Hugging Face's logo

Buckets:
hf-doc-build
/
doc-dev

Files
xet
hf-doc-build/doc-dev / transformers /pr_33913 /de /run_scripts.html
download
raw
70.8 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Trainieren mit einem Skript&quot;,&quot;local&quot;:&quot;trainieren-mit-einem-skript&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Einrichtung&quot;,&quot;local&quot;:&quot;einrichtung&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Ein Skript ausführen&quot;,&quot;local&quot;:&quot;ein-skript-ausführen&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Verteiltes Training und gemischte Präzision&quot;,&quot;local&quot;:&quot;verteiltes-training-und-gemischte-präzision&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Ein Skript auf einer TPU ausführen&quot;,&quot;local&quot;:&quot;ein-skript-auf-einer-tpu-ausführen&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Führen Sie ein Skript mit 🤗 Accelerate aus.&quot;,&quot;local&quot;:&quot;führen-sie-ein-skript-mit--accelerate-aus&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Verwenden Sie einen benutzerdefinierten Datensatz&quot;,&quot;local&quot;:&quot;verwenden-sie-einen-benutzerdefinierten-datensatz&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Testen Sie ein Skript&quot;,&quot;local&quot;:&quot;testen-sie-ein-skript&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Training vom Kontrollpunkt fortsetzen&quot;,&quot;local&quot;:&quot;training-vom-kontrollpunkt-fortsetzen&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Teilen Sie Ihr Modell&quot;,&quot;local&quot;:&quot;teilen-sie-ihr-modell&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
 <link href="/docs/transformers/pr_33913/de/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/entry/start.4cb54951.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/scheduler.987d3921.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/singletons.cc4df6a9.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/index.42b1abb7.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/paths.7a1265c0.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/entry/app.8a9a40b8.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/index.c8b1fed4.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/nodes/0.509a6df1.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/each.e59479a4.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/nodes/16.b3379ea7.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/CodeBlock.18094d58.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/Markdown.10b8ab04.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/globals.7f7f1b26.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/de/_app/immutable/chunks/EditOnGithub.a3fde557.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Trainieren mit einem Skript&quot;,&quot;local&quot;:&quot;trainieren-mit-einem-skript&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Einrichtung&quot;,&quot;local&quot;:&quot;einrichtung&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Ein Skript ausführen&quot;,&quot;local&quot;:&quot;ein-skript-ausführen&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Verteiltes Training und gemischte Präzision&quot;,&quot;local&quot;:&quot;verteiltes-training-und-gemischte-präzision&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Ein Skript auf einer TPU ausführen&quot;,&quot;local&quot;:&quot;ein-skript-auf-einer-tpu-ausführen&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Führen Sie ein Skript mit 🤗 Accelerate aus.