Buckets:

hf-doc-build
/

doc

Files
xet
hf-doc-build/doc / diffusers /main /ru /training /lora.html
mishig
about 2 hours ago
download
raw
41.9 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;LoRA&quot;,&quot;local&quot;:&quot;lora&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Параметры скрипта&quot;,&quot;local&quot;:&quot;script-parameters&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Скрипт обучения&quot;,&quot;local&quot;:&quot;training-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Запустите скрипт&quot;,&quot;local&quot;:&quot;launch-the-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Дальнейшие шаги&quot;,&quot;local&quot;:&quot;next-steps&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/>
		<link href="/docs/diffusers/main/en/_app/immutable/entry/start.D3LnKZ4R.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/BF9UNEM9.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DPnyVmXs.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/entry/app.C0rBoMxR.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DSGUT_OW.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/C045MWWx.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DnVIJfC5.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DsnmJJEf.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/nodes/0.DTZsP_3B.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/IaDvMxLQ.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/nodes/318.egALNfaM.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/bKdwVYpU.js" rel="modulepreload">
		<link href="/docs/diffusers/main/en/_app/immutable/chunks/DTd9IGSp.js" rel="modulepreload">
		<!--1b9mnc8--><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;LoRA&quot;,&quot;local&quot;:&quot;lora&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Script parameters&quot;,&quot;local&quot;:&quot;script-parameters&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Training script&quot;,&quot;local&quot;:&quot;training-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Launch the script&quot;,&quot;local&quot;:&quot;launch-the-script&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Next steps&quot;,&quot;local&quot;:&quot;next-steps&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/><!---->
		<link href="/docs/diffusers/main/en/_app/immutable/assets/0.tn0RQdqM.css" rel="modulepreload"> <!--[--><!--[0--><!--[--><!--[0--><!--[--><p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: inline-end; margin-inline-start: 10px; display: inline-flex; position: relative; z-index: 10;"><div dir="ltr" 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 dir="auto">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> <!--[-1--><!--]--></div><!----> <!--[0--><h1 class="relative group"><a id="lora" 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="#lora"><span><svg 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><!---->LoRA<!----></span></h1><!--]--><!----> <blockquote class="warning"><p>Это экспериментальная функция, и API может измениться в будущем.</p></blockquote> <p><a href="https://hf.co/papers/2106.09685" rel="nofollow">LoRA (Low-Rank Adaptation of Large Language Models)</a> — популярный и лёгкий метод обучения, который значительно сокращает количество обучаемых параметров. Он работает за счёт вставки в модель меньшего числа новых весов, и обучаются только они. Это делает обучение с LoRA намного быстрее, экономнее по памяти и приводит к получению меньших весов модели (несколько сотен МБ), которые проще хранить и передавать. LoRA также можно комбинировать с другими методами обучения, такими как DreamBooth, чтобы ускорить процесс обучения.</p> <blockquote class="tip"><p>LoRA очень универсален и поддерживается для <a href="https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora.py" rel="nofollow">DreamBooth</a>, <a href="https://github.com/huggingface/diffusers/blob/main/examples/kandinsky2_2/text_to_image/train_text_to_image_lora_decoder.py" rel="nofollow">Kandinsky 2.2</a>, <a href="https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora_sdxl.py" rel="nofollow">Stable Diffusion XL</a> и <a href="https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py" rel="nofollow">text-to-image</a>.</p></blockquote> <p>This guide will explore the <a href="https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py" rel="nofollow">train_text_to_image_lora.py</a> script to help you become more familiar with it, and how you can adapt it for your own use-case.</p> <p>Перед запуском скрипта убедитесь, что вы установили библиотеку из исходного кода:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-bash "><!---->git <span class="hljs-built_in">clone</span> https://github.com/huggingface/diffusers
<span class="hljs-built_in">cd</span> diffusers
pip install .<!----></pre></div><!----> <p>Перейдите в папку с примером, содержащую скрипт обучения, и установите необходимые зависимости для используемого вами скрипта:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-bash "><!----><span class="hljs-built_in">cd</span> examples/text_to_image
pip install -r requirements.txt<!----></pre></div><!----> <blockquote class="tip"><p>🤗 Accelerate — это библиотека, которая помогает обучать модели на нескольких GPU/TPU или с использованием смешанной точности. Она автоматически настраивает среду обучения в зависимости от вашего оборудования и окружения. Подробнее можно узнать в <a href="https://huggingface.co/docs/accelerate/quicktour" rel="nofollow">Кратком обзоре</a> 🤗 Accelerate.</p></blockquote> <p>Инициализируйте среду 🤗 Accelerate:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-bash "><!---->accelerate config<!----></pre></div><!----> <p>Чтобы настроить среду по умолчанию 🤗 Accelerate без выбора каких-либо конфигураций:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-bash "><!---->accelerate config default<!----></pre></div><!----> <p>Или, если ваша среда не поддерживает интерактивную оболочку, например, ноутбук, вы можете использовать:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-py "><!----><span class="hljs-keyword">from</span> accelerate.utils <span class="hljs-keyword">import</span> write_basic_config

