Buckets:

hf-doc-build
/

doc-dev

Files

xet

hf-doc-build/doc-dev / smol-course /pr_296 /ja /unit4 /3.html

HuggingFaceDocBuilder

about 2 months ago

download

raw

31.9 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"カスタムドメイン評価","local":"カスタムドメイン評価","sections":[{"title":"評価戦略の設計","local":"評価戦略の設計","sections":[],"depth":2},{"title":"LightEvalを使用した実装","local":"lightevalを使用した実装","sections":[],"depth":2},{"title":"カスタムメトリック","local":"カスタムメトリック","sections":[],"depth":2},{"title":"データセットの作成","local":"データセットの作成","sections":[],"depth":2},{"title":"ベストプラクティス","local":"ベストプラクティス","sections":[],"depth":2},{"title":"参考文献","local":"参考文献","sections":[],"depth":2}],"depth":1}">
	<link href="/docs/smol-course/pr_296/ja/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/entry/start.afa23ac6.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/scheduler.f38f3a0e.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/singletons.6933622c.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/paths.2a0c5cb3.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/entry/app.27746c11.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/preload-helper.c5d58819.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/index.a03143fc.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/nodes/0.117e80e6.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/each.e59479a4.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/nodes/15.0fbd4842.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/CopyLLMTxtMenu.8710b29c.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.4b36369c.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/CodeBlock.a6034da5.js">
	<link rel="modulepreload" href="/docs/smol-course/pr_296/ja/_app/immutable/chunks/CourseFloatingBanner.9d00122a.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"カスタムドメイン評価","local":"カスタムドメイン評価","sections":[{"title":"評価戦略の設計","local":"評価戦略の設計","sections":[],"depth":2},{"title":"LightEvalを使用した実装","local":"lightevalを使用した実装","sections":[],"depth":2},{"title":"カスタムメトリック","local":"カスタムメトリック","sections":[],"depth":2},{"title":"データセットの作成","local":"データセットの作成","sections":[],"depth":2},{"title":"ベストプラクティス","local":"ベストプラクティス","sections":[],"depth":2},{"title":"参考文献","local":"参考文献","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="カスタムドメイン評価" 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="#カスタムドメイン評価"><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>カスタムドメイン評価</span></h1> <div class="flex space-x-1 absolute z-10 right-0 top-0" style=""><a href="https://discuss.huggingface.co/t/chapter-10-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/smol-course/blob/main/notebooks/ja/4_evaluation/lighteval_evaluate_and_analyse_your_LLM.ipynb" target="_blank"><img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"></a> </div>

	標準ベンチマークは貴重な洞察を提供しますが、多くのアプリケーションでは特定のドメインやユースケースに合わせた評価アプローチが必要です。このガイドでは、ターゲットドメインでモデルの性能を正確に評価するためのカスタム評価パイプラインを作成する方法を説明します。
	<h2 class="relative group"><a id="評価戦略の設計" 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="#評価戦略の設計"><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>評価戦略の設計</span></h2> <p data-svelte-h="svelte-1v8ffdy">成功するカスタム評価戦略は、明確な目標から始まります。ドメインでモデルが示すべき具体的な能力を考慮してください。これには、技術的な知識、推論パターン、ドメイン固有の形式が含まれるかもしれません。これらの要件を慎重に文書化してください。これがタスク設計とメトリック選択の両方を導きます。</p> <p data-svelte-h="svelte-1pfrbs5">評価は、標準的なユースケースとエッジケースの両方をテストする必要があります。例えば、医療ドメインでは、一般的な診断シナリオと稀な状態の両方を評価するかもしれません。金融アプリケーションでは、通常の取引と複数の通貨や特別な条件を含む複雑なエッジケースの両方をテストするかもしれません。</p> <h2 class="relative group"><a id="lightevalを使用した実装" 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="#lightevalを使用した実装"><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>LightEvalを使用した実装</span></h2> <p data-svelte-h="svelte-1ojfrcc">LightEvalは、カスタム評価を実装するための柔軟なフレームワークを提供します。カスタムタスクを作成する方法は次のとおりです：</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> lighteval.tasks <span class="hljs-keyword">import</span> Task, Doc
	<span class="hljs-keyword">from</span> lighteval.metrics <span class="hljs-keyword">import</span> SampleLevelMetric, MetricCategory, MetricUseCase