&quot;,&quot;local&quot;:&quot;führen-sie-ein-skript-mit--accelerate-aus&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Verwenden Sie einen benutzerdefinierten Datensatz&quot;,&quot;local&quot;:&quot;verwenden-sie-einen-benutzerdefinierten-datensatz&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Testen Sie ein Skript&quot;,&quot;local&quot;:&quot;testen-sie-ein-skript&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Training vom Kontrollpunkt fortsetzen&quot;,&quot;local&quot;:&quot;training-vom-kontrollpunkt-fortsetzen&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Teilen Sie Ihr Modell&quot;,&quot;local&quot;:&quot;teilen-sie-ihr-modell&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="trainieren-mit-einem-skript" 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="#trainieren-mit-einem-skript"><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>Trainieren mit einem Skript</span></h1> <p data-svelte-h="svelte-10suli8">Neben den 🤗 Transformers <a href="./notebooks">notebooks</a> gibt es auch Beispielskripte, die zeigen, wie man ein Modell für eine Aufgabe mit <a href="https://github.com/huggingface/transformers/tree/main/examples/pytorch" rel="nofollow">PyTorch</a>, <a href="https://github.com/huggingface/transformers/tree/main/examples/tensorflow" rel="nofollow">TensorFlow</a> oder <a href="https://github.com/huggingface/transformers/tree/main/examples/flax" rel="nofollow">JAX/Flax</a> trainiert.</p> <p data-svelte-h="svelte-1u2bhpe">Sie werden auch Skripte finden, die wir in unseren <a href="https://github.com/huggingface/transformers/tree/main/examples/research_projects" rel="nofollow">Forschungsprojekten</a> und <a href="https://github.com/huggingface/transformers/tree/main/examples/legacy" rel="nofollow">Legacy-Beispielen</a> verwendet haben und die größtenteils von der Community stammen. Diese Skripte werden nicht aktiv gepflegt und erfordern eine bestimmte Version von 🤗 Transformers, die höchstwahrscheinlich nicht mit der neuesten Version der Bibliothek kompatibel ist.</p> <p data-svelte-h="svelte-1yjk519">Es wird nicht erwartet, dass die Beispielskripte bei jedem Problem sofort funktionieren. Möglicherweise müssen Sie das Skript an das Problem anpassen, das Sie zu lösen versuchen. Um Ihnen dabei zu helfen, legen die meisten Skripte vollständig offen, wie die Daten vorverarbeitet werden, so dass Sie sie nach Bedarf für Ihren Anwendungsfall bearbeiten können.</p> <p data-svelte-h="svelte-1p21xl">Für jede Funktion, die Sie in einem Beispielskript implementieren möchten, diskutieren Sie bitte im <a href="https://discuss.huggingface.co/" rel="nofollow">Forum</a> oder in einem <a href="https://github.com/huggingface/transformers/issues" rel="nofollow">issue</a>, bevor Sie einen Pull Request einreichen. Wir freuen uns zwar über Fehlerkorrekturen, aber es ist unwahrscheinlich, dass wir einen Pull Request zusammenführen, der mehr Funktionalität auf Kosten der Lesbarkeit hinzufügt.</p> <p data-svelte-h="svelte-f7vti5">Diese Anleitung zeigt Ihnen, wie Sie ein Beispiel für ein Trainingsskript zur Zusammenfassung in <a href="https://github.com/huggingface/transformers/tree/main/examples/pytorch/summarization" rel="nofollow">PyTorch</a> und <a href="https://github.com/huggingface/transformers/tree/main/examples/tensorflow/summarization" rel="nofollow">TensorFlow</a> ausführen können. Sofern nicht anders angegeben, sollten alle Beispiele mit beiden Frameworks funktionieren.</p> <h2 class="relative group"><a id="einrichtung" 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="#einrichtung"><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>Einrichtung</span></h2> <p data-svelte-h="svelte-1k7wbk1">Um die neueste Version der Beispielskripte erfolgreich auszuführen, <strong>müssen Sie 🤗 Transformers aus dem Quellcode</strong> in einer neuen virtuellen Umgebung installieren:</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 -->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-1v3grbn">Für ältere Versionen der Beispielskripte klicken Sie auf die Umschalttaste unten:</p> <details data-svelte-h="svelte-161spkh"><summary>Beispiele für ältere Versionen von 🤗 Transformers</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-18wy8ey">Dann stellen Sie Ihren aktuellen Klon von 🤗 Transformers auf eine bestimmte Version um, z.B. 