Buckets:

hf-doc-build
/

doc

Files

xet

hf-doc-build/doc / transformers /main /zh /run_scripts.html

HuggingFaceDocBuilder

1 day ago

download

raw

52.8 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"使用脚本进行训练","local":"使用脚本进行训练","sections":[{"title":"设置","local":"设置","sections":[],"depth":2},{"title":"运行脚本","local":"运行脚本","sections":[],"depth":2},{"title":"分布式训练和混合精度","local":"分布式训练和混合精度","sections":[],"depth":2},{"title":"在TPU上运行脚本","local":"在tpu上运行脚本","sections":[],"depth":2},{"title":"基于🤗 Accelerate运行脚本","local":"基于-accelerate运行脚本","sections":[],"depth":2},{"title":"使用自定义数据集","local":"使用自定义数据集","sections":[],"depth":2},{"title":"测试脚本","local":"测试脚本","sections":[],"depth":2},{"title":"从checkpoint恢复训练","local":"从checkpoint恢复训练","sections":[],"depth":2},{"title":"分享模型","local":"分享模型","sections":[],"depth":2}],"depth":1}">
	<link href="/docs/transformers/main/zh/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/entry/start.7a5389c5.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/scheduler.70fef3e7.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/singletons.de8a9de6.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/index.048d9726.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/paths.a6fe27ce.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/entry/app.ac7d4b06.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/preload-helper.77b1b155.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/index.747f74de.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/nodes/0.3408eb2f.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/each.e59479a4.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/nodes/61.49afcfdb.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/CopyLLMTxtMenu.25c10d36.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.04bd7ab0.js">
	<link rel="modulepreload" href="/docs/transformers/main/zh/_app/immutable/chunks/CodeBlock.15c43204.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"使用脚本进行训练","local":"使用脚本进行训练","sections":[{"title":"设置","local":"设置","sections":[],"depth":2},{"title":"运行脚本","local":"运行脚本","sections":[],"depth":2},{"title":"分布式训练和混合精度","local":"分布式训练和混合精度","sections":[],"depth":2},{"title":"在TPU上运行脚本","local":"在tpu上运行脚本","sections":[],"depth":2},{"title":"基于🤗 Accelerate运行脚本","local":"基于-accelerate运行脚本","sections":[],"depth":2},{"title":"使用自定义数据集","local":"使用自定义数据集","sections":[],"depth":2},{"title":"测试脚本","local":"测试脚本","sections":[],"depth":2},{"title":"从checkpoint恢复训练","local":"从checkpoint恢复训练","sections":[],"depth":2},{"title":"分享模型","local":"分享模型","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="使用脚本进行训练" 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="#使用脚本进行训练"><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>使用脚本进行训练</span></h1> <p data-svelte-h="svelte-pxoox7">除了 🤗 Transformers <a href="./notebooks">notebooks</a>，还有示例脚本演示了如何使用<a href="https://github.com/huggingface/transformers/tree/main/examples/pytorch" rel="nofollow">PyTorch</a>训练模型以解决特定任务。</p> <p data-svelte-h="svelte-1ziige">您还可以在这些示例中找到我们在<a href="https://github.com/huggingface/transformers-research-projects/" rel="nofollow">研究项目</a>和<a href="https://github.com/huggingface/transformers/tree/main/examples/legacy" rel="nofollow">遗留示例</a>中使用过的脚本，这些脚本主要是由社区贡献的。这些脚本已不再被积极维护，需要使用特定版本的🤗 Transformers，可能与库的最新版本不兼容。</p> <p data-svelte-h="svelte-y5rq4n">示例脚本可能无法在初始配置下直接解决每个问题，您可能需要根据要解决的问题调整脚本。为了帮助您，大多数脚本都完全暴露了数据预处理的方式，允许您根据需要对其进行编辑。</p> <p data-svelte-h="svelte-zuh3zf">如果您想在示例脚本中实现任何功能，请在<a href="https://discuss.huggingface.co/" rel="nofollow">论坛</a>或<a href="https://github.com/huggingface/transformers/issues" rel="nofollow">issue</a>上讨论，然后再提交Pull Request。虽然我们欢迎修复错误，但不太可能合并添加更多功能的Pull Request，因为这会降低可读性。</p> <p data-svelte-h="svelte-1fu3vi8">本指南将向您展示如何在<a href="https://github.com/huggingface/transformers/tree/main/examples/pytorch/summarization" rel="nofollow">PyTorch</a>中运行示例摘要训练脚本。</p> <h2 class="relative group"><a id="设置" 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="#设置"><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>设置</span></h2> <p data-svelte-h="svelte-2nj8cj">要成功运行示例脚本的最新版本，您必须在新虚拟环境中<strong>从源代码安装 🤗 Transformers</strong>：</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="language-bash "><!-- HTML_TAG_START -->git <span class="hljs-built_in">clone</span> https://github.com/huggingface/transformers