	<span class="hljs-keyword">class</span> <span class="hljs-title class_">CustomEvalTask</span>(<span class="hljs-title class_ inherited__">Task</span>):
	<span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self</span>):
	<span class="hljs-built_in">super</span>().__init__(
	name=<span class="hljs-string">"custom_task"</span>,
	version=<span class="hljs-string">"0.0.1"</span>,
	metrics=[<span class="hljs-string">"accuracy"</span>, <span class="hljs-string">"f1"</span>], <span class="hljs-comment"># 選択したメトリック</span>
	description=<span class="hljs-string">"カスタム評価タスクの説明"</span>
	)

	<span class="hljs-keyword">def</span> <span class="hljs-title function_">get_prompt</span>(<span class="hljs-params">self, sample</span>):
	<span class="hljs-comment"># 入力をプロンプトにフォーマット</span>
	<span class="hljs-keyword">return</span> <span class="hljs-string">f"質問: <span class="hljs-subst">{sample[<span class="hljs-string">'question'</span>
	]}</span>\n回答:"</span>

	<span class="hljs-keyword">def</span> <span class="hljs-title function_">process_response</span>(<span class="hljs-params">self, response, ref</span>):
	<span class="hljs-comment"># モデルの出力を処理し、参照と比較</span>
	<span class="hljs-keyword">return</span> response.strip() == ref.strip()<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="カスタムメトリック" 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="#カスタムメトリック"><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>カスタムメトリック</span></h2> <p data-svelte-h="svelte-cmxtrw">ドメイン固有のタスクには、専門的なメトリックが必要なことがよくあります。LightEvalは、ドメインに関連する性能の側面を捉えるカスタムメトリックを作成するための柔軟なフレームワークを提供します：</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> aenum <span class="hljs-keyword">import</span> extend_enum
	<span class="hljs-keyword">from</span> lighteval.metrics <span class="hljs-keyword">import</span> Metrics, SampleLevelMetric, SampleLevelMetricGrouping
	<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np

	<span class="hljs-comment"># サンプルレベルのメトリック関数を定義</span>
	<span class="hljs-keyword">def</span> <span class="hljs-title function_">custom_metric</span>(<span class="hljs-params">predictions: <span class="hljs-built_in">list</span>[<span class="hljs-built_in">str</span>], formatted_doc: Doc, **kwargs</span>) -> <span class="hljs-built_in">dict</span>:
	<span class="hljs-string">"""サンプルごとに複数のスコアを返す例のメトリック"""</span>
	response = predictions[<span class="hljs-number">0</span>]
	<span class="hljs-keyword">return</span> {
	<span class="hljs-string">"accuracy"</span>: response == formatted_doc.choices[formatted_doc.gold_index],
	<span class="hljs-string">"length_match"</span>: <span class="hljs-built_in">len</span>(response) == <span class="hljs-built_in">len</span>(formatted_doc.reference)
	}

	<span class="hljs-comment"># サンプルごとに複数の値を返すメトリックを作成</span>
	custom_metric_group = SampleLevelMetricGrouping(
	metric_name=[<span class="hljs-string">"accuracy"</span>, <span class="hljs-string">"length_match"</span>], <span class="hljs-comment"># サブメトリックの名前</span>
	higher_is_better={ <span class="hljs-comment"># 各メトリックで高い値が良いかどうか</span>
	<span class="hljs-string">"accuracy"</span>: <span class="hljs-literal">True</span>,
	<span class="hljs-string">"length_match"</span>: <span class="hljs-literal">True</span>
	},
	category=MetricCategory.CUSTOM,
	use_case=MetricUseCase.SCORING,
	sample_level_fn=custom_metric,
	corpus_level_fn={ <span class="hljs-comment"># 各メトリックを集計する方法</span>
	<span class="hljs-string">"accuracy"</span>: np.mean,
	<span class="hljs-string">"length_match"</span>: np.mean
	}
	)

	<span class="hljs-comment"># LightEvalにメトリックを登録</span>
	extend_enum(Metrics, <span class="hljs-string">"custom_metric_name"</span>, custom_metric_group)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-tni7mx">サンプルごとに1つのメトリック値のみが必要な場合の簡単なケース：</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">def</span> <span class="hljs-title function_">simple_metric</span>(<span class="hljs-params">predictions: <span class="hljs-built_in">list</span>[<span class="hljs-built_in">str</span>], formatted_doc: Doc, **kwargs</span>) -> <span class="hljs-built_in">bool</span>:
	<span class="hljs-string">"""サンプルごとに単一のスコアを返す例のメトリック"""</span>
	response = predictions[<span class="hljs-number">0</span>]
	<span class="hljs-keyword">return</span> response == formatted_doc.choices[formatted_doc.gold_index]