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=""><!-- HTML_TAG_START -->git checkout tags/v3.5.1<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1akgvwd">Nachdem Sie die richtige Bibliotheksversion eingerichtet haben, navigieren Sie zu dem Beispielordner Ihrer Wahl und installieren die beispielspezifischen Anforderungen:</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 -->pip install -r requirements.txt<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="ein-skript-ausführen" 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="#ein-skript-ausführen"><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>Ein Skript ausführen</span></h2> <div class="space-y-10 py-6 2xl:py-8 2xl:-mx-4"> <div class="border border-gray-200 rounded-xl px-4 relative"><div class="flex h-[22px] mt-[-12.5px] justify-between leading-none"><div class="flex px-1 items-center space-x-1 bg-white dark:bg-gray-950"><svg class="" 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> <span>Pytorch</span></div> <div class="cursor-pointer flex items-center justify-center space-x-1 text-sm px-2 bg-white dark:bg-gray-950 hover:underline leading-none"><svg class="" width="0.9em" height="0.9em" viewBox="0 0 10 9" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M1.39125 1.9725L0.0883333 0.669997L0.677917 0.0804138L8.9275 8.33041L8.33792 8.91958L6.95875 7.54041C6.22592 8.00523 5.37572 8.25138 4.50792 8.25C2.26125 8.25 0.392083 6.63333 0 4.5C0.179179 3.52946 0.667345 2.64287 1.39167 1.9725H1.39125ZM5.65667 6.23833L5.04667 5.62833C4.81335 5.73996 4.55116 5.77647 4.29622 5.73282C4.04129 5.68918 3.80617 5.56752 3.62328 5.38463C3.44039 5.20175 3.31874 4.96663 3.27509 4.71169C3.23144 4.45676 3.26795 4.19456 3.37958 3.96125L2.76958 3.35125C2.50447 3.75187 2.38595 4.2318 2.4341 4.70978C2.48225 5.18777 2.6941 5.63442 3.0338 5.97411C3.37349 6.31381 3.82015 6.52567 4.29813 6.57382C4.77611 6.62197 5.25605 6.50345 5.65667 6.23833ZM2.83042 1.06666C3.35 0.862497 3.91625 0.749997 4.50792 0.749997C6.75458 0.749997 8.62375 2.36666 9.01583 4.5C8.88816 5.19404 8.60119 5.84899 8.1775 6.41333L6.56917 4.805C6.61694 4.48317 6.58868 4.15463 6.48664 3.84569C6.3846 3.53675 6.21162 3.256 5.98156 3.02594C5.7515 2.79588 5.47075 2.6229 5.16181 2.52086C4.85287 2.41882 4.52433 2.39056 4.2025 2.43833L2.83042 1.06708V1.06666Z" fill="currentColor"></path></svg> <span>Hide Pytorch content</span></div></div> <div class="framework-content"> <p data-svelte-h="svelte-1m5xzyk">Das Beispielskript lädt einen Datensatz aus der 🤗 <a href="https://huggingface.co/docs/datasets/" rel="nofollow">Datasets</a> Bibliothek herunter und verarbeitet ihn vor. Dann nimmt das Skript eine Feinabstimmung eines Datensatzes mit dem <a href="https://huggingface.co/docs/transformers/main_classes/trainer" rel="nofollow">Trainer</a> auf einer Architektur vor, die eine Zusammenfassung unterstützt. Das folgende Beispiel zeigt, wie die Feinabstimmung von <a href="https://huggingface.co/google-t5/t5-small" rel="nofollow">T5-small</a> auf dem Datensatz <a href="https://huggingface.co/datasets/cnn_dailymail" rel="nofollow">CNN/DailyMail</a> durchgeführt wird. Das T5-Modell benötigt aufgrund der Art und Weise, wie es trainiert wurde, ein zusätzliches Argument <code>source_prefix</code>. Mit dieser Eingabeaufforderung weiß T5, dass es sich um eine Zusammenfassungsaufgabe handelt.</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 -->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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size=4 \
 --per_device_eval_batch_size=4 \
