때까지 대기</span>
	<span class="hljs-meta">>>> </span>endpoint = get_inference_endpoint(<span class="hljs-string">"batch-endpoint"</span>).resume().wait()

	<span class="hljs-comment"># 추론 실행</span>
	<span class="hljs-meta">>>> </span>async_client = endpoint.async_client
	<span class="hljs-meta">>>> </span>results = asyncio.gather(*[async_client.text_generation(...) <span class="hljs-keyword">for</span> job <span class="hljs-keyword">in</span> jobs])

	<span class="hljs-comment"># 엔드포인트 중지</span>
	<span class="hljs-meta">>>> </span>endpoint.pause()<!-- HTML_TAG_END --></pre></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/huggingface_hub/blob/main/docs/source/ko/guides/inference_endpoints.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 data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p>

	<script>
	{
	__sveltekit_1hcjqts = {
	assets: "/docs/huggingface_hub/pr_4113/ko",
	base: "/docs/huggingface_hub/pr_4113/ko",
	env: {}
	};

	const element = document.currentScript.parentElement;

	const data = [null,null];

	Promise.all([
	import("/docs/huggingface_hub/pr_4113/ko/_app/immutable/entry/start.e4d70577.js"),
	import("/docs/huggingface_hub/pr_4113/ko/_app/immutable/entry/app.3b0f3b68.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 8],
	data,
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 65.8 kB
Xet hash:: abb538a4d21ad6dfafc3c34671da31ecaa4110645ed65783e66d22f4d5168cb6

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.