Hugging Face's logo

Buckets:
hf-doc-build
/
doc-dev

Files
xet
hf-doc-build/doc-dev / trl /pr_4331 /en /distributing_training.html
rtrm's picture
rtrm
download
raw
43.2 kB
<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;Context Parallelism&quot;,&quot;local&quot;:&quot;context-parallelism&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Requirements and Limitations&quot;,&quot;local&quot;:&quot;requirements-and-limitations&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Configuration&quot;,&quot;local&quot;:&quot;configuration&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Accelerate Configuration&quot;,&quot;local&quot;:&quot;accelerate-configuration&quot;,&quot;sections&quot;:[],&quot;depth&quot;:4},{&quot;title&quot;:&quot;Training Configuration&quot;,&quot;local&quot;:&quot;training-configuration&quot;,&quot;sections&quot;:[],&quot;depth&quot;:4}],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Best Practices&quot;,&quot;local&quot;:&quot;best-practices&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Benchmarking Context Parallelism&quot;,&quot;local&quot;:&quot;benchmarking-context-parallelism&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Further Reading on Context Parallelism&quot;,&quot;local&quot;:&quot;further-reading-on-context-parallelism&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Multi-Node Training&quot;,&quot;local&quot;:&quot;multi-node-training&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
 <link href="/docs/trl/pr_4331/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/entry/start.6bbbc54b.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/scheduler.7b731bd4.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/singletons.55eb59f9.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/index.ac28c20f.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/paths.677b038d.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/entry/app.b003256e.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/preload-helper.71df5523.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/index.cc268345.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/nodes/0.c996cd3a.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/each.e59479a4.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/nodes/12.57f3a195.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.d403d039.js">
 <link rel="modulepreload" href="/docs/trl/pr_4331/en/_app/immutable/chunks/CodeBlock.17bc4142.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;Context Parallelism&quot;,&quot;local&quot;:&quot;context-parallelism&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Requirements and Limitations&quot;,&quot;local&quot;:&quot;requirements-and-limitations&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Configuration&quot;,&quot;local&quot;:&quot;configuration&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Accelerate Configuration&quot;,&quot;local&quot;:&quot;accelerate-configuration&quot;,&quot;sections&quot;:[],&quot;depth&quot;:4},{&quot;title&quot;:&quot;Training Configuration&quot;,&quot;local&quot;:&quot;training-configuration&quot;,&quot;sections&quot;:[],&quot;depth&quot;:4}],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Best Practices&quot;,&quot;local&quot;:&quot;best-practices&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Benchmarking Context Parallelism&quot;,&quot;local&quot;:&quot;benchmarking-context-parallelism&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Further Reading on Context Parallelism&quot;,&quot;local&quot;:&quot;further-reading-on-context-parallelism&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Multi-Node Training&quot;,&quot;local&quot;:&quot;multi-node-training&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;: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 max-sm:gap-0.5 h-6 max-sm:h-5 px-2 max-sm:px-1.5 text-[11px] max-sm:text-[9px] 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"><svg class="w-3 h-3 max-sm:w-2.5 max-sm:h-2.5" 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-6 max-sm:h-5 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 w-3 h-3 max-sm:w-2.5 max-sm:h-2.5 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="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> <blockquote class="warning" data-svelte-h="svelte-1gc28wp"><p>Section under construction. Feel free to contribute!