diffusers/main/en/optimization/habana.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Habana Gaudi&quot;,&quot;local&quot;:&quot;habana-gaudi&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Benchmark&quot;,&quot;local&quot;:&quot;benchmark&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/diffusers/main/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/entry/start.21e27d66.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/scheduler.8c3d61f6.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/singletons.1db06f6d.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/index.0997d446.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/paths.085e8bc8.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/entry/app.de4fb612.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/index.da70eac4.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/nodes/0.f6117ae5.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/nodes/169.1db72ff7.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/CodeBlock.00a903b3.js">
		<link rel="modulepreload" href="/docs/diffusers/main/en/_app/immutable/chunks/EditOnGithub.1e64e623.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Habana Gaudi&quot;,&quot;local&quot;:&quot;habana-gaudi&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Benchmark&quot;,&quot;local&quot;:&quot;benchmark&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>   <h1 class="relative group"><a id="habana-gaudi" 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="#habana-gaudi"><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>Habana Gaudi</span></h1> <p data-svelte-h="svelte-zkz97r">🤗 Diffusers is compatible with Habana Gaudi through 🤗 <a href="https://huggingface.co/docs/optimum/habana/usage_guides/stable_diffusion" rel="nofollow">Optimum</a>. Follow the <a href="https://docs.habana.ai/en/latest/Installation_Guide/index.html" rel="nofollow">installation</a> guide to install the SynapseAI and Gaudi drivers, and then install Optimum Habana:</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 -m pip install --upgrade-strategy eager optimum[habana]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ojl4a9">To generate images with Stable Diffusion 1 and 2 on Gaudi, you need to instantiate two instances:</p> <ul data-svelte-h="svelte-1x8zh9o"><li><code>GaudiStableDiffusionPipeline</code>, a pipeline for text-to-image generation.</li> <li><code>GaudiDDIMScheduler</code>, a Gaudi-optimized scheduler.</li></ul> <p data-svelte-h="svelte-1ncqr39">When you initialize the pipeline, you have to specify <code>use_habana=True</code> to deploy it on HPUs and to get the fastest possible generation, you should enable <strong>HPU graphs</strong> with <code>use_hpu_graphs=True</code>.</p> <p data-svelte-h="svelte-orl9jn">Finally, specify a <code>GaudiConfig</code> which can be downloaded from the <a href="https://huggingface.co/Habana" rel="nofollow">Habana</a> organization on the Hub.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> optimum.habana <span class="hljs-keyword">import</span> GaudiConfig
<span class="hljs-keyword">from</span> optimum.habana.diffusers <span class="hljs-keyword">import</span> GaudiDDIMScheduler, GaudiStableDiffusionPipeline

model_name = <span class="hljs-string">&quot;stabilityai/stable-diffusion-2-base&quot;</span>
scheduler = GaudiDDIMScheduler.from_pretrained(model_name, subfolder=<span class="hljs-string">&quot;scheduler&quot;</span>)
pipeline = GaudiStableDiffusionPipeline.from_pretrained(
    model_name,
    scheduler=scheduler,
    use_habana=<span class="hljs-literal">True</span>,
    use_hpu_graphs=<span class="hljs-literal">True</span>,
    gaudi_config=<span class="hljs-string">&quot;Habana/stable-diffusion-2&quot;</span>,
)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-g0h5z1">Now you can call the pipeline to generate images by batches from one or several prompts:</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 -->outputs = pipeline(
    prompt=[
        <span class="hljs-string">&quot;High quality photo of an astronaut riding a horse in space&quot;</span>,
        <span class="hljs-string">&quot;Face of a yellow cat, high resolution, sitting on a park bench&quot;</span>,
    ],
    num_images_per_prompt=<span class="hljs-number">10</span>,
    batch_size=<span class="hljs-number">4</span>,
)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1xk941i">For more information, check out 🤗 Optimum Habana’s <a href="https://huggingface.co/docs/optimum/habana/usage_guides/stable_diffusion" rel="nofollow">documentation</a> and the <a href="https://github.com/huggingface/optimum-habana/tree/main/examples/stable-diffusion" rel="nofollow">example</a> provided in the official GitHub repository.</p>  <h2 class="relative group"><a id="benchmark" 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="#benchmark"><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>Benchmark</span></h2> <p data-svelte-h="svelte-whvrtk">We benchmarked Habana’s first-generation Gaudi and Gaudi2 with the <a href="https://huggingface.co/Habana/stable-diffusion" rel="nofollow">Habana/stable-diffusion</a> and <a href="https://huggingface.co/Habana/stable-diffusion-2" rel="nofollow">Habana/stable-diffusion-2</a> Gaudi configurations (mixed precision bf16/fp32) to demonstrate their performance.</p> <p data-svelte-h="svelte-v3u7rq">For <a href="https://huggingface.co/runwayml/stable-diffusion-v1-5" rel="nofollow">Stable Diffusion v1.5</a> on 512x512 images:</p> <table data-svelte-h="svelte-1d97u7g"><thead><tr><th></th> <th align="center">Latency (batch size = 1)</th> <th align="center">Throughput</th></tr></thead> <tbody><tr><td>first-generation Gaudi</td> <td align="center">3.80s</td> <td align="center">0.308 images/s (batch size = 8)</td></tr> <tr><td>Gaudi2</td> <td align="center">1.33s</td> <td align="center">1.081 images/s (batch size = 8)</td></tr></tbody></table> <p data-svelte-h="svelte-1pi2k0u">For <a href="https://huggingface.co/stabilityai/stable-diffusion-2-1" rel="nofollow">Stable Diffusion v2.1</a> on 768x768 images:</p> <table data-svelte-h="svelte-1uvfcpf"><thead><tr><th></th> <th align="center">Latency (batch size = 1)</th> <th align="center">Throughput</th></tr></thead> <tbody><tr><td>first-generation Gaudi</td> <td align="center">10.2s</td> <td align="center">0.108 images/s (batch size = 4)</td></tr> <tr><td>Gaudi2</td> <td align="center">3.17s</td> <td align="center">0.379 images/s (batch size = 8)</td></tr></tbody></table> <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/optimization/habana.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_1eu7tzz = {
						assets: "/docs/diffusers/main/en",
						base: "/docs/diffusers/main/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/diffusers/main/en/_app/immutable/entry/start.21e27d66.js"),
						import("/docs/diffusers/main/en/_app/immutable/entry/app.de4fb612.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 169],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>