	<span class="hljs-built_in">cd</span> transformers
	pip install .<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-8nj16n">对于旧版本的示例脚本，请点击下面的切换按钮：</p> <details data-svelte-h="svelte-1ubwk1n"><summary>老版本🤗 Transformers示例</summary> <ul><li><a href="https://github.com/huggingface/transformers/tree/v4.5.1/examples">v4.5.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v4.4.2/examples">v4.4.2</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v4.3.3/examples">v4.3.3</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v4.2.2/examples">v4.2.2</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v4.1.1/examples">v4.1.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v4.0.1/examples">v4.0.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v3.5.1/examples">v3.5.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v3.4.0/examples">v3.4.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v3.3.1/examples">v3.3.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v3.2.0/examples">v3.2.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v3.1.0/examples">v3.1.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v3.0.2/examples">v3.0.2</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.11.0/examples">v2.11.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.10.0/examples">v2.10.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.9.1/examples">v2.9.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.8.0/examples">v2.8.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.7.0/examples">v2.7.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.6.0/examples">v2.6.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.5.1/examples">v2.5.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.4.0/examples">v2.4.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.3.0/examples">v2.3.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.2.0/examples">v2.2.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.1.0/examples">v2.1.1</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v2.0.0/examples">v2.0.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v1.2.0/examples">v1.2.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v1.1.0/examples">v1.1.0</a></li> <li><a href="https://github.com/huggingface/transformers/tree/v1.0.0/examples">v1.0.0</a></li></ul></details> <p data-svelte-h="svelte-1ebeyhp">然后切换您clone的 🤗 Transformers 仓到特定的版本，例如v3.5.1：</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="language-bash "><!-- HTML_TAG_START -->git checkout tags/v3.5.1<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-rb945x">在安装了正确的库版本后，进入您选择的版本的<code>example</code>文件夹并安装例子要求的环境：</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="language-bash "><!-- HTML_TAG_START -->pip install -r requirements.txt<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="运行脚本" 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="#运行脚本"><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>运行脚本</span></h2> <p data-svelte-h="svelte-1nzsvd0">示例脚本从🤗 <a href="https://huggingface.co/docs/datasets/" rel="nofollow">Datasets</a>库下载并预处理数据集。然后，脚本通过<a href="https://huggingface.co/docs/transformers/main_classes/trainer" rel="nofollow">Trainer</a>使用支持摘要任务的架构对数据集进行微调。以下示例展示了如何在<a href="https://huggingface.co/datasets/cnn_dailymail" rel="nofollow">CNN/DailyMail</a>数据集上微调<a href="https://huggingface.co/google-t5/t5-small" rel="nofollow">T5-small</a>。由于T5模型的训练方式，它需要一个额外的<code>source_prefix</code>参数。这个提示让T5知道这是一个摘要任务。</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="language-bash "><!-- HTML_TAG_START -->python examples/pytorch/summarization/run_summarization.py \
	--model_name_or_path google-t5/t5-small \
	--do_train \
