kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity
a79e2dc7fb
TensorRT-LLMs/features/parallel-strategy.html
Kaiyu Xie a79e2dc7fb Update GitHub pages in root to v1.2.0rc4
2025-11-25 03:40:39 +00:00

919 lines
51 KiB
HTML
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<!DOCTYPE html>
<html lang="en" data-content_root="../" >
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" /><meta name="viewport" content="width=device-width, initial-scale=1" />
<title>Parallelism in TensorRT LLM &#8212; TensorRT LLM</title>
<script data-cfasync="false">
document.documentElement.dataset.mode = localStorage.getItem("mode") || "";
document.documentElement.dataset.theme = localStorage.getItem("theme") || "";
</script>
<!--
this give us a css class that will be invisible only if js is disabled
-->
<noscript>
<style>
.pst-js-only { display: none !important; }
</style>
</noscript>
<!-- Loaded before other Sphinx assets -->
<link href="../_static/styles/theme.css?digest=8878045cc6db502f8baf" rel="stylesheet" />
<link href="../_static/styles/pydata-sphinx-theme.css?digest=8878045cc6db502f8baf" rel="stylesheet" />
<link rel="stylesheet" type="text/css" href="../_static/pygments.css?v=8f2a1f02" />
<link rel="stylesheet" type="text/css" href="../_static/styles/nvidia-sphinx-theme.css?v=933278ad" />
<link rel="stylesheet" type="text/css" href="../_static/copybutton.css?v=76b2166b" />
<link rel="stylesheet" type="text/css" href="../_static/autodoc_pydantic.css" />
<link rel="stylesheet" type="text/css" href="../_static/togglebutton.css?v=13237357" />
<link rel="stylesheet" type="text/css" href="../_static/custom.css?v=19d20f17" />
<!-- So that users can add custom icons -->
<script src="../_static/scripts/fontawesome.js?digest=8878045cc6db502f8baf"></script>
<!-- Pre-loaded scripts that we'll load fully later -->
<link rel="preload" as="script" href="../_static/scripts/bootstrap.js?digest=8878045cc6db502f8baf" />
<link rel="preload" as="script" href="../_static/scripts/pydata-sphinx-theme.js?digest=8878045cc6db502f8baf" />
<script src="../_static/documentation_options.js?v=5929fcd5"></script>
<script src="../_static/doctools.js?v=9a2dae69"></script>
<script src="../_static/sphinx_highlight.js?v=dc90522c"></script>
<script src="../_static/clipboard.min.js?v=a7894cd8"></script>
<script src="../_static/copybutton.js?v=65e89d2a"></script>
<script>let toggleHintShow = 'Click to show';</script>
<script>let toggleHintHide = 'Click to hide';</script>
<script>let toggleOpenOnPrint = 'true';</script>
<script src="../_static/togglebutton.js?v=4a39c7ea"></script>
<script>var togglebuttonSelector = '.toggle, .admonition.dropdown';</script>
<script>var togglebuttonSelector = '.toggle, .admonition.dropdown';</script>
<script>DOCUMENTATION_OPTIONS.pagename = 'features/parallel-strategy';</script>
<script>
DOCUMENTATION_OPTIONS.theme_version = '0.16.1';
DOCUMENTATION_OPTIONS.theme_switcher_json_url = './_static/switcher.json';
DOCUMENTATION_OPTIONS.theme_switcher_version_match = '1.2.0rc4';
DOCUMENTATION_OPTIONS.show_version_warning_banner =
false;
</script>
<link rel="icon" href="../_static/favicon.png"/>
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="Quantization" href="quantization.html" />
<link rel="prev" title="Paged Attention, IFB, and Request Scheduling" href="paged-attention-ifb-scheduler.html" />
<meta name="viewport" content="width=device-width, initial-scale=1"/>
<meta name="docsearch:language" content="en"/>
<meta name="docsearch:version" content="1.2.0rc4" />
</head>
<body data-bs-spy="scroll" data-bs-target=".bd-toc-nav" data-offset="180" data-bs-root-margin="0px 0px -60%" data-default-mode="">
<div id="pst-skip-link" class="skip-link d-print-none"><a href="#main-content">Skip to main content</a></div>
<div id="pst-scroll-pixel-helper"></div>
<button type="button" class="btn rounded-pill" id="pst-back-to-top">
<i class="fa-solid fa-arrow-up"></i>Back to top</button>
<dialog id="pst-search-dialog">
<form class="bd-search d-flex align-items-center"
action="../search.html"
method="get">
<i class="fa-solid fa-magnifying-glass"></i>
<input type="search"
class="form-control"
name="q"
placeholder="Search the docs ..."
