Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"파이프라인 내부 동작 과정","local":"behind-the-pipeline","sections":[{"title":"토크나이저를 이용한 전처리","local":"preprocessing-with-a-tokenizer","sections":[],"depth":2},{"title":"모델 살펴보기","local":"going-through-the-model","sections":[{"title":"고차원 벡터란?","local":"a-high-dimensional-vector","sections":[],"depth":3},{"title":"모델 헤드: 숫자로 이해하기","local":"model-heads-making-sense-out-of-numbers","sections":[],"depth":3}],"depth":2},{"title":"출력 후처리","local":"postprocessing-the-output","sections":[],"depth":2}],"depth":1}"> | |
| <link href="/docs/course/pr_1069/ko/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/entry/start.667ba29b.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/scheduler.37c15a92.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/singletons.8de3d8e2.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/index.18351ede.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/paths.6047fff5.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/entry/app.65e35200.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/index.2bf4358c.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/nodes/0.9169bc8c.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/nodes/14.abf17ac7.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/Tip.363c041f.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/Youtube.1e50a667.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/CodeBlock.4e987730.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/CourseFloatingBanner.9ff4c771.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/DocNotebookDropdown.efc1fb7c.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/FrameworkSwitchCourse.8d4d4ab6.js"> | |
| <link rel="modulepreload" href="/docs/course/pr_1069/ko/_app/immutable/chunks/getInferenceSnippets.24b50994.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"파이프라인 내부 동작 과정","local":"behind-the-pipeline","sections":[{"title":"토크나이저를 이용한 전처리","local":"preprocessing-with-a-tokenizer","sections":[],"depth":2},{"title":"모델 살펴보기","local":"going-through-the-model","sections":[{"title":"고차원 벡터란?","local":"a-high-dimensional-vector","sections":[],"depth":3},{"title":"모델 헤드: 숫자로 이해하기","local":"model-heads-making-sense-out-of-numbers","sections":[],"depth":3}],"depth":2},{"title":"출력 후처리","local":"postprocessing-the-output","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="bg-white leading-none border border-gray-100 rounded-lg flex p-0.5 w-56 text-sm mb-4"><a class="flex justify-center flex-1 py-1.5 px-2.5 focus:outline-none !no-underline rounded-l bg-red-50 dark:bg-transparent text-red-600" href="?fw=pt"><svg class="mr-1.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><defs><clipPath id="a"><rect x="3.05" y="0.5" width="25.73" height="31" fill="none"></rect></clipPath></defs><g clip-path="url(#a)"><path d="M24.94,9.51a12.81,12.81,0,0,1,0,18.16,12.68,12.68,0,0,1-18,0,12.81,12.81,0,0,1,0-18.16l9-9V5l-.84.83-6,6a9.58,9.58,0,1,0,13.55,0ZM20.44,9a1.68,1.68,0,1,1,1.67-1.67A1.68,1.68,0,0,1,20.44,9Z" fill="#ee4c2c"></path></g></svg> Pytorch </a><a class="flex justify-center flex-1 py-1.5 px-2.5 focus:outline-none !no-underline rounded-r text-gray-500 filter grayscale" href="?fw=tf"><svg class="mr-1.