 --overwrite_output_dir \
 --predict_with_generate<!-- HTML_TAG_END --></pre></div></div></div> <div class="border border-gray-200 rounded-xl px-4 relative"><div class="flex h-[22px] mt-[-12.5px] justify-between leading-none"><div class="flex px-1 items-center space-x-1 bg-white dark:bg-gray-950"><svg class="" 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> <span>TensorFlow</span></div> <div class="cursor-pointer flex items-center justify-center space-x-1 text-sm px-2 bg-white dark:bg-gray-950 hover:underline leading-none"><svg class="" width="0.9em" height="0.9em" viewBox="0 0 10 9" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M1.39125 1.9725L0.0883333 0.669997L0.677917 0.0804138L8.9275 8.33041L8.33792 8.91958L6.95875 7.54041C6.22592 8.00523 5.37572 8.25138 4.50792 8.25C2.26125 8.25 0.392083 6.63333 0 4.5C0.179179 3.52946 0.667345 2.64287 1.39167 1.9725H1.39125ZM5.65667 6.23833L5.04667 5.62833C4.81335 5.73996 4.55116 5.77647 4.29622 5.73282C4.04129 5.68918 3.80617 5.56752 3.62328 5.38463C3.44039 5.20175 3.31874 4.96663 3.27509 4.71169C3.23144 4.45676 3.26795 4.19456 3.37958 3.96125L2.76958 3.35125C2.50447 3.75187 2.38595 4.2318 2.4341 4.70978C2.48225 5.18777 2.6941 5.63442 3.0338 5.97411C3.37349 6.31381 3.82015 6.52567 4.29813 6.57382C4.77611 6.62197 5.25605 6.50345 5.65667 6.23833ZM2.83042 1.06666C3.35 0.862497 3.91625 0.749997 4.50792 0.749997C6.75458 0.749997 8.62375 2.36666 9.01583 4.5C8.88816 5.19404 8.60119 5.84899 8.1775 6.41333L6.56917 4.805C6.61694 4.48317 6.58868 4.15463 6.48664 3.84569C6.3846 3.53675 6.21162 3.256 5.98156 3.02594C5.7515 2.79588 5.47075 2.6229 5.16181 2.52086C4.85287 2.41882 4.52433 2.39056 4.2025 2.43833L2.83042 1.06708V1.06666Z" fill="currentColor"></path></svg> <span>Hide TensorFlow content</span></div></div> <div class="framework-content"> <p data-svelte-h="svelte-ja7ep7">Das Beispielskript lädt einen Datensatz aus der 🤗 <a href="https://huggingface.co/docs/datasets/" rel="nofollow">Datasets</a> Bibliothek herunter und verarbeitet ihn vor. Anschließend nimmt das Skript die Feinabstimmung eines Datensatzes mit Keras auf einer Architektur vor, die die Zusammenfassung unterstützt. Das folgende Beispiel zeigt, wie die Feinabstimmung von <a href="https://huggingface.co/google-t5/t5-small" rel="nofollow">T5-small</a> auf dem <a href="https://huggingface.co/datasets/cnn_dailymail" rel="nofollow">CNN/DailyMail</a> Datensatz durchgeführt wird. Das T5-Modell benötigt aufgrund der Art und Weise, wie es trainiert wurde, ein zusätzliches Argument <code>source_prefix</code>. Mit dieser Eingabeaufforderung weiß T5, dass es sich um eine Zusammenfassungsaufgabe handelt.</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 -->python examples/tensorflow/summarization/run_summarization.py \
 --model_name_or_path google-t5/t5-small \
 --dataset_name cnn_dailymail \
 --dataset_config <span class="hljs-string">&quot;3.0.0&quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size 8 \
 --per_device_eval_batch_size 16 \
 --num_train_epochs 3 \
 --do_train \
 --do_eval<!-- HTML_TAG_END --></pre></div> </div></div> </div> <h2 class="relative group"><a id="verteiltes-training-und-gemischte-präzision" 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="#verteiltes-training-und-gemischte-präzision"><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>Verteiltes Training und gemischte Präzision</span></h2> <p data-svelte-h="svelte-ygqg3g">Der <a href="https://huggingface.co/docs/transformers/main_classes/trainer" rel="nofollow">Trainer</a> unterstützt verteiltes Training und gemischte Präzision, d.h. Sie können ihn auch in einem Skript verwenden. So aktivieren Sie diese beiden Funktionen:</p> <ul data-svelte-h="svelte-1l33p46"><li>Fügen Sie das Argument <code>fp16</code> hinzu, um gemischte Genauigkeit zu aktivieren.</li> <li>Legen Sie die Anzahl der zu verwendenden GPUs mit dem Argument <code>nproc_per_node</code> fest.</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=""><!-- 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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size=4 \