aria-label="Search the docs ..."
autocomplete="off"
autocorrect="off"
autocapitalize="off"
spellcheck="false"/>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd>K</kbd></span>
</form>
</dialog>
<div class="pst-async-banner-revealer d-none">
<aside id="bd-header-version-warning" class="d-none d-print-none" aria-label="Version warning"></aside>
</div>
<header class="bd-header navbar navbar-expand-lg bd-navbar d-print-none">
<div class="bd-header__inner bd-page-width">
<button class="pst-navbar-icon sidebar-toggle primary-toggle" aria-label="Site navigation">
<span class="fa-solid fa-bars"></span>
</button>
<div class="col-lg-3 navbar-header-items__start">
<div class="navbar-item">
<a class="navbar-brand logo" href="../index.html">
<img src="../_static/nvidia-logo-horiz-rgb-blk-for-screen.svg" class="logo__image only-light" alt="TensorRT LLM - Home"/>
<img src="../_static/nvidia-logo-horiz-rgb-wht-for-screen.svg" class="logo__image only-dark pst-js-only" alt="TensorRT LLM - Home"/>
<p class="title logo__title">TensorRT LLM</p>
</a></div>
</div>
<div class="col-lg-9 navbar-header-items">
<div class="me-auto navbar-header-items__center">
<div class="navbar-item">
<div class="version-switcher__container dropdown pst-js-only">
<button id="pst-version-switcher-button-2"
type="button"
class="version-switcher__button btn btn-sm dropdown-toggle"
data-bs-toggle="dropdown"
aria-haspopup="listbox"
aria-controls="pst-version-switcher-list-2"
aria-label="Version switcher list"
>
Choose version <!-- this text may get changed later by javascript -->
<span class="caret"></span>
</button>
<div id="pst-version-switcher-list-2"
class="version-switcher__menu dropdown-menu list-group-flush py-0"
role="listbox" aria-labelledby="pst-version-switcher-button-2">
<!-- dropdown will be populated by javascript on page load -->
</div>
</div></div>
</div>
<div class="navbar-header-items__end">
<div class="navbar-item navbar-persistent--container">
<button class="btn search-button-field search-button__button pst-js-only" title="Search" aria-label="Search" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="fa-solid fa-magnifying-glass"></i>
<span class="search-button__default-text">Search</span>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd class="kbd-shortcut__modifier">K</kbd></span>
</button>
</div>
<div class="navbar-item">
<button class="btn btn-sm nav-link pst-navbar-icon theme-switch-button pst-js-only" aria-label="Color mode" data-bs-title="Color mode" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="theme-switch fa-solid fa-sun fa-lg" data-mode="light" title="Light"></i>
<i class="theme-switch fa-solid fa-moon fa-lg" data-mode="dark" title="Dark"></i>
<i class="theme-switch fa-solid fa-circle-half-stroke fa-lg" data-mode="auto" title="System Settings"></i>
</button></div>
</div>
</div>
<div class="navbar-persistent--mobile">
<button class="btn search-button-field search-button__button pst-js-only" title="Search" aria-label="Search" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="fa-solid fa-magnifying-glass"></i>
<span class="search-button__default-text">Search</span>
<span class="search-button__kbd-shortcut"><kbd class="kbd-shortcut__modifier">Ctrl</kbd>+<kbd class="kbd-shortcut__modifier">K</kbd></span>
</button>
</div>
<button class="pst-navbar-icon sidebar-toggle secondary-toggle" aria-label="On this page">
<span class="fa-solid fa-outdent"></span>
</button>
</div>
</header>
<div class="bd-container">
<div class="bd-container__inner bd-page-width">
<dialog id="pst-primary-sidebar-modal"></dialog>
<div id="pst-primary-sidebar" class="bd-sidebar-primary bd-sidebar">
<a class="navbar-brand logo" href="../index.html">
<img src="../_static/nvidia-logo-horiz-rgb-blk-for-screen.svg" class="logo__image only-light" alt="TensorRT LLM - Home"/>
<img src="../_static/nvidia-logo-horiz-rgb-wht-for-screen.svg" class="logo__image only-dark pst-js-only" alt="TensorRT LLM - Home"/>
<p class="title logo__title">TensorRT LLM</p>
</a>
<div class="sidebar-header-items sidebar-primary__section">
<div class="sidebar-header-items__center">
<div class="navbar-item">
<div class="version-switcher__container dropdown pst-js-only">
<button id="pst-version-switcher-button-3"
type="button"
class="version-switcher__button btn btn-sm dropdown-toggle"
data-bs-toggle="dropdown"
aria-haspopup="listbox"
aria-controls="pst-version-switcher-list-3"
aria-label="Version switcher list"
>
Choose version <!-- this text may get changed later by javascript -->
<span class="caret"></span>
</button>
<div id="pst-version-switcher-list-3"
class="version-switcher__menu dropdown-menu list-group-flush py-0"
role="listbox" aria-labelledby="pst-version-switcher-button-3">
<!-- dropdown will be populated by javascript on page load -->
</div>
</div></div>
</div>
<div class="sidebar-header-items__end">
<div class="navbar-item">
<button class="btn btn-sm nav-link pst-navbar-icon theme-switch-button pst-js-only" aria-label="Color mode" data-bs-title="Color mode" data-bs-placement="bottom" data-bs-toggle="tooltip">
<i class="theme-switch fa-solid fa-sun fa-lg" data-mode="light" title="Light"></i>