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width="0.94em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 274"><path d="M145.726 42.065v42.07l72.861 42.07v-42.07l-72.86-42.07zM0 84.135v42.07l36.43 21.03V105.17L0 84.135zm109.291 21.035l-36.43 21.034v126.2l36.43 21.035v-84.135l36.435 21.035v-42.07l-36.435-21.034V105.17z" fill="#E55B2D"></path><path d="M145.726 42.065L36.43 105.17v42.065l72.861-42.065v42.065l36.435-21.03v-84.14zM255.022 63.1l-36.435 21.035v42.07l36.435-21.035V63.1zm-72.865 84.135l-36.43 21.035v42.07l36.43-21.036v-42.07zm-36.43 63.104l-36.436-21.035v84.135l36.435-21.035V210.34z" fill="#ED8E24"></path><path d="M145.726 0L0 84.135l36.43 21.035l109.296-63.105l72.861 42.07L255.022 63.1L145.726 0zm0 126.204l-36.435 21.03l36.435 21.036l36.43-21.035l-36.43-21.03z" fill="#F8BF3C"></path></svg> TensorFlow </a></div> <h1 class="relative group"><a id="behind-the-pipeline" 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="#behind-the-pipeline"><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>파이프라인 내부 동작 과정</span></h1> <div class="flex space-x-1 absolute z-10 right-0 top-0"><a href="https://discuss.huggingface.co/t/chapter-2-questions" target="_blank"><img alt="Ask a Question" class="!m-0" src="https://img.shields.io/badge/Ask%20a%20question-ffcb4c.svg?logo=data:image/svg+xml;base64,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"></a> <a href="https://colab.research.google.com/github/huggingface/notebooks/blob/master/course/en/chapter2/section2_pt.ipynb" target="_blank"><img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"></a> <a href="https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/master/course/en/chapter2/section2_pt.ipynb" target="_blank"><img alt="Open In Studio Lab" class="!m-0" src="https://studiolab.sagemaker.aws/studiolab.svg"></a></div> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400">PyTorch를 사용하는지, TensorFlow를 사용하는지에 따라 내용이 약간 달라지는 첫 번째 섹션입니다. 제목 상단 스위치를 이용해 선호하는 플랫폼을 선택하세요!</div> <iframe class="w-full xl:w-4/6 h-80" src="https://www.youtube-nocookie.com/embed/1pedAIvTWXk" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe> <p data-svelte-h="svelte-g9xmns">완전한 예제를 이용해, 아래의 <a href="/course/chapter1">제1단원</a> 코드를 수행했을 때 뒤에서 어떤 일이 일어나고 있는지 알아봅시다.</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> pipeline | |
| classifier = pipeline(<span class="hljs-string">"sentiment-analysis"</span>) | |
| classifier( | |
| [ | |
| <span class="hljs-string">"I've been waiting for a HuggingFace course my whole life."</span>, | |
| <span class="hljs-string">"I hate this so much!"</span>, | |
| ] | |
| )<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-wl1uu7">다음과 같은 출력이 나오게 됩니다.</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-string">'label'</span>: <span class="hljs-string">'POSITIVE'</span>, <span class="hljs-string">'score'</span>: <span class="hljs-number">0.9598047137260437</span>}, | |
| {<span class="hljs-string">'label'</span>: <span class="hljs-string">'NEGATIVE'</span>, <span class="hljs-string">'score'</span>: <span class="hljs-number">0.9994558095932007</span>}]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1iq5920"><a href="/course/chapter1">제1단원</a>에서 확인했듯이, 이 파이프라인 그룹은 세 단계(전처리, 입력을 모델에 넘겨주는 것, 후처리)를 함께 수행합니다.</p> <div class="flex justify-center" data-svelte-h="svelte-fbe70j"><img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface-course/documentation-images/resolve/main/en/chapter2/full_nlp_pipeline.svg" alt="The full NLP pipeline: tokenization of text, conversion to IDs, and inference through the Transformer model and the model head."