write_basic_config()<!----></pre></div><!----> <p>Наконец, если вы хотите обучить модель на собственном наборе данных, ознакомьтесь с руководством <a href="create_dataset">Create a dataset for training</a>, чтобы узнать, как создать набор данных, совместимый со скриптом обучения.</p> <blockquote class="tip"><p>В следующих разделах выделены части скрипта обучения, важные для понимания того, как его модифицировать, но не все аспекты скрипта описаны подробно. Если вы хотите узнать больше, не стесняйтесь прочитать <a href="https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image_lora.py" rel="nofollow">скрипт</a> и сообщить нам, если у вас возникнут вопросы или замечания.</p></blockquote> <!--[1--><h2 class="relative group"><a id="script-parameters" 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="#script-parameters"><span><svg 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><!---->Параметры скрипта<!----></span></h2><!--]--><!----> <p>Скрипт обучения содержит множество параметров, позволяющих настроить процесс обучения. Все параметры и их описания находятся в функции <a href="https://github.com/huggingface/diffusers/blob/dd9a5caf61f04d11c0fa9f3947b69ab0010c9a0f/examples/text_to_image/train_text_to_image_lora.py#L85" rel="nofollow"><code>parse_args()</code></a>. Для большинства параметров заданы значения по умолчанию, которые хорошо работают, но при необходимости вы можете указать свои значения в команде запуска обучения.</p> <p>Например, чтобы увеличить количество эпох обучения:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-bash "><!---->accelerate launch train_text_to_image_lora.py \
  --num_train_epochs=150 \<!----></pre></div><!----> <p>Многие базовые и важные параметры описаны в руководстве по обучению <a href="text2image#script-parameters">Text-to-image</a>, поэтому в данном руководстве рассматриваются только параметры, относящиеся к LoRA:</p> <ul><li><code>--rank</code>: внутреннее измерение матриц низкого ранга, подлежащих обучению; чем выше ранг, тем больше обучаемых параметров</li> <li><code>--learning_rate</code>: скорость обучения по умолчанию составляет 1e-4, но при использовании LoRA можно применять более высокую скорость обучения</li></ul> <!--[1--><h2 class="relative group"><a id="training-script" 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-script"><span><svg 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><!---->Скрипт обучения<!----></span></h2><!--]--><!----> <p>Код предобработки набора данных и цикл обучения находятся в функции <a href="https://github.com/huggingface/diffusers/blob/dd9a5caf61f04d11c0fa9f3947b69ab0010c9a0f/examples/text_to_image/train_text_to_image_lora.py#L371" rel="nofollow"><code>main()</code></a>, и если вам нужно адаптировать скрипт обучения, именно здесь вы внесёте изменения.</p> <p>Как и в случае с параметрами скрипта, пошаговое описание скрипта обучения приведено в руководстве по обучению <a href="text2image#training-script">Text-to-image</a>. В данном руководстве рассматриваются только те части скрипта, которые относятся к LoRA.</p> <div class="flex space-x-2 items-center my-1.5 mr-8 h-7 !pl-0 -mx-3 md:mx-0"><!--[--><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd border-gray-800 bg-black dark:bg-gray-700 text-white">UNet</div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">text encoder</div><!--]--></div> <div class="language-select"><div><p>Diffusers использует <a href="https://huggingface.co/docs/peft/main/en/package_reference/lora#peft.LoraConfig" rel="nofollow">LoraConfig</a> из библиотеки <a href="https://hf.co/docs/peft" rel="nofollow">PEFT</a> для настройки параметров адаптера LoRA, таких как ранг, альфа и модули, в которые вставляются веса LoRA. Адаптер добавляется в UNet, и для оптимизации в <code>lora_layers</code> фильтруются только слои LoRA.</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-py "><!---->unet_lora_config = LoraConfig(
    r=args.rank,
    lora_alpha=args.rank,
    init_lora_weights=<span class="hljs-string">&quot;gaussian&quot;</span>,
    target_modules=[<span class="hljs-string">&quot;to_k&quot;</span>, <span class="hljs-string">&quot;to_q&quot;</span>, <span class="hljs-string">&quot;to_v&quot;</span>, <span class="hljs-string">&quot;to_out.0&quot;</span>],
)