 --per_device_eval_batch_size=4 \
 --overwrite_output_dir \
 --predict_with_generate<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1hk9llr">TensorFlow-Skripte verwenden eine <a href="https://www.tensorflow.org/guide/distributed_training#mirroredstrategy" rel="nofollow"><code>MirroredStrategy</code></a> für verteiltes Training, und Sie müssen dem Trainingsskript keine zusätzlichen Argumente hinzufügen. Das TensorFlow-Skript verwendet standardmäßig mehrere GPUs, wenn diese verfügbar sind.</p> <h2 class="relative group"><a id="ein-skript-auf-einer-tpu-ausführen" 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="#ein-skript-auf-einer-tpu-ausführen"><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>Ein Skript auf einer TPU ausführen</span></h2> <div class="space-y-10 py-6 2xl:py-8 2xl:-mx-4"> <div class="border border-gray-200 rounded-xl px-4 relative"><div class="flex h-[22px] mt-[-12.5px] justify-between leading-none"><div class="flex px-1 items-center space-x-1 bg-white dark:bg-gray-950"><svg class="" 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> <span>Pytorch</span></div> <div class="cursor-pointer flex items-center justify-center space-x-1 text-sm px-2 bg-white dark:bg-gray-950 hover:underline leading-none"><svg class="" width="0.9em" height="0.9em" viewBox="0 0 10 9" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M1.39125 1.9725L0.0883333 0.669997L0.677917 0.0804138L8.9275 8.33041L8.33792 8.91958L6.95875 7.54041C6.22592 8.00523 5.37572 8.25138 4.50792 8.25C2.26125 8.25 0.392083 6.63333 0 4.5C0.179179 3.52946 0.667345 2.64287 1.39167 1.9725H1.39125ZM5.65667 6.23833L5.04667 5.62833C4.81335 5.73996 4.55116 5.77647 4.29622 5.73282C4.04129 5.68918 3.80617 5.56752 3.62328 5.38463C3.44039 5.20175 3.31874 4.96663 3.27509 4.71169C3.23144 4.45676 3.26795 4.19456 3.37958 3.96125L2.76958 3.35125C2.50447 3.75187 2.38595 4.2318 2.4341 4.70978C2.48225 5.18777 2.6941 5.63442 3.0338 5.97411C3.37349 6.31381 3.82015 6.52567 4.29813 6.57382C4.77611 6.62197 5.25605 6.50345 5.65667 6.23833ZM2.83042 1.06666C3.35 0.862497 3.91625 0.749997 4.50792 0.749997C6.75458 0.749997 8.62375 2.36666 9.01583 4.5C8.88816 5.19404 8.60119 5.84899 8.1775 6.41333L6.56917 4.805C6.61694 4.48317 6.58868 4.15463 6.48664 3.84569C6.3846 3.53675 6.21162 3.256 5.98156 3.02594C5.7515 2.79588 5.47075 2.6229 5.16181 2.52086C4.85287 2.41882 4.52433 2.39056 4.2025 2.43833L2.83042 1.06708V1.06666Z" fill="currentColor"></path></svg> <span>Hide Pytorch content</span></div></div> <div class="framework-content"> <p data-svelte-h="svelte-1x2esg2">Tensor Processing Units (TPUs) sind speziell für die Beschleunigung der Leistung konzipiert. PyTorch unterstützt TPUs mit dem <a href="https://www.tensorflow.org/xla" rel="nofollow">XLA</a> Deep Learning Compiler (siehe <a href="https://github.com/pytorch/xla/blob/master/README.md" rel="nofollow">hier</a> für weitere Details). Um eine TPU zu verwenden, starten Sie das Skript <code>xla_spawn.py</code> und verwenden das Argument <code>num_cores</code>, um die Anzahl der TPU-Kerne festzulegen, die Sie verwenden möchten.</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 -->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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size=4 \
 --per_device_eval_batch_size=4 \
 --overwrite_output_dir \