> <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface-course/documentation-images/resolve/main/en/chapter2/full_nlp_pipeline-dark.svg" alt="The full NLP pipeline: tokenization of text, conversion to IDs, and inference through the Transformer model and the model head."></div> <p data-svelte-h="svelte-vi0ems">각 단계에 대해 빠르게 살펴보겠습니다.</p> <h2 class="relative group"><a id="preprocessing-with-a-tokenizer" 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="#preprocessing-with-a-tokenizer"><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>토크나이저를 이용한 전처리</span></h2> <p data-svelte-h="svelte-e4ngzp">다른 신경망처럼 Transformer 모델도 원시 텍스트를 바로 처리할 수 없기 때문에 파이프라인의 첫 번째 단계는 텍스트 입력을 모델이 이해할 수 있는 숫자로 변환하는 것입니다. 이 과정을 위해 다음 기능들을 수행하는 <em>토크나이저</em>를 사용합니다.</p> <ul data-svelte-h="svelte-cxgkxr"><li>입력을 <em>토큰</em>이라고 부르는 단어나 하위 단어, 또는 심볼(예-구두점)로 분할</li> <li>각 토큰을 하나의 정수에 매핑</li> <li>모델에 유용할 수 있는 부가적인 입력 추가</li></ul> <p data-svelte-h="svelte-w17t67">이 모든 전처리 과정은 모델이 사전학습될 때와 완전히 동일한 방식으로 진행되어야 하기 때문에 <a href="https://huggingface.co/models" rel="nofollow">Model Hub</a>에서 정보를 다운로드 해야합니다. 전처리를 위해 <code>AutoTokenizer</code> 클래스와 AutoTokenizer의 <code>from_pretrained()</code> 메서드를 사용합니다. 모델의 체크포인트 이름을 사용하여 모델의 토크나이저와 연관된 데이터를 자동으로 가져와 저장합니다. (따라서 아래 코드를 처음 실행할 때만 다운로드됩니다.)</p> <p data-svelte-h="svelte-79gaho"><code>sentiment-analysis</code> 파이프라인의 기본 체크포인트는 <code>distilbert-base-uncased-finetuned-sst-2-english</code>(모델 카드 확인은 <a href="https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english" rel="nofollow">여기</a>서 가능)이므로 아래 코드를 실행합니다.</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| checkpoint = <span class="hljs-string">"distilbert-base-uncased-finetuned-sst-2-english"</span> | |
| tokenizer = AutoTokenizer.from_pretrained(checkpoint)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1rskcen">토크나이저가 있다면 문장을 토크나이저에 입력하여 우리의 모델에 전달할 준비가 된 딕셔너리를 출력으로 얻게 됩니다! 남은 것은 입력 ID 목록을 tensor로 변환하는 것입니다.</p> <p data-svelte-h="svelte-1yfq89p">여러분은 어떤 ML 프레임워크-PyTorch나 TensorFlow, 또는 몇몇 모델을 위한 Flax-가 백엔드로 사용되는지 걱정하지 않고 🤗 Transformers를 사용할 수 있습니다. 하지만 Transformer 모델은 <em>tensor</em>만을 입력으로 받습니다. 만약 tensor에 대해 처음 들어봤다면, NumPy 배열을 생각하면 됩니다. NumPy 배열은 스칼라(0D), 벡터(1D), 행렬(2D) 또는 더 많은 차원을 가질 수 있습니다. 이것은 사실상 텐서입니다. 다른 ML 프레임워크의 텐서도 유사하게 동작하며, NumPy 배열처럼 인스턴스화가 쉽습니다.</p> <p data-svelte-h="svelte-yp4x9m">얻고자 하는 tensor의 타입(PyTorch, TensorFlow, 일반 NumPy)을 지정하기 위해, <code>return_tensors</code> 전달인자를 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 -->raw_inputs = [ | |
| <span class="hljs-string">"I've been waiting for a HuggingFace course my whole life."</span>, | |
| <span class="hljs-string">"I hate this so much!"</span>, | |
| ] | |
| inputs = tokenizer(raw_inputs, padding=<span class="hljs-literal">True</span>, truncation=<span class="hljs-literal">True</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| <span class="hljs-built_in">print</span>(inputs)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-12luxqf">padding과 truncation에 대해 벌써 걱정하지 마세요. 나중에 설명하도록 하겠습니다. 여기서 기억해야 할 중요한 점은 하나의 문장 또는 여러 개의 문장 리스트를 토크나이저 함수로 전달할 수 있을 뿐만 아니라 얻고 싶은 텐서 유형까지 지정할 수 있다는 것입니다. (텐서 유형이 지정되지 않으면 이중 리스트를 결과로 얻게 됩니다)</p> <p data-svelte-h="svelte-224u1o">PyTorch tensor로 지정했을 때의 결과입니다.</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-string">'input_ids'</span>: tensor([ | |