unet.add_adapter(unet_lora_config)
lora_layers = <span class="hljs-built_in">filter</span>(<span class="hljs-keyword">lambda</span> p: p.requires_grad, unet.parameters())<!----></pre></div><!----><!----></div><!----> <div hidden=""><p>Diffusers также поддерживает дообучение текстового энкодера с помощью LoRA из библиотеки <a href="https://hf.co/docs/peft" rel="nofollow">PEFT</a> в тех случаях, когда это необходимо, например, при дообучении Stable Diffusion XL (SDXL). <a href="https://huggingface.co/docs/peft/main/en/package_reference/lora#peft.LoraConfig" rel="nofollow">LoraConfig</a> используется для настройки параметров адаптера LoRA, которые затем добавляются к текстовому энкодеру, и для обучения фильтруются только слои LoRA.</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-py "><!---->text_lora_config = LoraConfig(
    r=args.rank,
    lora_alpha=args.rank,
    init_lora_weights=<span class="hljs-string">&quot;gaussian&quot;</span>,
    target_modules=[<span class="hljs-string">&quot;q_proj&quot;</span>, <span class="hljs-string">&quot;k_proj&quot;</span>, <span class="hljs-string">&quot;v_proj&quot;</span>, <span class="hljs-string">&quot;out_proj&quot;</span>],
)

text_encoder_one.add_adapter(text_lora_config)
text_encoder_two.add_adapter(text_lora_config)
text_lora_parameters_one = <span class="hljs-built_in">list</span>(<span class="hljs-built_in">filter</span>(<span class="hljs-keyword">lambda</span> p: p.requires_grad, text_encoder_one.parameters()))
text_lora_parameters_two = <span class="hljs-built_in">list</span>(<span class="hljs-built_in">filter</span>(<span class="hljs-keyword">lambda</span> p: p.requires_grad, text_encoder_two.parameters()))<!----></pre></div><!----><!----></div><!----><!----></div><!----> <p><a href="https://github.com/huggingface/diffusers/blob/e4b8f173b97731686e290b2eb98e7f5df2b1b322/examples/text_to_image/train_text_to_image_lora.py#L529" rel="nofollow">Оптимизатор</a> инициализируется с <code>lora_layers</code>, поскольку именно эти веса подлежат оптимизации:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-py "><!---->optimizer = optimizer_cls(
    lora_layers,
    lr=args.learning_rate,
    betas=(args.adam_beta1, args.adam_beta2),
    weight_decay=args.adam_weight_decay,
    eps=args.adam_epsilon,
)<!----></pre></div><!----> <p>Aside from setting up the LoRA layers, the training script is more or less the same as train_text_to_image.py!</p> <!--[1--><h2 class="relative group"><a id="launch-the-script" 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="#launch-the-script"><span><svg 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><!---->Запустите скрипт<!----></span></h2><!--]--><!----> <p>Когда вы внесёте все изменения или решите оставить конфигурацию по умолчанию, можно запускать скрипт обучения! 🚀</p> <p>Обучимся на датасете <a href="https://huggingface.co/datasets/lambdalabs/naruto-blip-captions" rel="nofollow">Naruto BLIP captions</a>, чтобы генерировать собственных персонажей Наруто. Установите переменные окружения <code>MODEL_NAME</code> и <code>DATASET_NAME</code> для указания модели и датасета соответственно. Также укажите, куда сохранять модель, в <code>OUTPUT_DIR</code>, и имя модели для сохранения на Hub в <code>HUB_MODEL_ID</code>. Скрипт создаёт и сохраняет в ваш репозиторий следующие файлы:</p> <ul><li>сохранённые контрольные точки модели</li> <li><code>pytorch_lora_weights.safetensors</code> (обученные веса LoRA)</li></ul> <p>Если вы обучаете модель на нескольких GPU, добавьте параметр <code>--multi_gpu</code> к команде <code>accelerate launch</code>.</p> <blockquote class="warning"><p>Полный цикл обучения занимает около 5 часов на GPU 2080 Ti с 11 ГБ видеопамяти.</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-bash "><!----><span class="hljs-built_in">export</span> MODEL_NAME=<span class="hljs-string">&quot;stable-diffusion-v1-5/stable-diffusion-v1-5&quot;</span>
<span class="hljs-built_in">export</span> OUTPUT_DIR=<span class="hljs-string">&quot;/sddata/finetune/lora/naruto&quot;</span>
<span class="hljs-built_in">export</span> HUB_MODEL_ID=<span class="hljs-string">&quot;naruto-lora&quot;</span>
<span class="hljs-built_in">export</span> DATASET_NAME=<span class="hljs-string">&quot;lambdalabs/naruto-blip-captions&quot;</span>