 --predict_with_generate<!-- HTML_TAG_END --></pre></div></div></div> <div class="border border-gray-200 rounded-xl px-4 relative"><div class="flex h-[22px] mt-[-12.5px] justify-between leading-none"><div class="flex px-1 items-center space-x-1 bg-white dark:bg-gray-950"><svg class="" 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> <span>TensorFlow</span></div> <div class="cursor-pointer flex items-center justify-center space-x-1 text-sm px-2 bg-white dark:bg-gray-950 hover:underline leading-none"><svg class="" width="0.9em" height="0.9em" viewBox="0 0 10 9" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M1.39125 1.9725L0.0883333 0.669997L0.677917 0.0804138L8.9275 8.33041L8.33792 8.91958L6.95875 7.54041C6.22592 8.00523 5.37572 8.25138 4.50792 8.25C2.26125 8.25 0.392083 6.63333 0 4.5C0.179179 3.52946 0.667345 2.64287 1.39167 1.9725H1.39125ZM5.65667 6.23833L5.04667 5.62833C4.81335 5.73996 4.55116 5.77647 4.29622 5.73282C4.04129 5.68918 3.80617 5.56752 3.62328 5.38463C3.44039 5.20175 3.31874 4.96663 3.27509 4.71169C3.23144 4.45676 3.26795 4.19456 3.37958 3.96125L2.76958 3.35125C2.50447 3.75187 2.38595 4.2318 2.4341 4.70978C2.48225 5.18777 2.6941 5.63442 3.0338 5.97411C3.37349 6.31381 3.82015 6.52567 4.29813 6.57382C4.77611 6.62197 5.25605 6.50345 5.65667 6.23833ZM2.83042 1.06666C3.35 0.862497 3.91625 0.749997 4.50792 0.749997C6.75458 0.749997 8.62375 2.36666 9.01583 4.5C8.88816 5.19404 8.60119 5.84899 8.1775 6.41333L6.56917 4.805C6.61694 4.48317 6.58868 4.15463 6.48664 3.84569C6.3846 3.53675 6.21162 3.256 5.98156 3.02594C5.7515 2.79588 5.47075 2.6229 5.16181 2.52086C4.85287 2.41882 4.52433 2.39056 4.2025 2.43833L2.83042 1.06708V1.06666Z" fill="currentColor"></path></svg> <span>Hide TensorFlow content</span></div></div> <div class="framework-content"> <p data-svelte-h="svelte-q7y3mk">Tensor Processing Units (TPUs) sind speziell für die Beschleunigung der Leistung konzipiert. TensorFlow Skripte verwenden eine <a href="https://www.tensorflow.org/guide/distributed_training#tpustrategy" rel="nofollow"><code>TPUStrategy</code></a> für das Training auf TPUs. Um eine TPU zu verwenden, übergeben Sie den Namen der TPU-Ressource an das Argument <code>tpu</code>.</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 -->python run_summarization.py \
 --tpu name_of_tpu_resource \
 --model_name_or_path google-t5/t5-small \
 --dataset_name cnn_dailymail \
 --dataset_config <span class="hljs-string">&quot;3.0.0&quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size 8 \
 --per_device_eval_batch_size 16 \
 --num_train_epochs 3 \
 --do_train \
 --do_eval<!-- HTML_TAG_END --></pre></div> </div></div> </div> <h2 class="relative group"><a id="führen-sie-ein-skript-mit--accelerate-aus" 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="#führen-sie-ein-skript-mit--accelerate-aus"><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>Führen Sie ein Skript mit 🤗 Accelerate aus.</span></h2> <p data-svelte-h="svelte-hlvrr2">🤗 <a href="https://huggingface.co/docs/accelerate" rel="nofollow">Accelerate</a> ist eine reine PyTorch-Bibliothek, die eine einheitliche Methode für das Training eines Modells auf verschiedenen Arten von Setups (nur CPU, mehrere GPUs, TPUs) bietet und dabei die vollständige Transparenz der PyTorch-Trainingsschleife beibehält. Stellen Sie sicher, dass Sie 🤗 Accelerate installiert haben, wenn Sie es nicht bereits haben:</p> <blockquote><p data-svelte-h="svelte-1oqwqfa">Hinweis: Da Accelerate schnell weiterentwickelt wird, muss die Git-Version von Accelerate installiert sein, um die Skripte auszuführen.</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 -->pip install git+https://github.com/huggingface/accelerate<!