bitsandbytes/main/en/nonpytorchcuda.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;How to use a CUDA version that is different from PyTorch&quot;,&quot;local&quot;:&quot;how-to-use-a-cuda-version-that-is-different-from-pytorch&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Installing or determining the CUDA installation&quot;,&quot;local&quot;:&quot;installing-or-determining-the-cuda-installation&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Setting the environmental variables BNB_CUDA_VERSION , and LD_LIBRARY_PATH&quot;,&quot;local&quot;:&quot;setting-the-environmental-variables-bnbcudaversion--and-ldlibrarypath&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/bitsandbytes/main/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/entry/start.908aed57.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/scheduler.0f20f9e5.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/singletons.8ebd9a37.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/index.e6666749.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/paths.63876dae.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/entry/app.5720ff48.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/index.fbd8b8e3.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/nodes/0.68949aec.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/nodes/10.a5d63a67.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/CodeBlock.c14486c2.js">
		<link rel="modulepreload" href="/docs/bitsandbytes/main/en/_app/immutable/chunks/Heading.2451f3bb.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;How to use a CUDA version that is different from PyTorch&quot;,&quot;local&quot;:&quot;how-to-use-a-cuda-version-that-is-different-from-pytorch&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Installing or determining the CUDA installation&quot;,&quot;local&quot;:&quot;installing-or-determining-the-cuda-installation&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Setting the environmental variables BNB_CUDA_VERSION , and LD_LIBRARY_PATH&quot;,&quot;local&quot;:&quot;setting-the-environmental-variables-bnbcudaversion--and-ldlibrarypath&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>  <h1 class="relative group"><a id="how-to-use-a-cuda-version-that-is-different-from-pytorch" 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-a-cuda-version-that-is-different-from-pytorch"><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 a CUDA version that is different from PyTorch</span></h1> <p data-svelte-h="svelte-3za3u7">Some features of <code>bitsandbytes</code> may need a newer CUDA version than regularly supported by PyTorch binaries from conda / pip. In that case you can use the following instructions to load a precompiled <code>bitsandbytes</code> binary that works for you.</p>  <h2 class="relative group"><a id="installing-or-determining-the-cuda-installation" 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="#installing-or-determining-the-cuda-installation"><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>Installing or determining the CUDA installation</span></h2> <p data-svelte-h="svelte-1qsfu71">Determine the path of the CUDA version that you want to use. Common paths paths are:</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 -->/usr/local/cuda
/usr/local/cuda-XX.X<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-hrid8g">where XX.X is the CUDA version number.</p> <p data-svelte-h="svelte-hwgjn">You can also install CUDA version that you need locally with a script provided by <code>bitsandbytes</code> as follows:</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 -->wget https://raw.githubusercontent.com/TimDettmers/bitsandbytes/main/install_cuda.sh
<span class="hljs-comment"># Syntax cuda_install CUDA_VERSION INSTALL_PREFIX EXPORT_TO_BASH</span>
<span class="hljs-comment">#   CUDA_VERSION in {110, 111, 112, 113, 114, 115, 116, 117, 118, 120, 121, 122}</span>
<span class="hljs-comment">#   EXPORT_TO_BASH in {0, 1} with 0=False and 1=True</span>

<span class="hljs-comment"># For example, the following installs CUDA 11.7 to ~/local/cuda-11.7 and exports the path to your .bashrc</span>

bash cuda_install.sh 117 ~/local 1<!-- HTML_TAG_END --></pre></div>  <h2 class="relative group"><a id="setting-the-environmental-variables-bnbcudaversion--and-ldlibrarypath" 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="#setting-the-environmental-variables-bnbcudaversion--and-ldlibrarypath"><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>Setting the environmental variables BNB_CUDA_VERSION , and LD_LIBRARY_PATH</span></h2> <p data-svelte-h="svelte-19qigbx">To manually override the PyTorch installed CUDA version you need to set to variable, like so:</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-built_in">export</span> BNB_CUDA_VERSION=&lt;VERSION&gt;
<span class="hljs-built_in">export</span> LD_LIBRARY_PATH=<span class="hljs-variable">$LD_LIBRARY_PATH</span>:&lt;PATH&gt;<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1s8m2s6">For example, to use the local install path from above:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-built_in">export</span> BNB_CUDA_VERSION=117
<span class="hljs-built_in">export</span> LD_LIBRARY_PATH=<span class="hljs-variable">$LD_LIBRARY_PATH</span>:/home/tim/local/cuda-11.7<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-tbg1n0">It is best to add these lines to the <code>.bashrc</code> file to make them permanent.</p> <p data-svelte-h="svelte-1geu9m0">If you now launch bitsandbytes with these environmental variables the PyTorch CUDA version will be overridden by the new CUDA version and a different bitsandbytes library is loaded (in this case version 117).</p>  <p></p> 
			
			<script>
				{
					__sveltekit_1gczcdv = {
						assets: "/docs/bitsandbytes/main/en",
						base: "/docs/bitsandbytes/main/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/bitsandbytes/main/en/_app/immutable/entry/start.908aed57.js"),
						import("/docs/bitsandbytes/main/en/_app/immutable/entry/app.5720ff48.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 10],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>