	simple_metric_obj = SampleLevelMetric(
	metric_name=<span class="hljs-string">"simple_accuracy"</span>,
	higher_is_better=<span class="hljs-literal">True</span>,
	category=MetricCategory.CUSTOM,
	use_case=MetricUseCase.SCORING,
	sample_level_fn=simple_metric,
	corpus_level_fn=np.mean <span class="hljs-comment"># サンプル全体で集計する方法</span>
	)

	extend_enum(Metrics, <span class="hljs-string">"simple_metric"</span>, simple_metric_obj)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1yjlxdn">カスタムメトリックを評価タスクで参照することで、評価タスクで自動的に計算され、指定された関数に従って集計されます。</p> <p data-svelte-h="svelte-pvy7ub">より複雑なメトリックの場合、次のことを検討してください：</p> <ul data-svelte-h="svelte-zvms9u"><li>フォーマットされたドキュメントのメタデータを使用してスコアを重み付けまたは調整</li> <li>コーパスレベルの統計のためのカスタム集計関数の実装</li> <li>メトリック入力の検証チェックの追加</li> <li>エッジケースと期待される動作の文書化</li></ul> <p data-svelte-h="svelte-1plvrw">これらの概念を実装する完全な例については、<a href="./project/README">ドメイン評価プロジェクト</a>を参照してください。</p> <h2 class="relative group"><a id="データセットの作成" 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="#データセットの作成"><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>データセットの作成</span></h2> <p data-svelte-h="svelte-1o4jyyw">高品質の評価には、慎重にキュレーションされたデータセットが必要です。データセット作成のアプローチを次に示します：</p> <ol data-svelte-h="svelte-17okzmc"><li><p>専門家のアノテーション：ドメインの専門家と協力して評価例を作成および検証します。<a href="https://github.com/argilla-io/argilla" rel="nofollow">Argilla</a>のようなツールを使用すると、このプロセスがより効率的になります。</p></li> <li><p>実世界のデータ：実際の使用データを収集し、匿名化して、実際の展開シナリオを反映します。</p></li> <li><p>合成生成：LLMを使用して初期例を生成し、専門家がそれを検証および洗練します。</p></li></ol> <h2 class="relative group"><a id="ベストプラクティス" 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="#ベストプラクティス"><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>ベストプラクティス</span></h2> <ul data-svelte-h="svelte-1lax18e"><li>評価方法論を徹底的に文書化し、仮定や制限を含める</li> <li>ドメインのさまざまな側面をカバーする多様なテストケースを含める</li> <li>自動メトリックと人間の評価の両方を適用する</li> <li>評価データセットとコードをバージョン管理する</li> <li>新しいエッジケースや要件を発見するたびに評価スイートを定期的に更新する</li></ul> <h2 class="relative group"><a id="参考文献" 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="#参考文献"><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>参考文献</span></h2> <ul data-svelte-h="svelte-vl9na5"><li><a href="https://github.com/huggingface/lighteval/wiki/Adding-a-Custom-Task" rel="nofollow">LightEvalカスタムタスクガイド</a></li> <li><a href="https://github.com/huggingface/lighteval/wiki/Adding-a-New-Metric" rel="nofollow">LightEvalカスタムメトリック</a></li> <li>データセットアノテーションのための<a href="https://docs.argilla.io" rel="nofollow">Argillaドキュメント</a></li> <li>一般的な評価原則のための<a href="https://github.com/huggingface/evaluation-guidebook" rel="nofollow">評価ガイドブック</a></li></ul> <h1 class="relative group"><a id="次のステップ" 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="#次のステップ"><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>次のステップ</span></h1> <p data-svelte-h="svelte-6p1xhz">⏩ これらの概念を実装する完全な例については、<a href="./project/README">ドメイン評価プロジェクト</a>を参照してください。</p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/smol-course/blob/main/units/ja/unit4/3.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_1lq7jru = {
	assets: "/docs/smol-course/pr_296/ja",
	base: "/docs/smol-course/pr_296/ja",
	env: {}
	};

	const element = document.currentScript.parentElement;

	const data = [null,null];

	Promise.all([
	import("/docs/smol-course/pr_296/ja/_app/immutable/entry/start.afa23ac6.js"),
	import("/docs/smol-course/pr_296/ja/_app/immutable/entry/app.27746c11.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 15],
	data,
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 31.9 kB
Xet hash:: d190146f4ed4d816e5077db8f8958ee7e9abab9df9eecf586fa86f3267a54efe

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