trl/pr_3395/en/vllm_integration.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;vLLM Integration&quot;,&quot;local&quot;:&quot;vllm-integration&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;TRL vLLM server&quot;,&quot;local&quot;:&quot;trl-vllm-server&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Find the best distributed setup&quot;,&quot;local&quot;:&quot;find-the-best-distributed-setup&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/trl/pr_3395/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/entry/start.5a7d6017.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/scheduler.d627b047.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/singletons.37ee724a.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/index.a57a1c33.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/paths.6eab44a0.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/entry/app.d9770f1a.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/index.73c51727.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/nodes/0.7922892a.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/nodes/49.3335feed.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/Tip.a82942ec.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/CodeBlock.5f78c87f.js">
		<link rel="modulepreload" href="/docs/trl/pr_3395/en/_app/immutable/chunks/index.8df1db42.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;vLLM Integration&quot;,&quot;local&quot;:&quot;vllm-integration&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;TRL vLLM server&quot;,&quot;local&quot;:&quot;trl-vllm-server&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Find the best distributed setup&quot;,&quot;local&quot;:&quot;find-the-best-distributed-setup&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>  <h1 class="relative group"><a id="vllm-integration" 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="#vllm-integration"><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>vLLM Integration</span></h1>  <div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p data-svelte-h="svelte-txx63b">Section under construction. Feel free to contribute!</p></div>  <h2 class="relative group"><a id="trl-vllm-server" 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="#trl-vllm-server"><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>TRL vLLM server</span></h2> <p data-svelte-h="svelte-1b1y5j5">TRL provides a way to speedup generation using a dedicated vLLM server.</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 -->$ trl vllm-serve --<span class="hljs-built_in">help</span>
usage: trl vllm-serve [-h] --model MODEL [--revision REVISION] [--tensor_parallel_size TENSOR_PARALLEL_SIZE] [--host HOST]
                      [--port PORT] [--gpu_memory_utilization GPU_MEMORY_UTILIZATION] [--dtype DTYPE]
                      [--max_model_len MAX_MODEL_LEN] [--enable_prefix_caching ENABLE_PREFIX_CACHING]

options:
  -h, --<span class="hljs-built_in">help</span>            Show this <span class="hljs-built_in">help</span> message and <span class="hljs-built_in">exit</span>
  --model MODEL         Model name or path to load the model from. (default: None)
  --revision REVISION   Revision to use <span class="hljs-keyword">for</span> the model. If not specified, the default branch will be used. (default: None)
  --tensor_parallel_size TENSOR_PARALLEL_SIZE, --tensor-parallel-size TENSOR_PARALLEL_SIZE
                        Number of tensor parallel workers to use. (default: 1)
  --host HOST           Host address to run the server on. (default: 0.0.0.0)
  --port PORT           Port to run the server on. (default: 8000)
  --gpu_memory_utilization GPU_MEMORY_UTILIZATION, --gpu-memory-utilization GPU_MEMORY_UTILIZATION
                        Ratio (between 0 and 1) of GPU memory to reserve <span class="hljs-keyword">for</span> the model weights, activations, and KV cache on the device
                        dedicated to generation powered by vLLM. Higher values will increase the KV cache size and thus improve the
                        model<span class="hljs-string">&#x27;s throughput. However, if the value is too high, it may cause out-of-memory (OOM) errors during
                        initialization. (default: 0.9)
  --dtype DTYPE         Data type to use for vLLM generation. If set to &#x27;</span>auto<span class="hljs-string">&#x27;, the data type will be automatically determined based on
                        the model configuration. Find the supported values in the vLLM documentation. (default: auto)
  --max_model_len MAX_MODEL_LEN, --max-model-len MAX_MODEL_LEN
                        If set, the `max_model_len` to use for vLLM. This can be useful when running with reduced
                        `vllm_gpu_memory_utilization`, leading to a reduced KV cache size. If not set, vLLM will use the model context
                        size, which might be much larger than the KV cache, leading to inefficiencies. (default: None)
  --enable_prefix_caching ENABLE_PREFIX_CACHING, --enable-prefix-caching ENABLE_PREFIX_CACHING
                        Whether to enable prefix caching in vLLM. If set to `True`, ensure that the model and the hardware support this
                        feature. (default: None)</span><!-- HTML_TAG_END --></pre></div>  <h3 class="relative group"><a id="find-the-best-distributed-setup" 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="#find-the-best-distributed-setup"><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>Find the best distributed setup</span></h3> <p data-svelte-h="svelte-gecap4"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/tp_dp_throughput_8_gpus.png"></p> <p data-svelte-h="svelte-rp1uk4"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/tp_dp_throughput_4_gpus.png"></p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/trl/blob/main/docs/source/vllm_integration.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_13ckwie = {
						assets: "/docs/trl/pr_3395/en",
						base: "/docs/trl/pr_3395/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/trl/pr_3395/en/_app/immutable/entry/start.5a7d6017.js"),
						import("/docs/trl/pr_3395/en/_app/immutable/entry/app.d9770f1a.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 49],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>