accelerate launch --mixed_precision=<span class="hljs-string">&quot;fp16&quot;</span>  train_text_to_image_lora.py \
  --pretrained_model_name_or_path=<span class="hljs-variable">$MODEL_NAME</span> \
  --dataset_name=<span class="hljs-variable">$DATASET_NAME</span> \
  --dataloader_num_workers=8 \
  --resolution=512 \
  --center_crop \
  --random_flip \
  --train_batch_size=1 \
  --gradient_accumulation_steps=4 \
  --max_train_steps=15000 \
  --learning_rate=1e-04 \
  --max_grad_norm=1 \
  --lr_scheduler=<span class="hljs-string">&quot;cosine&quot;</span> \
  --lr_warmup_steps=0 \
  --output_dir=<span class="hljs-variable">${OUTPUT_DIR}</span> \
  --push_to_hub \
  --hub_model_id=<span class="hljs-variable">${HUB_MODEL_ID}</span> \
  --report_to=wandb \
  --checkpointing_steps=500 \
  --validation_prompt=<span class="hljs-string">&quot;A naruto with blue eyes.&quot;</span> \
  --seed=1337<!----></pre></div><!----> <p>После завершения обучения вы можете использовать свою модель для инференса:</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 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> Скопировано</div><!----></button><!----></div> <pre class="language-py "><!----><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> AutoPipelineForText2Image
<span class="hljs-keyword">import</span> torch

pipeline = AutoPipelineForText2Image.from_pretrained(<span class="hljs-string">&quot;stable-diffusion-v1-5/stable-diffusion-v1-5&quot;</span>, dtype=torch.float16).to(<span class="hljs-string">&quot;cuda&quot;</span>)  <span class="hljs-comment"># or &quot;mps&quot;, &quot;xpu&quot;, &quot;cpu&quot;</span>
pipeline.load_lora_weights(<span class="hljs-string">&quot;path/to/lora/model&quot;</span>, weight_name=<span class="hljs-string">&quot;pytorch_lora_weights.safetensors&quot;</span>)
image = pipeline(<span class="hljs-string">&quot;A naruto with blue eyes&quot;</span>).images[<span class="hljs-number">0</span>]<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="next-steps" 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="#next-steps"><span><svg 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><!---->Дальнейшие шаги<!----></span></h2><!--]--><!----> <p>Поздравляем с обучением новой модели с помощью LoRA! Чтобы узнать, как использовать новую модель, вам могут помочь следующие руководства:</p> <ul><li>Узнайте, как <a href="../tutorials/using_peft_for_inference">загружать различные форматы LoRA</a>, обученные с помощью комьюнити-тренажеров, таких как Kohya и TheLastBen.</li> <li>Узнайте, как использовать и <a href="../tutorials/using_peft_for_inference">комбинировать несколько LoRA</a> с PEFT для инференса.</li></ul> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/diffusers/blob/main/docs/source/en/training/lora.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><span class="underline">Update</span> on GitHub</span></a><!----> <p></p><!--]--><!----><!--]--><!--]--><!--]--> <!--[-1--><!--]--><!--]-->
			
			<script>
				{
					__sveltekit_g37pnb = {
						base: "/docs/diffusers/main/en",
						assets: "/docs/diffusers/main/en"
					};
					const element = document.currentScript.parentElement;
					Promise.all([
						import("/docs/diffusers/main/en/_app/immutable/entry/start.D3LnKZ4R.js"),
						import("/docs/diffusers/main/en/_app/immutable/entry/app.C0rBoMxR.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 318],
							data: [null,null],
							form: null,
							error: null
						});
					});
				}
			</script>
		
Xet Storage Details

Size:: 41.9 kB
Xet hash:: 60b925430459b9b77846b6ae7646cd5c057029aa69f4a46857d71aeb56e03481
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.