| [ <span class="hljs-number">101</span>, <span class="hljs-number">1045</span>, <span class="hljs-number">1005</span>, <span class="hljs-number">2310</span>, <span class="hljs-number">2042</span>, <span class="hljs-number">3403</span>, <span class="hljs-number">2005</span>, <span class="hljs-number">1037</span>, <span class="hljs-number">17662</span>, <span class="hljs-number">12172</span>, <span class="hljs-number">2607</span>, <span class="hljs-number">2026</span>, <span class="hljs-number">2878</span>, <span class="hljs-number">2166</span>, <span class="hljs-number">1012</span>, <span class="hljs-number">102</span>], | |
| [ <span class="hljs-number">101</span>, <span class="hljs-number">1045</span>, <span class="hljs-number">5223</span>, <span class="hljs-number">2023</span>, <span class="hljs-number">2061</span>, <span class="hljs-number">2172</span>, <span class="hljs-number">999</span>, <span class="hljs-number">102</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>] | |
| ]), | |
| <span class="hljs-string">'attention_mask'</span>: tensor([ | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>] | |
| ]) | |
| }<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-4zxdhb">출력은 <code>input_ids</code>와 <code>attention_mask</code> 두 개의 키를 갖는 딕셔너리입니다. <code>input_ids</code>는 각 문장 내 토큰의 고유 식별자인 정수로 이루어진 2개의 행(한 행이 하나의 문장)을 가지고 있습니다. <code>attention_mask</code>는 이 챕터의 뒤쪽에서 설명할 것입니다.</p> <h2 class="relative group"><a id="going-through-the-model" 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="#going-through-the-model"><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 살펴보기</span></h2> <p data-svelte-h="svelte-9oefvu">토크나이저를 다운받은 방식과 동일한 방식으로 사전학습된 모델을 다운받을 수 있습니다. 🤗 Transformers는 <code>from_pretrained</code> 메서드를 가진 <code>AutoModel</code> 클래스를 제공합니다.</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoModel | |
| checkpoint = <span class="hljs-string">"distilbert-base-uncased-finetuned-sst-2-english"</span> | |
| model = AutoModel.from_pretrained(checkpoint)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1kq8h6f">이 코드 스니펫에서 이전 파이프라인에서 사용된 것과 동일한 체크포인트를 다운로드(실제로는 이미 저장되어 있어야 합니다.)하고 그것으로 모델을 인스턴스화했습니다.</p> <p data-svelte-h="svelte-1m8n2tn">해당 아키텍처는 기본 Transformer 모듈만 포함하고 있습니다. 입력이 주어지면 <em>features</em>라고도 불리는 <em>hidden states</em>를 출력합니다. 각 모델의 입력에 대해 <strong>Transformer 모델에 의해 수행된 입력의 문맥적 이해</strong>로 표현할 수 있는 고차원 벡터를 가져옵니다.</p> <p data-svelte-h="svelte-xwzhyw">이 내용이 이해되지 않더라도 걱정하지 마세요. 뒤에서 설명할 것입니다.</p> <p data-svelte-h="svelte-1j1df6u">이러한 hidden states는 그 자체로 유용할 수 있지만 일반적으로 <em>head</em>라고 알려진 모델의 다른 부분에 입력으로 들어갑니다. <a href="/course/chapter1">제1단원</a>에서, 동일한 구조로 다른 태스크를 수행할 수 있었는데 이 태스크들은 서로 다른 헤드와 연관되어 있습니다.</p> <h3 class="relative group"><a id="a-high-dimensional-vector" 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="#a-high-dimensional-vector"><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>고차원 벡터란?</span></h3> <p data-svelte-h="svelte-118vqvh">Transformer 모듈에 의한 출력 벡터는 일반적으로 크며 보통 3개의 차원을 가집니다.</p> <ul data-svelte-h="svelte-1yer64p"><li><strong>Batch size</strong>: 한 번에 처리되는 시퀀스의 수 (예제에서는 2)</li> <li><strong>Sequence length</strong>: 시퀀스의 숫자 표현 길이 (예제에서는 16)</li> <li><strong>Hidden size</strong>: 각 모델 입력 벡터 차원</li></ul> <p data-svelte-h="svelte-1ldcz1p">위에서 마지막 값으로 인해 “고차원”이라고 불립니다. hidden size는 매우 클 수 있습니다 (작은 모델은 768이 일반적이며 큰 모델은 3072나 그 이상의 값이 될 수 있습니다)</p> <p data-svelte-h="svelte-foiu8f">전처리 과정을 거친 입력을 모델로 넘기면 아래 결과를 확인할 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 -->outputs = model(**inputs) | |