	--do_eval \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--predict_with_generate<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="分布式训练和混合精度" 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="#分布式训练和混合精度"><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>分布式训练和混合精度</span></h2> <p data-svelte-h="svelte-1xszoa1"><a href="https://huggingface.co/docs/transformers/main_classes/trainer" rel="nofollow">Trainer</a> 支持分布式训练和混合精度，这意味着你也可以在脚本中使用它。要启用这两个功能，可以做如下设置：</p> <ul data-svelte-h="svelte-zaadqd"><li>添加 <code>fp16</code> 参数以启用混合精度。</li> <li>使用 <code>nproc_per_node</code> 参数设置使用的GPU数量。</li></ul> <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="language-bash "><!-- HTML_TAG_START -->torchrun \
	--nproc_per_node 8 pytorch/summarization/run_summarization.py \
	--fp16 \
	--model_name_or_path google-t5/t5-small \
	--do_train \
	--do_eval \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--predict_with_generate<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="在tpu上运行脚本" 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="#在tpu上运行脚本"><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>在TPU上运行脚本</span></h2> <p data-svelte-h="svelte-4tu5tb">张量处理单元（TPUs）是专门设计用于加速性能的。PyTorch使用 <a href="https://github.com/pytorch/xla/blob/master/README.md" rel="nofollow">PyTorch/XLA</a> 支持TPU。要使用TPU，请启动<code>xla_spawn.py</code>脚本并使用<code>num_cores</code>参数设置要使用的TPU核心数量。</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="language-bash "><!-- HTML_TAG_START -->python xla_spawn.py --num_cores 8 \
	summarization/run_summarization.py \
	--model_name_or_path google-t5/t5-small \
	--do_train \
	--do_eval \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--predict_with_generate<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="基于-accelerate运行脚本" 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="#基于-accelerate运行脚本"><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>基于🤗 Accelerate运行脚本</span></h2> <p data-svelte-h="svelte-1d5gx6n">🤗 <a href="https://huggingface.co/docs/accelerate" rel="nofollow">Accelerate</a> 是一个仅支持 PyTorch 的库，它提供了一种统一的方法来在不同类型的设置（仅 CPU、多个 GPU、多个TPU）上训练模型，同时保持对 PyTorch 训练循环的完全可见性。如果你还没有安装 🤗 Accelerate，请确保你已经安装了它：</p> <blockquote data-svelte-h="svelte-1hi61t"><p>注意：由于 Accelerate 正在快速发展，因此必须安装 git 版本的 accelerate 来运行脚本。</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="language-bash "><!-- HTML_TAG_START -->pip install git+https://github.com/huggingface/accelerate<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1sqmzpd">你需要使用<code>run_summarization_no_trainer.py</code>脚本，而不是<code>run_summarization.py</code>脚本。🤗 Accelerate支持的脚本需要在文件夹中有一个<code>task_no_trainer.py</code>文件。首先运行以下命令以创建并保存配置文件：</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="language-bash "><!-- HTML_TAG_START -->accelerate config<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1sleoq5">检测您的设置以确保配置正确：</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="language-bash "><!-- HTML_TAG_START -->accelerate <span class="hljs-built_in">test</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-jem89t">现在您可以开始训练模型了：</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="language-bash "><!-- HTML_TAG_START -->accelerate launch run_summarization_no_trainer.py \
	--model_name_or_path google-t5/t5-small \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir ~/tmp/tst-summarization<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="使用自定义数据集" 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="#使用自定义数据集"><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>使用自定义数据集</span></h2> <p data-svelte-h="svelte-1sbqsjd">摘要脚本支持自定义数据集，只要它们是CSV或JSON Line文件。当你使用自己的数据集时，需要指定一些额外的参数：</p> <ul data-svelte-h="svelte-1gfb9n1"><li><code>train_file</code> 和 <code>validation_file</code> 分别指定您的训练和验证文件的路径。</li> <li><code>text_column</code> 是输入要进行摘要的文本。</li> <li><code>summary_column</code> 是目标输出的文本。</li></ul> <p data-svelte-h="svelte-287ma3">使用自定义数据集的摘要脚本看起来是这样的：</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="language-bash "><!-- HTML_TAG_START -->python examples/pytorch/summarization/run_summarization.py \
	--model_name_or_path google-t5/t5-small \
	--do_train \
	--do_eval \
	--train_file path_to_csv_or_jsonlines_file \
	--validation_file path_to_csv_or_jsonlines_file \
	--text_column text_column_name \