-- HTML_TAG_END --></pre></div></blockquote> <p data-svelte-h="svelte-sajudj">Anstelle des Skripts <code>run_summarization.py</code> müssen Sie das Skript <code>run_summarization_no_trainer.py</code> verwenden. Die von Accelerate unterstützten Skripte haben eine Datei <code>task_no_trainer.py</code> im Ordner. Beginnen Sie mit dem folgenden Befehl, um eine Konfigurationsdatei zu erstellen und zu speichern:</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 -->accelerate config<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ba3uqv">Testen Sie Ihre Einrichtung, um sicherzustellen, dass sie korrekt konfiguriert ist:</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 -->accelerate <span class="hljs-built_in">test</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-189gcch">Jetzt sind Sie bereit, das Training zu starten:</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 -->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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir ~/tmp/tst-summarization<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="verwenden-sie-einen-benutzerdefinierten-datensatz" 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="#verwenden-sie-einen-benutzerdefinierten-datensatz"><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>Verwenden Sie einen benutzerdefinierten Datensatz</span></h2> <p data-svelte-h="svelte-70977j">Das Verdichtungsskript unterstützt benutzerdefinierte Datensätze, solange es sich um eine CSV- oder JSON-Line-Datei handelt. Wenn Sie Ihren eigenen Datensatz verwenden, müssen Sie mehrere zusätzliche Argumente angeben:</p> <ul data-svelte-h="svelte-1lpyzfn"><li><code>train_file</code> und <code>validation_file</code> geben den Pfad zu Ihren Trainings- und Validierungsdateien an.</li> <li><code>text_column</code> ist der Eingabetext, der zusammengefasst werden soll.</li> <li>Summary_column” ist der auszugebende Zieltext.</li></ul> <p data-svelte-h="svelte-wbqae1">Ein Zusammenfassungsskript, das einen benutzerdefinierten Datensatz verwendet, würde wie folgt aussehen:</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 -->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">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --overwrite_output_dir \
 --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="testen-sie-ein-skript" 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="#testen-sie-ein-skript"><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>Testen Sie ein Skript</span></h2> <p data-svelte-h="svelte-1etiu5b">Es ist oft eine gute Idee, Ihr Skript an einer kleineren Anzahl von Beispielen für Datensätze auszuführen, um sicherzustellen, dass alles wie erwartet funktioniert, bevor Sie sich auf einen ganzen Datensatz festlegen, dessen Fertigstellung Stunden dauern kann. Verwenden Sie die folgenden Argumente, um den Datensatz auf eine maximale Anzahl von Stichproben zu beschränken:</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=""><!-- 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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size=4 \
 --per_device_eval_batch_size=4 \
 --overwrite_output_dir \
 --predict_with_generate<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-17jr0zz">Nicht alle Beispielskripte unterstützen das Argument <code>max_predict_samples</code>. Wenn Sie sich nicht sicher sind, ob Ihr Skript dieses Argument unterstützt, fügen Sie das Argument <code>-h</code> hinzu, um dies zu überprüfen:</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 -->examples/pytorch/summarization/run_summarization.py -h<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="training-vom-kontrollpunkt-fortsetzen" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#training-vom-kontrollpunkt-fortsetzen"><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>Training vom Kontrollpunkt fortsetzen</span></h2> <p data-svelte-h="svelte-7kw3hv">Eine weitere hilfreiche Option, die Sie aktivieren können, ist die Wiederaufnahme des Trainings von einem früheren Kontrollpunkt aus. Auf diese Weise können Sie im Falle einer Unterbrechung Ihres Trainings dort weitermachen, wo Sie aufgehört haben, ohne von vorne beginnen zu müssen. Es gibt zwei Methoden, um das Training von einem Kontrollpunkt aus wieder aufzunehmen.