</p></blockquote> <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-jw4lic"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/multi_gpu.png" alt="multi gpu"></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> <blockquote class="tip" data-svelte-h="svelte-yb03hq"><p>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">DeepSpeed Integration</a> guide for more details.</p></blockquote> <h2 class="relative group"><a id="context-parallelism" 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="#context-parallelism"><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>Context Parallelism</span></h2> <p data-svelte-h="svelte-dacuvn">Context Parallelism (CP) is a parallelization technique that enables training with longer sequences by splitting the sequence dimension across multiple GPUs. Each GPU processes a portion of the sequence, allowing you to train with sequences longer than what would fit on a single GPU’s memory.</p> <p data-svelte-h="svelte-1ftyw1f">For more details on CP, see the <a href="https://huggingface.co/spaces/nanotron/ultrascale-playbook?section=context_parallelism" rel="nofollow">Ultrascale Playbook - Context Parallelism</a>.</p> <p data-svelte-h="svelte-5cqgjt">CP is particularly useful when:</p> <ul data-svelte-h="svelte-1los5ox"><li>You want to train with very long sequences (&gt;32k tokens)</li> <li>Single GPU memory is insufficient for your desired sequence length</li> <li>You need to maintain sequence coherence across the full context</li></ul> <h3 class="relative group"><a id="requirements-and-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="#requirements-and-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>Requirements and Limitations</span></h3> <p data-svelte-h="svelte-z275fr">CP has specific requirements:</p> <ol data-svelte-h="svelte-1gm4trx"><li><strong>Accelerate 1.10 or higher</strong> is required</li> <li><strong>FSDP2 (PyTorch FSDP v2)</strong> is required as the distributed training backend</li> <li><strong>SDPA attention</strong> - Flash Attention is currently not supported with CP</li> <li><strong>Sequence length divisibility</strong> - sequences must be divisible by <code>cp_size * 2</code>. This is now automatically handled using the <code>pad_to_multiple_of</code> parameter in the data collator, which works seamlessly with both standard and padding-free modes.</li></ol> <h3 class="relative group"><a id="configuration" 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="#configuration"><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>Configuration</span></h3> <p data-svelte-h="svelte-qtiqiz">To enable CP, you need to configure both Accelerate and your training arguments:</p> <h4 class="relative group"><a id="accelerate-configuration" 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="#accelerate-configuration"><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>Accelerate Configuration</span></h4> <p data-svelte-h="svelte-14prv5">Use one of the provided accelerate config files (e.g. <a href="https://github.com/huggingface/trl/blob/main/examples/accelerate_configs/context_parallel_2gpu.yaml" rel="nofollow"><code>context_parallel_2gpu.yaml</code></a> for 2 GPUs):</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-attr">compute_environment:</span> <span class="hljs-string">LOCAL_MACHINE</span>
<span class="hljs-attr">debug:</span> <span class="hljs-literal">false</span>
<span class="hljs-attr">distributed_type:</span> <span class="hljs-string">FSDP</span>
<span class="hljs-attr">downcast_bf16:</span> <span class="hljs-string">&#x27;no&#x27;</span>
<span class="hljs-attr">enable_cpu_affinity:</span> <span class="hljs-literal">false</span>
<span class="hljs-attr">fsdp_config:</span>
 <span class="hljs-attr">fsdp_activation_checkpointing:</span> <span class="hljs-literal">true</span> <span class="hljs-comment"># Enable activation checkpointing for memory efficiency</span>
 <span class="hljs-attr">fsdp_auto_wrap_policy:</span> <span class="hljs-string">TRANSFORMER_BASED_WRAP</span>
 <span class="hljs-attr">fsdp_cpu_ram_efficient_loading:</span> <span class="hljs-literal">true</span>
 <span class="hljs-attr">fsdp_offload_params:</span> <span class="hljs-literal">false</span>
 <span class="hljs-attr">fsdp_reshard_after_forward:</span> <span class="hljs-literal">true</span>
 <span class="hljs-attr">fsdp_state_dict_type:</span> <span class="hljs-string">FULL_STATE_DICT</span>
 <span class="hljs-attr">fsdp_version:</span> <span class="hljs-number">2</span>
<span class="hljs-attr">machine_rank:</span> <span class="hljs-number">0</span>
<span class="hljs-attr">main_training_function:</span> <span class="hljs-string">main</span>
