Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Using Local SGD with Accelerate","local":"using-local-sgd-with-accelerate","sections":[{"title":"Converting it to Accelerate","local":"converting-it-to-accelerate","sections":[],"depth":2},{"title":"Letting Accelerate handle model synchronization","local":"letting-accelerate-handle-model-synchronization","sections":[],"depth":2},{"title":"Limitations","local":"limitations","sections":[],"depth":2},{"title":"References","local":"references","sections":[],"depth":2}],"depth":1}"> | |
| <link href="/docs/accelerate/pr_4021/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/entry/start.8a49e72b.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/scheduler.b9285784.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/singletons.7547c222.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/index.6d423e5c.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/paths.d42c9205.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/entry/app.1df4d18e.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/preload-helper.b0bd19d1.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/index.26bc89a1.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/nodes/0.0e7c56e8.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/nodes/50.e991b299.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.7a0ae628.js"> | |
| <link rel="modulepreload" href="/docs/accelerate/pr_4021/en/_app/immutable/chunks/CodeBlock.844ff9c3.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Using Local SGD with Accelerate","local":"using-local-sgd-with-accelerate","sections":[{"title":"Converting it to Accelerate","local":"converting-it-to-accelerate","sections":[],"depth":2},{"title":"Letting Accelerate handle model synchronization","local":"letting-accelerate-handle-model-synchronization","sections":[],"depth":2},{"title":"Limitations","local":"limitations","sections":[],"depth":2},{"title":"References","local":"references","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" 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></span> <span>Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-7 max-sm:h-7 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible sm:size-3.5 size-3 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg></button></div> </div> <h1 class="relative group"><a id="using-local-sgd-with-accelerate" 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="#using-local-sgd-with-accelerate"><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>Using Local SGD with Accelerate</span></h1> <p data-svelte-h="svelte-1gzsxbq">Local SGD is a technique for distributed training where gradients are not synchronized every step. Thus, each process updates its own version of the model weights and after a given number of steps these weights are synchronized by averaging across all processes. This improves communication efficiency and can lead to substantial training speed up especially when a computer lacks a faster interconnect such as NVLink. | |
| Unlike gradient accumulation (where improving communication efficiency requires increasing the effective batch size), Local SGD does not require changing a batch size or a learning rate / schedule. However, if necessary, Local SGD can be combined with gradient accumulation as well.</p> <p data-svelte-h="svelte-1d63qys">In this tutorial you will see how to quickly setup Local SGD Accelerate. Compared to a standard Accelerate setup, this requires only two extra lines of code.</p> <p data-svelte-h="svelte-n9bxrd">This example will use a very simplistic PyTorch training loop that performs gradient accumulation every two batches:</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 -->device = <span class="hljs-string">"cuda"</span> | |
| model.to(device) | |
| gradient_accumulation_steps = <span class="hljs-number">2</span> | |
| <span class="hljs-keyword">for</span> index, batch <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(training_dataloader): | |
| inputs, targets = batch | |
| inputs = inputs.to(device) | |
| targets = targets.to(device) | |
| outputs = model(inputs) | |
| loss = loss_function(outputs, targets) | |
| loss = loss / gradient_accumulation_steps | |
| loss.backward() | |
| <span class="hljs-keyword">if</span> (index + <span class="hljs-number">1</span>) % gradient_accumulation_steps == <span class="hljs-number">0</span>: | |
| optimizer.step() | |
| scheduler.step() | |
| optimizer.zero_grad()<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="converting-it-to-accelerate" 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="#converting-it-to-accelerate"><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>Converting it to Accelerate</span></h2> <p data-svelte-h="svelte-19w42d7">First the code shown earlier will be converted to use Accelerate with neither a LocalSGD or a gradient accumulation helper:</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-addition">+ from accelerate import Accelerator</span> | |