	--summary_column summary_column_name \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--predict_with_generate<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="测试脚本" 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="#测试脚本"><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>测试脚本</span></h2> <p data-svelte-h="svelte-1h7zdvc">通常，在提交整个数据集之前，最好先在较少的数据集示例上运行脚本，以确保一切按预期工作,因为完整数据集的处理可能需要花费几个小时的时间。使用以下参数将数据集截断为最大样本数：</p> <ul data-svelte-h="svelte-14g5f8g"><li><code>max_train_samples</code></li> <li><code>max_eval_samples</code></li> <li><code>max_predict_samples</code></li></ul> <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="language-bash "><!-- HTML_TAG_START -->python examples/pytorch/summarization/run_summarization.py \
	--model_name_or_path google-t5/t5-small \
	--max_train_samples 50 \
	--max_eval_samples 50 \
	--max_predict_samples 50 \
	--do_train \
	--do_eval \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--predict_with_generate<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1e4r7qi">并非所有示例脚本都支持<code>max_predict_samples</code>参数。如果您不确定您的脚本是否支持此参数，请添加<code>-h</code>参数进行检查：</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="language-bash "><!-- HTML_TAG_START -->examples/pytorch/summarization/run_summarization.py -h<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="从checkpoint恢复训练" 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="#从checkpoint恢复训练"><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>从checkpoint恢复训练</span></h2> <p data-svelte-h="svelte-ixjp8">另一个有用的选项是从之前的checkpoint恢复训练。这将确保在训练中断时，您可以从之前停止的地方继续进行，而无需重新开始。有两种方法可以从checkpoint恢复训练。</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="language-bash "><!-- HTML_TAG_START -->python examples/pytorch/summarization/run_summarization.py
	--model_name_or_path google-t5/t5-small \
	--do_train \
	--do_eval \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--resume_from_checkpoint path_to_specific_checkpoint \
	--predict_with_generate<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="分享模型" 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="#分享模型"><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>分享模型</span></h2> <p data-svelte-h="svelte-14o7a7e">所有脚本都可以将您的最终模型上传到<a href="https://huggingface.co/models" rel="nofollow">Model Hub</a>。在开始之前，请确保您已登录Hugging Face：</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="language-bash "><!-- HTML_TAG_START -->hf auth login<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-s1qcjt">然后，在脚本中添加<code>push_to_hub</code>参数。这个参数会创建一个带有您Hugging Face用户名和<code>output_dir</code>中指定的文件夹名称的仓库。</p> <p data-svelte-h="svelte-1dapele">为了给您的仓库指定一个特定的名称，使用<code>push_to_hub_model_id</code>参数来添加它。该仓库将自动列出在您的命名空间下。</p> <p data-svelte-h="svelte-dckksr">以下示例展示了如何上传具有特定仓库名称的模型：</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="language-bash "><!-- HTML_TAG_START -->python examples/pytorch/summarization/run_summarization.py
	--model_name_or_path google-t5/t5-small \
	--do_train \
	--do_eval \
	--dataset_name cnn_dailymail \
	--dataset_config <span class="hljs-string">"3.0.0"</span> \
	--source_prefix <span class="hljs-string">"summarize: "</span> \
	--push_to_hub \
	--push_to_hub_model_id finetuned-t5-cnn_dailymail \
	--output_dir /tmp/tst-summarization \
	--per_device_train_batch_size=4 \
	--per_device_eval_batch_size=4 \
	--predict_with_generate<!-- 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/transformers/blob/main/docs/source/zh/run_scripts.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_zpy4e4 = {
	assets: "/docs/transformers/main/zh",
	base: "/docs/transformers/main/zh",
	env: {}
	};

	const element = document.currentScript.parentElement;

	const data = [null,null];

	Promise.all([
	import("/docs/transformers/main/zh/_app/immutable/entry/start.7a5389c5.js"),
	import("/docs/transformers/main/zh/_app/immutable/entry/app.ac7d4b06.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 61],
	data,
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 52.8 kB
Xet hash:: b543f3992a329940c4234cefd1dd590324b5d339be02f6b7aafbf2b657aadf94

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