<span class="hljs-attr">mixed_precision:</span> <span class="hljs-string">bf16</span>
<span class="hljs-attr">num_machines:</span> <span class="hljs-number">1</span>
<span class="hljs-attr">num_processes:</span> <span class="hljs-number">2</span> <span class="hljs-comment"># Number of GPUs</span>
<span class="hljs-attr">rdzv_backend:</span> <span class="hljs-string">static</span>
<span class="hljs-attr">same_network:</span> <span class="hljs-literal">true</span>
<span class="hljs-attr">tpu_env:</span> []
<span class="hljs-attr">tpu_use_cluster:</span> <span class="hljs-literal">false</span>
<span class="hljs-attr">tpu_use_sudo:</span> <span class="hljs-literal">false</span>
<span class="hljs-attr">use_cpu:</span> <span class="hljs-literal">false</span>
<span class="hljs-attr">parallelism_config:</span>
 <span class="hljs-attr">parallelism_config_dp_replicate_size:</span> <span class="hljs-number">1</span>
 <span class="hljs-attr">parallelism_config_dp_shard_size:</span> <span class="hljs-number">1</span>
 <span class="hljs-attr">parallelism_config_tp_size:</span> <span class="hljs-number">1</span>
 <span class="hljs-attr">parallelism_config_cp_size:</span> <span class="hljs-number">2</span> <span class="hljs-comment"># Context parallel size</span><!-- HTML_TAG_END --></pre></div> <h4 class="relative group"><a id="training-configuration" 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="#training-configuration"><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>Training Configuration</span></h4> <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">from</span> trl <span class="hljs-keyword">import</span> SFTConfig
training_args = SFTConfig(
 <span class="hljs-comment"># required</span>
 pad_to_multiple_of=<span class="hljs-number">4</span>, <span class="hljs-comment"># ensures divisibility by cp_size * 2</span>
 <span class="hljs-comment"># to get the most out of CP</span>
 max_length=<span class="hljs-number">16384</span>, <span class="hljs-comment"># long sequence length</span>
 packing=<span class="hljs-literal">True</span>, <span class="hljs-comment"># use packing to reduce padding</span>
 use_liger_kernel=<span class="hljs-literal">True</span>, <span class="hljs-comment"># compatible with CP</span>
 gradient_checkpointing=<span class="hljs-literal">False</span>, <span class="hljs-comment"># The activation_checkpointing in FSDP config and the gradient_checkpointing in training arg can&#x27;t be set to True simultaneously</span>
 per_device_train_batch_size=<span class="hljs-number">1</span>,
 ...
)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1opb19">Then, launch your training script with the appropriate accelerate config file:</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 context_parallel_2gpu.yaml train.py<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="best-practices" 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="#best-practices"><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>Best Practices</span></h3> <ol data-svelte-h="svelte-mfle9"><li><p><strong>Use the <code>pad_to_multiple_of</code> parameter</strong> - This is now the recommended way to ensure sequence length divisibility:</p> <ul><li>For <code>cp_size=2</code>: use <code>pad_to_multiple_of=4</code> (since <code>cp_size * 2 = 4</code>)</li> <li>For <code>cp_size=4</code>: use <code>pad_to_multiple_of=8</code> (since <code>cp_size * 2 = 8</code>)</li> <li>The data collator automatically pads sequences to the required multiple, ensuring compatibility with CP</li></ul></li> <li><p><strong>Use packing with padding</strong> - The default BFD (Best Fit Decreasing) strategy works perfectly:</p> <ul><li>Preserves sequence boundaries and maintains training quality</li> <li>Works seamlessly with both <code>padding_free=True</code> and standard padding modes</li></ul></li> <li><p><strong>Combine with other memory optimizations</strong> like Liger kernels, bfloat16, and gradient checkpointing</p></li> <li><p><strong>Start with smaller context parallel sizes</strong> (2-4 GPUs) before scaling up</p></li> <li><p><strong>Monitor memory usage</strong> across all GPUs to ensure balanced workload</p></li></ol> <h3 class="relative group"><a id="benchmarking-context-parallelism" 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="#benchmarking-context-parallelism"><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>Benchmarking Context Parallelism</span></h3> <p data-svelte-h="svelte-bxtdlf">We benchmarked CP to highlight its potential improvements in training efficiency.<br>
