Buckets:

hf-doc-build
/

doc-dev

Files

xet

hf-doc-build/doc-dev / inference-endpoints /pr_118 /en /frameworks /llama_cpp.html

rtrm

about 1 month ago

download

raw

18.1 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"llama.cpp","local":"llamacpp","sections":[{"title":"Deployment Steps","local":"deployment-steps","sections":[],"depth":2},{"title":"Configurations","local":"configurations","sections":[{"title":"Basic Configurations","local":"basic-configurations","sections":[],"depth":3},{"title":"Advanced Configurations","local":"advanced-configurations","sections":[],"depth":3}],"depth":2},{"title":"Troubleshooting","local":"troubleshooting","sections":[],"depth":2}],"depth":1}">
	<link href="/docs/inference-endpoints/pr_118/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/entry/start.9b6bb78f.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/chunks/scheduler.389d799c.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/chunks/singletons.93749600.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/chunks/paths.72a954e9.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/entry/app.f206cea9.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/chunks/index.8f81d18f.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/nodes/0.1525eca1.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/nodes/5.52527c71.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/chunks/CodeBlock.c0898180.js">
	<link rel="modulepreload" href="/docs/inference-endpoints/pr_118/en/_app/immutable/chunks/getInferenceSnippets.4dd433fd.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"llama.cpp","local":"llamacpp","sections":[{"title":"Deployment Steps","local":"deployment-steps","sections":[],"depth":2},{"title":"Configurations","local":"configurations","sections":[{"title":"Basic Configurations","local":"basic-configurations","sections":[],"depth":3},{"title":"Advanced Configurations","local":"advanced-configurations","sections":[],"depth":3}],"depth":2},{"title":"Troubleshooting","local":"troubleshooting","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="llamacpp" 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="#llamacpp"><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>llama.cpp</span></h1> <p data-svelte-h="svelte-14uwew1">You can deploy any llama.cpp compatible GGUF on the Hugging Face Endpoints. When you create an endpoint with a <a href="https://huggingface.co/docs/hub/en/gguf" rel="nofollow">GGUF</a> model, a <a href="https://github.com/ggerganov/llama.cpp" rel="nofollow">llama.cpp</a> container is automatically selected using the latest image built from the <code>master</code> branch of the llama.cpp repository. Upon successful deployment, a server with an OpenAI-compatible endpoint becomes available.</p> <p data-svelte-h="svelte-4ukih7">Llama.cpp supports multiple endpoints like <code>/tokenize</code>, <code>/health</code>, <code>/embedding</code> and many more. For a comprehensive list of available endpoints, please refer to the <a href="https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md#api-endpoints" rel="nofollow">API documentation</a>.</p> <h2 class="relative group"><a id="deployment-steps" 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="#deployment-steps"><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>Deployment Steps</span></h2> <p data-svelte-h="svelte-1viuqt7">To deploy an endpoint with a llama.cpp container, follow these steps:</p> <ol data-svelte-h="svelte-zqz83d"><li><a href="./create_endpoint">Create a new endpoint</a> and select a repository containing a GGUF model. The llama.cpp container will be automatically selected.</li></ol> <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/endpoints/llamacpp_1.png" alt="Select model"> <ol start="2" data-svelte-h="svelte-a3ltnp"><li>Choose the desired GGUF file, noting that memory requirements will vary depending on the selected file. For example, an F16 model requires more memory than a Q4_K_M model.</li></ol> <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/endpoints/llamacpp_2.png" alt="Select GGUF file"> <ol start="3" data-svelte-h="svelte-1mdwz6y"><li>Select your desired hardware configuration.</li></ol> <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/endpoints/llamacpp_3.png" alt="Select hardware"> <ol start="4" data-svelte-h="svelte-1vuz66q"><li><p>Optionally, you can customize the container’s configuration settings like <code>Max Tokens</code>, <code>Number of Concurrent Requests</code>. For more information on those, please refer to the <strong>Configurations</strong> section below.</p></li> <li><p>Click the <strong>Create Endpoint</strong> button to complete the deployment.</p></li></ol> <p data-svelte-h="svelte-174w0va">Alternatively, you can follow the video tutorial below for a step-by-step guide on deploying an endpoint with a llama.cpp container:</p> <video width="1280" height="720" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/endpoints/llamacpp_guide_video.mp4" controls="true"></video> <h2 class="relative group"><a id="configurations" 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="#configurations"><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>Configurations</span></h2> <p data-svelte-h="svelte-jdgis4">The llama.cpp container offers several configuration options that can be adjusted. After deployment, you can modify these settings by accessing the <strong>Settings</strong> tab on the endpoint details page.</p> <h3 class="relative group"><a id="basic-configurations" 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="#basic-configurations"><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>Basic Configurations</span></h3> <ul data-svelte-h="svelte-d1iatn"><li><strong>Max Tokens (per Request)</strong>: The maximum number of tokens that can be sent in a single request.</li> <li><strong>Max Concurrent Requests</strong>: The maximum number of concurrent requests allowed for this deployment. Increasing this limit requires additional memory allocation.