</p> <p data-svelte-h="svelte-10lk07m">Die erste Methode verwendet das Argument <code>output_dir previous_output_dir</code>, um das Training ab dem letzten in <code>output_dir</code> gespeicherten Kontrollpunkt wieder aufzunehmen. In diesem Fall sollten Sie <code>overwrite_output_dir</code> entfernen:</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 -->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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size=4 \
 --per_device_eval_batch_size=4 \
 --output_dir previous_output_dir \
 --predict_with_generate<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1jvl0eq">Die zweite Methode verwendet das Argument <code>Resume_from_checkpoint path_to_specific_checkpoint</code>, um das Training ab einem bestimmten Checkpoint-Ordner wieder aufzunehmen.</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 -->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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</span> \
 --output_dir /tmp/tst-summarization \
 --per_device_train_batch_size=4 \
 --per_device_eval_batch_size=4 \
 --overwrite_output_dir \
 --resume_from_checkpoint path_to_specific_checkpoint \
 --predict_with_generate<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="teilen-sie-ihr-modell" 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="#teilen-sie-ihr-modell"><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>Teilen Sie Ihr Modell</span></h2> <p data-svelte-h="svelte-9ovwzu">Alle Skripte können Ihr endgültiges Modell in den <a href="https://huggingface.co/models" rel="nofollow">Model Hub</a> hochladen. Stellen Sie sicher, dass Sie bei Hugging Face angemeldet sind, bevor Sie beginnen:</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 -->huggingface-cli login<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-x4eh9q">Dann fügen Sie dem Skript das Argument <code>push_to_hub</code> hinzu. Mit diesem Argument wird ein Repository mit Ihrem Hugging Face-Benutzernamen und dem in <code>output_dir</code> angegebenen Ordnernamen erstellt.</p> <p data-svelte-h="svelte-1mbkf4">Wenn Sie Ihrem Repository einen bestimmten Namen geben möchten, fügen Sie ihn mit dem Argument <code>push_to_hub_model_id</code> hinzu. Das Repository wird automatisch unter Ihrem Namensraum aufgeführt.</p> <p data-svelte-h="svelte-kk496d">Das folgende Beispiel zeigt, wie Sie ein Modell mit einem bestimmten Repository-Namen hochladen können:</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 -->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">&quot;3.0.0&quot;</span> \
 --source_prefix <span class="hljs-string">&quot;summarize: &quot;</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 \
 --overwrite_output_dir \
 --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/de/run_scripts.md" target="_blank"><span data-svelte-h="svelte-1kd6by1">&lt;</span> <span data-svelte-h="svelte-x0xyl0">&gt;</span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p>
<script>
 {
 __sveltekit_4w7cbd = {
 assets: "/docs/transformers/pr_33913/de",
 base: "/docs/transformers/pr_33913/de",
 env: {}
 };
 const element = document.currentScript.parentElement;
 const data = [null,null];
 Promise.all([
 import("/docs/transformers/pr_33913/de/_app/immutable/entry/start.4cb54951.js"),
 import("/docs/transformers/pr_33913/de/_app/immutable/entry/app.8a9a40b8.js")
 ]).then(([kit, app]) => {
 kit.start(app, element, {
 node_ids: [0, 16],
 data,
 form: null,
 error: null
 });
 });
 }
 </script>

Xet Storage Details

Size:
70.8 kB
·
Xet hash:
691d8c0b4879e31842b7e009c41f81edad3f7c8df74a78ff9fbba586384d2b9b

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