diffusers/pr_10312/en/quantization/gguf.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;GGUF&quot;,&quot;local&quot;:&quot;gguf&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Supported Quantization Types&quot;,&quot;local&quot;:&quot;supported-quantization-types&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/diffusers/pr_10312/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/entry/start.203b6290.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/scheduler.8c3d61f6.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/singletons.8b179f45.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/index.0997d446.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/paths.3fd58d56.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/entry/app.423ea23f.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/index.da70eac4.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/nodes/0.e544eae6.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/nodes/203.3094f5c5.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/CodeBlock.00a903b3.js">
		<link rel="modulepreload" href="/docs/diffusers/pr_10312/en/_app/immutable/chunks/EditOnGithub.1e64e623.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;GGUF&quot;,&quot;local&quot;:&quot;gguf&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Supported Quantization Types&quot;,&quot;local&quot;:&quot;supported-quantization-types&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>   <h1 class="relative group"><a id="gguf" 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="#gguf"><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>GGUF</span></h1> <p data-svelte-h="svelte-th49fe">The GGUF file format is typically used to store models for inference with <a href="https://github.com/ggerganov/ggml" rel="nofollow">GGML</a> and supports a variety of block wise quantization options. Diffusers supports loading checkpoints prequantized and saved in the GGUF format via <code>from_single_file</code> loading with Model classes. Loading GGUF checkpoints via Pipelines is currently not supported.</p> <p data-svelte-h="svelte-15l4qp1">The following example will load the <a href="https://huggingface.co/black-forest-labs/FLUX.1-dev" rel="nofollow">FLUX.1 DEV</a> transformer model using the GGUF Q2_K quantization variant.</p> <p data-svelte-h="svelte-1s5awbc">Before starting please install gguf in your environment</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 -U gguf<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1cgxlhc">Since GGUF is a single file format, use <code>~FromSingleFileMixin.from_single_file</code> to load the model and pass in the <a href="/docs/diffusers/pr_10312/en/api/quantization#diffusers.GGUFQuantizationConfig">GGUFQuantizationConfig</a>.</p> <p data-svelte-h="svelte-1tiy01a">When using GGUF checkpoints, the quantized weights remain in a low memory <code>dtype</code>(typically <code>torch.uint8</code>) and are dynamically dequantized and cast to the configured <code>compute_dtype</code> during each module’s forward pass through the model. The <code>GGUFQuantizationConfig</code> allows you to set the <code>compute_dtype</code>.</p> <p data-svelte-h="svelte-2xs1wc">The functions used for dynamic dequantizatation are based on the great work done by <a href="https://github.com/city96/ComfyUI-GGUF" rel="nofollow">city96</a>, who created the Pytorch ports of the original <a href="https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/gguf/quants.py" rel="nofollow"><code>numpy</code></a> implementation by <a href="https://github.com/compilade" rel="nofollow">compilade</a>.</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">import</span> torch

<span class="hljs-keyword">from</span> diffusers <span class="hljs-keyword">import</span> FluxPipeline, FluxTransformer2DModel, GGUFQuantizationConfig

ckpt_path = (
    <span class="hljs-string">&quot;https://huggingface.co/city96/FLUX.1-dev-gguf/blob/main/flux1-dev-Q2_K.gguf&quot;</span>
)
transformer = FluxTransformer2DModel.from_single_file(
    ckpt_path,
    quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
    torch_dtype=torch.bfloat16,
)
pipe = FluxPipeline.from_pretrained(
    <span class="hljs-string">&quot;black-forest-labs/FLUX.1-dev&quot;</span>,
    transformer=transformer,
    generator=torch.manual_seed(<span class="hljs-number">0</span>),
    torch_dtype=torch.bfloat16,
)
pipe.enable_model_cpu_offload()
prompt = <span class="hljs-string">&quot;A cat holding a sign that says hello world&quot;</span>
image = pipe(prompt).images[<span class="hljs-number">0</span>]
image.save(<span class="hljs-string">&quot;flux-gguf.png&quot;</span>)<!-- HTML_TAG_END --></pre></div>  <h2 class="relative group"><a id="supported-quantization-types" 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="#supported-quantization-types"><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>Supported Quantization Types</span></h2> <ul data-svelte-h="svelte-1dde1tf"><li>BF16</li> <li>Q4_0</li> <li>Q4_1</li> <li>Q5_0</li> <li>Q5_1</li> <li>Q8_0</li> <li>Q2_K</li> <li>Q3_K</li> <li>Q4_K</li> <li>Q5_K</li> <li>Q6_K</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/quantization/gguf.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_zvhs44 = {
						assets: "/docs/diffusers/pr_10312/en",
						base: "/docs/diffusers/pr_10312/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/diffusers/pr_10312/en/_app/immutable/entry/start.203b6290.js"),
						import("/docs/diffusers/pr_10312/en/_app/immutable/entry/app.423ea23f.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 203],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>