transformers/pr_36839/en/quantization/hqq.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;HQQ&quot;,&quot;local&quot;:&quot;hqq&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Backends&quot;,&quot;local&quot;:&quot;backends&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Resources&quot;,&quot;local&quot;:&quot;resources&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/transformers/pr_36839/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/entry/start.6be8d590.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/scheduler.01eeda35.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/singletons.177df05e.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/index.4862150a.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/paths.517376d1.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/entry/app.09748b4b.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/index.6dd51b66.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/nodes/0.8897c14d.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/nodes/442.5fea1b6c.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/CodeBlock.864da1b0.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/EditOnGithub.7faefd25.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/HfOption.f7f04550.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/stores.318eade7.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;HQQ&quot;,&quot;local&quot;:&quot;hqq&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Backends&quot;,&quot;local&quot;:&quot;backends&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Resources&quot;,&quot;local&quot;:&quot;resources&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>   <h1 class="relative group"><a id="hqq" 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="#hqq"><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>HQQ</span></h1> <p data-svelte-h="svelte-fwh8ox"><a href="https://github.com/mobiusml/hqq/" rel="nofollow">Half-Quadratic Quantization (HQQ)</a> supports fast on-the-fly quantization for 8, 4, 3, 2, and even 1-bits. It doesn’t require calibration data, and it is compatible with any model modality (LLMs, vision, etc.).</p> <p data-svelte-h="svelte-14m3csp">HQQ further supports fine-tuning with <a href="https://huggingface.co/docs/peft" rel="nofollow">PEFT</a> and is fully compatible with <a href="https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html" rel="nofollow">torch.compile</a> for even faster inference and training.</p> <p data-svelte-h="svelte-qgr4qn">Install HQQ with the following command to get the latest version and to build its corresponding CUDA kernels.</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 hqq<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1n7dug2">You can choose to either replace all the linear layers in a model with the same quantization config or dedicate a specific quantization config for specific linear layers.</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">replace all layers </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">specific layers only </div></div> <div class="language-select"><p data-svelte-h="svelte-1a5fi98">Quantize a model by creating a <a href="/docs/transformers/pr_36839/en/main_classes/quantization#transformers.HqqConfig">HqqConfig</a> and specifying the <code>nbits</code> and <code>group_size</code> to replace for all the linear layers (<a href="https://pytorch.org/docs/stable/generated/torch.nn.Linear.html" rel="nofollow">torch.nn.Linear</a>) of the model.</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> transformers <span class="hljs-keyword">import</span> AutoModelForCausalLM, AutoTokenizer, HqqConfig

quant_config = HqqConfig(nbits=<span class="hljs-number">8</span>, group_size=<span class="hljs-number">64</span>)
model = transformers.AutoModelForCausalLM.from_pretrained(
    <span class="hljs-string">&quot;meta-llama/Llama-3.1-8B&quot;</span>, 
    torch_dtype=torch.float16, 
    device_map=<span class="hljs-string">&quot;cuda&quot;</span>, 
    quantization_config=quant_config
)<!-- HTML_TAG_END --></pre></div> </div>  <h2 class="relative group"><a id="backends" 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="#backends"><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>Backends</span></h2> <p data-svelte-h="svelte-1fe8tjm">HQQ supports various backends, including pure PyTorch and custom dequantization CUDA kernels. These backends are suitable for older GPUs and PEFT/QLoRA training.</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> hqq.core.quantize <span class="hljs-keyword">import</span> *

HQQLinear.set_backend(HQQBackend.PYTORCH)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-4w4k3t">For faster inference, HQQ supports 4-bit fused kernels (torchao and Marlin) after a model is quantized. These can reach up to 200 tokens/sec on a single 4090. The example below demonstrates enabling the torchao_int4 backend.</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> hqq.utils.patching <span class="hljs-keyword">import</span> prepare_for_inference

prepare_for_inference(<span class="hljs-string">&quot;model&quot;</span>, backend=<span class="hljs-string">&quot;torchao_int4&quot;</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-h25g7y">Refer to the <a href="https://github.com/mobiusml/hqq/#backend" rel="nofollow">Backend</a> guide for more details.</p>  <h2 class="relative group"><a id="resources" 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="#resources"><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>Resources</span></h2> <p data-svelte-h="svelte-ubx7z0">Read the <a href="https://mobiusml.github.io/hqq_blog/" rel="nofollow">Half-Quadratic Quantization of Large Machine Learning Models</a> blog post for more details about HQQ.</p> <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/en/quantization/hqq.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_1bm5psi = {
						assets: "/docs/transformers/pr_36839/en",
						base: "/docs/transformers/pr_36839/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/transformers/pr_36839/en/_app/immutable/entry/start.6be8d590.js"),
						import("/docs/transformers/pr_36839/en/_app/immutable/entry/app.09748b4b.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 442],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>