| <span class="hljs-addition">+ accelerator = Accelerator()</span> | |
| <span class="hljs-addition">+ model, optimizer, training_dataloader, scheduler = accelerator.prepare(</span> | |
| <span class="hljs-addition">+ model, optimizer, training_dataloader, scheduler</span> | |
| <span class="hljs-addition">+ )</span> | |
| for index, batch in enumerate(training_dataloader): | |
| inputs, targets = batch | |
| <span class="hljs-deletion">- inputs = inputs.to(device)</span> | |
| <span class="hljs-deletion">- targets = targets.to(device)</span> | |
| outputs = model(inputs) | |
| loss = loss_function(outputs, targets) | |
| loss = loss / gradient_accumulation_steps | |
| <span class="hljs-addition">+ accelerator.backward(loss)</span> | |
| if (index+1) % gradient_accumulation_steps == 0: | |
| optimizer.step() | |
| scheduler.step()<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="letting-accelerate-handle-model-synchronization" 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="#letting-accelerate-handle-model-synchronization"><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>Letting Accelerate handle model synchronization</span></h2> <p data-svelte-h="svelte-192byqc">All that is left now is to let Accelerate handle model parameter synchronization <strong>and</strong> the gradient accumulation for us. For simplicity let us assume we need to synchronize every 8 steps. This is | |
| achieved by adding one <code>with LocalSGD</code> statement and one call <code>local_sgd.step()</code> after every optimizer step:</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-addition">+local_sgd_steps=8</span> | |
| <span class="hljs-addition">+with LocalSGD(accelerator=accelerator, model=model, local_sgd_steps=8, enabled=True) as local_sgd:</span> | |
| for batch in training_dataloader: | |
| with accelerator.accumulate(model): | |
| inputs, targets = batch | |
| outputs = model(inputs) | |
| loss = loss_function(outputs, targets) | |
| accelerator.backward(loss) | |
| optimizer.step() | |
| scheduler.step() | |
| optimizer.zero_grad() | |
| <span class="hljs-addition">+ local_sgd.step()</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-avyf7q">Under the hood, the Local SGD code <strong>disables</strong> automatic gradient synchronization (but accumulation still works as expected!). Instead it averages model parameters every <code>local_sgd_steps</code> steps (as well as at the end of the training loop).</p> <h2 class="relative group"><a id="limitations" 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="#limitations"><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>Limitations</span></h2> <p data-svelte-h="svelte-10m76zr">The current implementation works only with basic multi-GPU (or multi-CPU) training without, e.g., <a href="https://github.com/deepspeedai/DeepSpeed" rel="nofollow">DeepSpeed.</a>.</p> <h2 class="relative group"><a id="references" 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="#references"><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>References</span></h2> <p data-svelte-h="svelte-1uio6eh">Although we are not aware of the true origins of this simple approach, the idea of local SGD is quite old and goes | |
| back to at least:</p> <p data-svelte-h="svelte-xebpa">Zhang, J., De Sa, C., Mitliagkas, I., & Ré, C. (2016). <a href="https://huggingface.co/papers/1606.07365" rel="nofollow">Parallel SGD: When does averaging help?. arXiv preprint | |
| arXiv:1606.07365.</a></p> <p data-svelte-h="svelte-5vkth6">We credit the term Local SGD to the following paper (but there might be earlier references we are not aware of).</p> <p data-svelte-h="svelte-1xdvzf3">Stich, Sebastian Urban. <a href="https://huggingface.co/papers/1805.09767" rel="nofollow">“Local SGD Converges Fast and Communicates Little.” ICLR 2019-International Conference on | |
| Learning Representations. No. CONF. 2019.</a></p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/accelerate/blob/main/docs/source/usage_guides/local_sgd.md" target="_blank"><svg class="mr-1" 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="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p> | |
| <script> | |
| { | |
| __sveltekit_1q7nz6m = { | |
| assets: "/docs/accelerate/pr_4021/en", | |
| base: "/docs/accelerate/pr_4021/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/accelerate/pr_4021/en/_app/immutable/entry/start.8a49e72b.js"), | |
| import("/docs/accelerate/pr_4021/en/_app/immutable/entry/app.1df4d18e.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 50], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 20.9 kB
- Xet hash:
- f5fc6c22df6c4236e47ce9f702be4b88aac4987af8d2500a8697a31a711504b3
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.