trl/pr_3582/en/distributing_training.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Distributing Training&quot;,&quot;local&quot;:&quot;distributing-training&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Multi-GPU Training with TRL&quot;,&quot;local&quot;:&quot;multi-gpu-training-with-trl&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Multi-Nodes Training&quot;,&quot;local&quot;:&quot;multi-nodes-training&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/trl/pr_3582/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/entry/start.0f0f318c.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/scheduler.d627b047.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/singletons.affb0d47.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/index.a57a1c33.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/paths.15dc14db.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/entry/app.b27a462f.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/index.73c51727.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/nodes/0.8cd8e450.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/nodes/15.6f6eb875.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/Tip.a82942ec.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/CodeBlock.5f78c87f.js">
		<link rel="modulepreload" href="/docs/trl/pr_3582/en/_app/immutable/chunks/getInferenceSnippets.256dfbf1.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Distributing Training&quot;,&quot;local&quot;:&quot;distributing-training&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Multi-GPU Training with TRL&quot;,&quot;local&quot;:&quot;multi-gpu-training-with-trl&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Multi-Nodes Training&quot;,&quot;local&quot;:&quot;multi-nodes-training&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>  <h1 class="relative group"><a id="distributing-training" 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="#distributing-training"><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>Distributing Training</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">Section under construction. Feel free to contribute!</div>  <h2 class="relative group"><a id="multi-gpu-training-with-trl" 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="#multi-gpu-training-with-trl"><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>Multi-GPU Training with TRL</span></h2> <p data-svelte-h="svelte-1awbvnz">The trainers in TRL use <a href="https://github.com/huggingface/accelerate" rel="nofollow">🤗 Accelerate</a> to enable distributed training across multiple GPUs or nodes. To do so, first create an <a href="https://github.com/huggingface/accelerate" rel="nofollow">🤗 Accelerate</a> config file by running</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 -->accelerate config<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-n2kdl3">and answering the questions according to your multi-GPU / multi-node setup. You can then launch distributed training by running:</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 -->accelerate launch train.py<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-15pg0q3">We also provide config files in the <a href="https://github.com/huggingface/trl/tree/main/examples/accelerate_configs" rel="nofollow">examples folder</a> that can be used as templates. To use these templates, simply pass the path to the config file when launching a job, e.g.:</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 -->accelerate launch --config_file examples/accelerate_configs/multi_gpu.yaml train.py &lt;SCRIPT_ARGS&gt;<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1n2xjt">This automatically distributes the workload across all available GPUs.</p> <p data-svelte-h="svelte-zfygwf">Under the hood, <a href="https://github.com/huggingface/accelerate" rel="nofollow">🤗 Accelerate</a> creates one model per GPU. Each process:</p> <ul data-svelte-h="svelte-3s441m"><li>Processes its own batch of data</li> <li>Computes the loss and gradients for that batch</li> <li>Shares gradient updates across all GPUs</li></ul> <p data-svelte-h="svelte-qt5st7"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/multi_gpu.png"></p> <p>The effective batch size is calculated as:
<!-- HTML_TAG_START --><span class="katex-display"><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML" display="block"><semantics><mrow><mtext>Batch Size</mtext><mo>=</mo><mtext>per_device_train_batch_size</mtext><mo>×</mo><mtext>num_devices</mtext><mo>×</mo><mtext>gradient_accumulation_steps</mtext></mrow><annotation encoding="application/x-tex">
\text{Batch Size} = \text{per\_device\_train\_batch\_size} \times \text{num\_devices} \times \text{gradient\_accumulation\_steps}
</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.6944em;"></span><span class="mord text"><span class="mord">Batch Size</span></span><span class="mspace" style="margin-right:0.2778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2778em;"></span></span><span class="base"><span class="strut" style="height:1.0044em;vertical-align:-0.31em;"></span><span class="mord text"><span class="mord">per_device_train_batch_size</span></span><span class="mspace" style="margin-right:0.2222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222em;"></span></span><span class="base"><span class="strut" style="height:1.0044em;vertical-align:-0.31em;"></span><span class="mord text"><span class="mord">num_devices</span></span><span class="mspace" style="margin-right:0.2222em;"></span><span class="mbin">×</span><span class="mspace" style="margin-right:0.2222em;"></span></span><span class="base"><span class="strut" style="height:1.0044em;vertical-align:-0.31em;"></span><span class="mord text"><span class="mord">gradient_accumulation_steps</span></span></span></span></span></span><!-- HTML_TAG_END --></p> <p data-svelte-h="svelte-19d9mn0">To maintain a consistent batch size when scaling to multiple GPUs, make sure to update <code>per_device_train_batch_size</code> and <code>gradient_accumulation_steps</code> accordingly.</p> <p data-svelte-h="svelte-1npf4a2">Example, these configurations are equivalent, and should yield the same results:</p> <table data-svelte-h="svelte-nztwgr"><thead><tr><th>Number of GPUs</th> <th>Per device batch size</th> <th>Gradient accumulation steps</th> <th>Comments</th></tr></thead> <tbody><tr><td>1</td> <td>32</td> <td>1</td> <td>Possibly high memory usage, but faster training</td></tr> <tr><td>1</td> <td>4</td> <td>8</td> <td>Lower memory usage, slower training</td></tr> <tr><td>8</td> <td>4</td> <td>1</td> <td>Multi-GPU to get the best of both worlds</td></tr></tbody></table>  <div class="course-tip  bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-1wlllm2">Having one model per GPU can lead to high memory usage, which may not be feasible for large models or low-memory GPUs. In such cases, you can leverage <a href="https://github.com/deepspeedai/DeepSpeed" rel="nofollow">DeepSpeed</a>, which provides optimizations like model sharding, Zero Redundancy Optimizer, mixed precision training, and offloading to CPU or NVMe. Check out our <a href="deepspeed_integration.md">DeepSpeed Integration</a> guide for more details.</p></div>  <h2 class="relative group"><a id="multi-nodes-training" 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="#multi-nodes-training"><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>Multi-Nodes Training</span></h2> <p data-svelte-h="svelte-cn0u2z">We’re working on a guide for multi-node training. Stay tuned! 🚀</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/distributing_training.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_4tczb2 = {
						assets: "/docs/trl/pr_3582/en",
						base: "/docs/trl/pr_3582/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/trl/pr_3582/en/_app/immutable/entry/start.0f0f318c.js"),
						import("/docs/trl/pr_3582/en/_app/immutable/entry/app.b27a462f.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 15],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>