	For instance, setting this value to 4 requests with 1024 tokens maximum per request requires memory capacity for 4096 tokens in total.</li></ul> <h3 class="relative group"><a id="advanced-configurations" 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="#advanced-configurations"><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>Advanced Configurations</span></h3> <p data-svelte-h="svelte-io22b7">In addition to the basic configurations, you can also modify specific settings by setting environment variables. A list of available environment variables can be found in the <a href="https://github.com/ggerganov/llama.cpp/blob/master/examples/server/README.md#usage" rel="nofollow">API documentation</a>.</p> <p data-svelte-h="svelte-omzti">Please note that the following environment variables are reserved by the system and cannot be modified:</p> <ul data-svelte-h="svelte-apzx5r"><li><code>LLAMA_ARG_MODEL</code></li> <li><code>LLAMA_ARG_HTTP_THREADS</code></li> <li><code>LLAMA_ARG_N_GPU_LAYERS</code></li> <li><code>LLAMA_ARG_EMBEDDINGS</code></li> <li><code>LLAMA_ARG_HOST</code></li> <li><code>LLAMA_ARG_PORT</code></li> <li><code>LLAMA_ARG_NO_MMAP</code></li> <li><code>LLAMA_ARG_CTX_SIZE</code></li> <li><code>LLAMA_ARG_N_PARALLEL</code></li> <li><code>LLAMA_ARG_ENDPOINT_METRICS</code></li></ul> <h2 class="relative group"><a id="troubleshooting" 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="#troubleshooting"><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>Troubleshooting</span></h2> <p data-svelte-h="svelte-sywduw">In case the deployment fails, please watch the log output for any error messages.</p> <p data-svelte-h="svelte-dnupoj">You can access the logs by clicking on the <strong>Logs</strong> tab on the endpoint details page. To learn more, refer to the <a href="./logs">Logs</a> documentation.</p> <ul><li><p data-svelte-h="svelte-14vukni"><strong>Malloc failed: out of memory</strong><br>
	If you see this error message in the log:</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 -->ggml_backend_cuda_buffer_type_alloc_buffer: allocating <span class="hljs-number">67200.00</span> MiB <span class="hljs-keyword">on</span> device <span class="hljs-number">0</span>: cuda
	Malloc failed: out of memory
	llama_kv_cache_init: failed <span class="hljs-keyword">to</span> allocate buffer for kv <span class="hljs-keyword">cache</span>
	llama_new_context_with_model: llama_kv_cache_init() failed for <span class="hljs-built_in">self</span><span class="hljs-params">-attention</span> <span class="hljs-keyword">cache</span>
	<span class="hljs-params">...</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ygfqn1">That means the selected hardware configuration does not have enough memory to accommodate the selected GGUF model. You can try to:</p> <ul data-svelte-h="svelte-7cq89u"><li>Lower the number of maximum tokens per request</li> <li>Lower the number of concurrent requests</li> <li>Select a smaller GGUF model</li> <li>Select a larger hardware configuration</li></ul></li> <li data-svelte-h="svelte-9ofzm8"><p><strong>Workload evicted, storage limit exceeded</strong><br>
	This error message indicates that the hardware has too little memory to accommodate the selected GGUF model. Try selecting a smaller model or select a larger hardware configuration.</p></li> <li data-svelte-h="svelte-xbgvsk"><p><strong>Other problems</strong><br>
	For other problems, please refer to the <a href="https://github.com/ggerganov/llama.cpp/issues" rel="nofollow">llama.cpp issues page</a>. In case you want to create a new issue, please also include the full log output in your bug report.</p></li></ul> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/hf-endpoints-documentation/blob/main/docs/source/frameworks/llama_cpp.mdx" target="_blank"><span data-svelte-h="svelte-1kd6by1"><</span> <span data-svelte-h="svelte-x0xyl0">></span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p>

	<script>
	{
	__sveltekit_1ym9hhx = {
	assets: "/docs/inference-endpoints/pr_118/en",
	base: "/docs/inference-endpoints/pr_118/en",
	env: {}
	};

	const element = document.currentScript.parentElement;

	const data = [null,null];

	Promise.all([
	import("/docs/inference-endpoints/pr_118/en/_app/immutable/entry/start.9b6bb78f.js"),
	import("/docs/inference-endpoints/pr_118/en/_app/immutable/entry/app.f206cea9.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 5],
	data,
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 18.1 kB
Xet hash:: c5835e5808889d200f6995925cb8247374e9ae56d0048ad34a0e3f574c02bde5

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.