| <span class="hljs-built_in">print</span>(outputs.last_hidden_state.shape)<!-- HTML_TAG_END --></pre></div> <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 -->torch.Size([<span class="hljs-number">2</span>, <span class="hljs-number">16</span>, <span class="hljs-number">768</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-it37lc">🤗 Transformers 모델의 출력은 <code>namedtuple</code> 또는 딕셔너리 형태입니다. 속성이나 키(<code>outputs["last_hidden_state"]</code>)를 이용해 요소에 접근할 수 있고 찾고자 하는 것의 정확한 위치를 안다면 인덱스(<code>outputs[0]</code>)도 사용할 수 있습니다.</p> <h3 class="relative group"><a id="model-heads-making-sense-out-of-numbers" 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="#model-heads-making-sense-out-of-numbers"><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 헤드: 숫자로 이해하기</span></h3> <p data-svelte-h="svelte-13d2s5r">모델 헤드는 hidden state의 고차원 벡터를 입력으로 받아 다른 차원으로 투영합니다. 모델 헤드는 보통 하나 이상의 선형 레이어로 이루어져 있습니다.</p> <div class="flex justify-center" data-svelte-h="svelte-8mo6lt"><img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface-course/documentation-images/resolve/main/en/chapter2/transformer_and_head.svg" alt="A Transformer network alongside its head."> <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface-course/documentation-images/resolve/main/en/chapter2/transformer_and_head-dark.svg" alt="A Transformer network alongside its head."></div> <p data-svelte-h="svelte-k1g54v">Transformer 모델의 출력은 처리할 모델 헤드로 바로 전달됩니다.</p> <p data-svelte-h="svelte-did2wt">이 다이어그램에서, 모델은 모델의 임베딩 레이어와 후속 레이어로 표현됩니다. 임베딩 레이어는 토큰화된 각각의 입력 ID를 연관된 토큰을 나타내는 벡터로 변환합니다. 후속 레이어는 문장의 최종 표현을 만들기 위해 어텐션 메커니즘을 이용해 이 벡터들을 처리합니다.</p> <p data-svelte-h="svelte-1bvhwer">🤗 Transformer에는 다양한 아키텍처가 있으며, 각각의 아키텍처는 특정 작업을 처리하도록 설계되었습니다. 아래는 일부 아키텍처입니다.</p> <ul data-svelte-h="svelte-nf90ig"><li><code>*Model</code> (retrieve the hidden states)</li> <li><code>*ForCausalLM</code></li> <li><code>*ForMaskedLM</code></li> <li><code>*ForMultipleChoice</code></li> <li><code>*ForQuestionAnswering</code></li> <li><code>*ForSequenceClassification</code></li> <li><code>*ForTokenClassification</code></li> <li>그 외 🤗</li></ul> <p data-svelte-h="svelte-vxxqdi">이 예제를 위해서는 문장을 긍정 또는 부정으로 분류할 수 있게 하는 시퀀스 분류 헤드를 가진 모델이 필요합니다. 따라서 <code>AutoModel</code> 클래스가 아닌 <code>AutoModelForSequenceClassification</code>을 사용할 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-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoModelForSequenceClassification | |
| checkpoint = <span class="hljs-string">"distilbert-base-uncased-finetuned-sst-2-english"</span> | |
| model = AutoModelForSequenceClassification.from_pretrained(checkpoint) | |
| outputs = model(**inputs)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-19v4iwk">출력 형태를 보면 차원이 훨씬 적은 것을 알 수 있습니다. 모델 헤드는 이전에 봤던 고차원 벡터를 입력으로 받아 2개의 값(레이블 당 하나)으로 이루어진 벡터를 출력합니다.</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-built_in">print</span>(outputs.logits.shape)<!-- HTML_TAG_END --></pre></div> <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 -->torch.Size([<span class="hljs-number">2</span>, <span class="hljs-number">2</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-vcufcz">우리는 2개의 문장과 2개의 레이블만 있기 때문에 모델로부터 얻은 출력 형태는 2 x 2입니다.</p> <h2 class="relative group"><a id="postprocessing-the-output" 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="#postprocessing-the-output"><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>출력 후처리</span></h2> <p data-svelte-h="svelte-1fm5gwp">모델의 출력 값이 그 자체로 의미있는 것은 아닙니다. 한 번 보도록 합시다.</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-built_in">print</span>(outputs.logits)<!-- HTML_TAG_END --></pre></div> <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 -->tensor([[-<span class="hljs-number">1.5607</span>, <span class="hljs-number">1.6123</span>], | |
