diffusers/v0.18.2/en/optimization/mps.html

<meta charset="utf-8" /><meta http-equiv="content-security-policy" content=""><meta name="hf:doc:metadata" content="{&quot;local&quot;:&quot;how-to-use-stable-diffusion-in-apple-silicon-m1m2&quot;,&quot;sections&quot;:[{&quot;local&quot;:&quot;requirements&quot;,&quot;title&quot;:&quot;Requirements&quot;},{&quot;local&quot;:&quot;inference-pipeline&quot;,&quot;title&quot;:&quot;Inference Pipeline&quot;},{&quot;local&quot;:&quot;performance-recommendations&quot;,&quot;title&quot;:&quot;Performance Recommendations&quot;},{&quot;local&quot;:&quot;known-issues&quot;,&quot;title&quot;:&quot;Known Issues&quot;}],&quot;title&quot;:&quot;How to use Stable Diffusion in Apple Silicon (M1/M2)&quot;}" data-svelte="svelte-1phssyn">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/assets/pages/__layout.svelte-hf-doc-builder.css">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/start-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/chunks/vendor-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/chunks/paths-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/pages/__layout.svelte-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/pages/optimization/mps.mdx-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/chunks/Tip-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/chunks/IconCopyLink-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/diffusers/v0.18.2/en/_app/chunks/CodeBlock-hf-doc-builder.js"> 






<h1 class="relative group"><a id="how-to-use-stable-diffusion-in-apple-silicon-m1m2" 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="#how-to-use-stable-diffusion-in-apple-silicon-m1m2"><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>How to use Stable Diffusion in Apple Silicon (M1/M2)
	</span></h1>

<p>🤗 Diffusers is compatible with Apple silicon for Stable Diffusion inference, using the PyTorch <code>mps</code> device. These are the steps you need to follow to use your M1 or M2 computer with Stable Diffusion.</p>
<h2 class="relative group"><a id="requirements" 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="#requirements"><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>Requirements
	</span></h2>

<ul><li>Mac computer with Apple silicon (M1/M2) hardware.</li>
<li>macOS 12.6 or later (13.0 or later recommended).</li>
<li>arm64 version of Python.</li>
<li>PyTorch 2.0 (recommended) or 1.13 (minimum version supported for <code>mps</code>). You can install it with <code>pip</code> or <code>conda</code> using the instructions in <a href="https://pytorch.org/get-started/locally/" rel="nofollow">https://pytorch.org/get-started/locally/</a>.</li></ul>
<h2 class="relative group"><a id="inference-pipeline" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#inference-pipeline"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a>
	<span>Inference Pipeline
	</span></h2>

<p>The snippet below demonstrates how to use the <code>mps</code> backend using the familiar <code>to()</code> interface to move the Stable Diffusion pipeline to your M1 or M2 device.</p>


<div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p><strong>If you are using PyTorch 1.13</strong> you need to “prime” the pipeline using an additional one-time pass through it. This is a temporary workaround for a weird issue we detected: the first inference pass produces slightly different results than subsequent ones. You only need to do this pass once, and it’s ok to use just one inference step and discard the result.</p></div>
<p>We strongly recommend you use PyTorch 2 or better, as it solves a number of problems like the one described in the previous tip.</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><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained(<span class="hljs-string">&quot;runwayml/stable-diffusion-v1-5&quot;</span>)
pipe = pipe.to(<span class="hljs-string">&quot;mps&quot;</span>)

<span class="hljs-comment"># Recommended if your computer has &lt; 64 GB of RAM</span>
pipe.enable_attention_slicing()

prompt = <span class="hljs-string">&quot;a photo of an astronaut riding a horse on mars&quot;</span>

<span class="hljs-comment"># First-time &quot;warmup&quot; pass if PyTorch version is 1.13 (see explanation above)</span>
_ = pipe(prompt, num_inference_steps=<span class="hljs-number">1</span>)

<span class="hljs-comment"># Results match those from the CPU device after the warmup pass.</span>
image = pipe(prompt).images[<span class="hljs-number">0</span>]<!-- HTML_TAG_END --></pre></div>
<h2 class="relative group"><a id="performance-recommendations" 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="#performance-recommendations"><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>Performance Recommendations
	</span></h2>

<p>M1/M2 performance is very sensitive to memory pressure. The system will automatically swap if it needs to, but performance will degrade significantly when it does.</p>
<p>We recommend you use <em>attention slicing</em> to reduce memory pressure during inference and prevent swapping, particularly if your computer has less than 64 GB of system RAM, or if you generate images at non-standard resolutions larger than 512 × 512 pixels. Attention slicing performs the costly attention operation in multiple steps instead of all at once. It usually has a performance impact of ~20% in computers without universal memory, but we have observed <em>better performance</em> in most Apple Silicon computers, unless you have 64 GB or more.</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><!-- HTML_TAG_START -->pipeline.enable_attention_slicing()<!-- HTML_TAG_END --></pre></div>
<h2 class="relative group"><a id="known-issues" 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="#known-issues"><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>Known Issues
	</span></h2>

<ul><li>Generating multiple prompts in a batch <a href="https://github.com/huggingface/diffusers/issues/363" rel="nofollow">crashes or doesn’t work reliably</a>. We believe this is related to the <a href="https://github.com/pytorch/pytorch/issues/84039" rel="nofollow"><code>mps</code> backend in PyTorch</a>. This is being resolved, but for now we recommend to iterate instead of batching.</li></ul>


		<script type="module" data-hydrate="nmbuoz">
		import { start } from "/docs/diffusers/v0.18.2/en/_app/start-hf-doc-builder.js";
		start({
			target: document.querySelector('[data-hydrate="nmbuoz"]').parentNode,
			paths: {"base":"/docs/diffusers/v0.18.2/en","assets":"/docs/diffusers/v0.18.2/en"},
			session: {},
			route: false,
			spa: false,
			trailing_slash: "never",
			hydrate: {
				status: 200,
				error: null,
				nodes: [
					import("/docs/diffusers/v0.18.2/en/_app/pages/__layout.svelte-hf-doc-builder.js"),
						import("/docs/diffusers/v0.18.2/en/_app/pages/optimization/mps.mdx-hf-doc-builder.js")
				],
				params: {}
			}
		});
	</script>