Our experiments were conducted using <strong>1, 2, 4, and 8 H100 GPUs</strong>, though the results can be extended to larger clusters with more nodes and GPUs.</p> <p data-svelte-h="svelte-1ar3p4q">For the setup, we fine-tuned an <strong>8B model</strong> (<a href="https://huggingface.co/Qwen/Qwen3-8B" rel="nofollow">Qwen/Qwen3-8B</a>) using the provided accelerate configuration<br>
(<a href="https://github.com/huggingface/trl/blob/main/examples/accelerate_configs/context_parallel_2gpu.yaml" rel="nofollow"><code>context_parallel_2gpu.yaml</code></a>).<br>
We adjusted <code>num_processes</code> and <code>parallelism_config_cp_size</code> based on the number of GPUs for each run.<br>
Training was performed with the <a href="https://github.com/huggingface/trl/blob/main/trl/scripts/sft.py" rel="nofollow">sft.py</a> example script, combined with the parameters described above.</p> <p data-svelte-h="svelte-iruthr">The results below summarize the <strong>maximum trainable sequence length</strong> and <strong>iterations per second</strong> for different numbers of GPUs. A value marked as <code>OOM</code> indicates that the configuration ran out of memory and could not be trained.</p> <p data-svelte-h="svelte-1yjupp2">These results show that <strong>Context Parallelism (CP) scales effectively with more GPUs</strong>, enabling training on much longer sequences. With <strong>8 GPUs</strong>, context lengths of over <strong>300k tokens</strong> become feasible, unlocking training with extremely long contexts while maintaining reasonable throughput.</p> <div class="flex justify-center" data-svelte-h="svelte-66t12q"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/context_parallelism_max_length_plot.png" alt="CP Max content length" width="45%"> <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/context_parallelism_s_it_plot.png" alt="CP seconds/iteration" width="45%"></div> <blockquote class="tip" data-svelte-h="svelte-x1ndil"><p>Accelerate also supports <strong>N-Dimensional Parallelism (ND-parallelism)</strong>, which enables you to combine different parallelization strategies to efficiently distribute model training across multiple GPUs.</p> <p>You can learn more and explore configuration examples in the <a href="https://github.com/huggingface/accelerate/blob/main/examples/torch_native_parallelism/README.md#nd-parallelism" rel="nofollow">Accelerate ND-parallelism guide</a>.</p></blockquote> <h3 class="relative group"><a id="further-reading-on-context-parallelism" 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="#further-reading-on-context-parallelism"><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>Further Reading on Context Parallelism</span></h3> <ul data-svelte-h="svelte-1o3jk6e"><li><a href="https://github.com/huggingface/accelerate/blob/main/docs/source/concept_guides/context_parallelism.md" rel="nofollow">Accelerate: Context Parallelism Guide</a></li> <li><a href="https://github.com/huggingface/accelerate/blob/main/examples/torch_native_parallelism/README.md#context-parallelism-128k-sequence-length" rel="nofollow">Accelerate Example: 128k Sequence Length</a></li> <li><a href="https://huggingface.co/blog/axolotl-ai-co/long-context-with-sequence-parallelism-in-axolotl" rel="nofollow">Hugging Face Blog: Enabling Long-Context Training with Sequence Parallelism in Axolotl</a></li> <li><a href="https://www.snowflake.com/en/engineering-blog/arctic-long-sequence-training-multi-million-token-ai/" rel="nofollow">Snowflake Engineering Blog: Arctic Long Sequence Training (ALST) — Scalable and Efficient Training for Multi-Million Token Sequences (Note that they use a different strategy)</a></li></ul> <h2 class="relative group"><a id="multi-node-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-node-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-Node 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"><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_l5mord = {
 assets: "/docs/trl/pr_4331/en",
 base: "/docs/trl/pr_4331/en",
 env: {}
 };
 const element = document.currentScript.parentElement;
 const data = [null,null];
 Promise.all([
 import("/docs/trl/pr_4331/en/_app/immutable/entry/start.6bbbc54b.js"),
 import("/docs/trl/pr_4331/en/_app/immutable/entry/app.b003256e.js")
 ]).then(([kit, app]) => {
 kit.start(app, element, {
 node_ids: [0, 12],
 data,
 form: null,
 error: null
 });
 });
 }
 </script>

Xet Storage Details

Size:
43.2 kB
·
Xet hash:
7c3737d117af638b6e5657942fb913b39c9ebf1d86e2687f236b4602f6a7f23c

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