| [ <span class="hljs-number">4.1692</span>, -<span class="hljs-number">3.3464</span>]], grad_fn=<AddmmBackward>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-a9kzjr">모델은 첫 번째 문장에 대해 <code>[-1.5607, 1.6123]</code>으로 예측했고 두 번째 문장에 대해 <code>[ 4.1692, -3.3464]</code>으로 예측했습니다. 이 값들은 확률이 아니라 모델의 마지막 층에 의해 출력된 정규화되지 않은 점수인 <em>logits</em>입니다. 확률로 변환되기 위해 logits은 <a href="https://en.wikipedia.org/wiki/Softmax_function" rel="nofollow">SoftMax</a> 층을 거쳐야 합니다. 학습을 위한 손실 함수가 일반적으로 SoftMax와 같은 마지막 활성함수와 교차 엔트로피와 같은 실제 손실 함수를 모두 사용하기 때문에 모든 🤗 Transformers 모델의 출력은 logit입니다.</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-keyword">import</span> torch | |
| predictions = torch.nn.functional.softmax(outputs.logits, dim=-<span class="hljs-number">1</span>) | |
| <span class="hljs-built_in">print</span>(predictions)<!-- HTML_TAG_END --></pre></div> <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 -->tensor([[<span class="hljs-number">4.0195e-02</span>, <span class="hljs-number">9.5980e-01</span>], | |
| [<span class="hljs-number">9.9946e-01</span>, <span class="hljs-number">5.4418e-04</span>]], grad_fn=<SoftmaxBackward>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-11qcsxk">모델은 첫 번째 문장에 대해 <code>[0.0402, 0.9598]</code>로 예측했고 두 번째 모델에 대해 <code>[0.9995, 0.0005]</code>로 예측했습니다. 이 값들은 확실하게 확률값입니다.</p> <p data-svelte-h="svelte-16uv0pg">각 위치에 해당하는 레이블을 얻기 위해, 모델 config의 <code>id2label</code> 속성을 살펴봅시다. Config에 대한 더 많은 내용은 다음 섹션에서 진행됩니다.</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 -->model.config.id2label<!-- HTML_TAG_END --></pre></div> <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-number">0</span>: <span class="hljs-string">'NEGATIVE'</span>, <span class="hljs-number">1</span>: <span class="hljs-string">'POSITIVE'</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1qu5r5q">모델의 예측 결과를 아래처럼 결론 지을 수 있습니다.</p> <ul data-svelte-h="svelte-a79kkl"><li>첫 번째 문장: NEGATIVE: 0.0402, POSITIVE: 0.9598</li> <li>두 번째 문장: NEGATIVE: 0.9995, POSITIVE: 0.0005</li></ul> <p data-svelte-h="svelte-ptsq7t">파이프라인 세 단계-토크나이저를 이용한 전처리, 모델에 입력 넣어주기, 후처리-를 성공적으로 재현했습니다! 이제 각 단계별로 좀 더 깊게 알아보는 시간을 가져봅시다.</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-1t16ayp">✏️ <strong>직접 해보세요!</strong> 2개 이상의 문장을 골라 <code>sentiment-analysis</code> 파이프라인을 적용해보세요. 이 챕터에서 본 내용을 그대로 수행해보고 같은 결과가 나오는지 확인해보세요!</p></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/course/blob/main/chapters/ko/chapter2/2.mdx" target="_blank"><span data-svelte-h="svelte-1kd6by1"><</span> <span data-svelte-h="svelte-x0xyl0">></span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p> | |
| <script> | |
| { | |
| __sveltekit_5pcgyl = { | |
| assets: "/docs/course/pr_1069/ko", | |
| base: "/docs/course/pr_1069/ko", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/course/pr_1069/ko/_app/immutable/entry/start.667ba29b.js"), | |
| import("/docs/course/pr_1069/ko/_app/immutable/entry/app.65e35200.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 14], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 56.1 kB
- Xet hash:
- 70b69d392e57f9c6bd5ebce74c2b24910e87efd5ffb0ddec3ccedd0e6fafddf4
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.