<i class="theme-switch fa-solid fa-moon fa-lg" data-mode="dark" title="Dark"></i>
<i class="theme-switch fa-solid fa-circle-half-stroke fa-lg" data-mode="auto" title="System Settings"></i>
</button></div>
</div>
</div>
<div class="sidebar-primary-items__start sidebar-primary__section">
<div class="sidebar-primary-item">
<nav class="bd-docs-nav bd-links"
aria-label="Table of Contents">
<p class="bd-links__title" role="heading" aria-level="1">Table of Contents</p>
<div class="bd-toc-item navbar-nav"><p aria-level="2" class="caption" role="heading"><span class="caption-text">Getting Started</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../overview.html">Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../quick-start-guide.html">Quick Start Guide</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../installation/index.html">Installation</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../installation/containers.html">Pre-built release container images on NGC</a></li>
<li class="toctree-l2"><a class="reference internal" href="../installation/linux.html">Installing on Linux via <code class="docutils literal notranslate"><span class="pre">pip</span></code></a></li>
<li class="toctree-l2"><a class="reference internal" href="../installation/build-from-source-linux.html">Building from Source Code on Linux</a></li>
</ul>
</details></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Deployment Guide</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1 has-children"><a class="reference internal" href="../examples/llm_api_examples.html">LLM Examples</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_inference.html">Generate text</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_inference_async.html">Generate text asynchronously</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_inference_async_streaming.html">Generate text in streaming</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_inference_distributed.html">Distributed LLM Generation</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_guided_decoding.html">Generate text with guided decoding</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_logits_processor.html">Control generated text using logits processor</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_multilora.html">Generate text with multiple LoRA adapters</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_sparse_attention.html">Sparse Attention</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_speculative_decoding.html">Speculative Decoding</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_kv_cache_connector.html">KV Cache Connector</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_kv_cache_offloading.html">KV Cache Offloading</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_runtime.html">Runtime Configuration Examples</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_sampling.html">Sampling Techniques Showcase</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_mgmn_llm_distributed.html">Run LLM-API with pytorch backend on Slurm</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_mgmn_trtllm_bench.html">Run trtllm-bench with pytorch backend on Slurm</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/llm_mgmn_trtllm_serve.html">Run trtllm-serve with pytorch backend on Slurm</a></li>
</ul>
</details></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../examples/trtllm_serve_examples.html">Online Serving Examples</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../examples/curl_chat_client.html">Curl Chat Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/curl_chat_client_for_multimodal.html">Curl Chat Client For Multimodal</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/curl_completion_client.html">Curl Completion Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/deepseek_r1_reasoning_parser.html">Deepseek R1 Reasoning Parser</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/genai_perf_client.html">Genai Perf Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/genai_perf_client_for_multimodal.html">Genai Perf Client For Multimodal</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/openai_chat_client.html">OpenAI Chat Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/openai_chat_client_for_multimodal.html">OpenAI Chat Client for Multimodal</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/openai_completion_client.html">OpenAI Completion Client</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/openai_completion_client_for_lora.html">Openai Completion Client For Lora</a></li>
<li class="toctree-l2"><a class="reference internal" href="../examples/openai_completion_client_json_schema.html">OpenAI Completion Client with JSON Schema</a></li>
</ul>
</details></li>
<li class="toctree-l1"><a class="reference internal" href="../examples/dynamo_k8s_example.html">Dynamo K8s Example</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../deployment-guide/index.html">Model Recipes</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../deployment-guide/deployment-guide-for-deepseek-r1-on-trtllm.html">Deployment Guide for DeepSeek R1 on TensorRT LLM - Blackwell &amp; Hopper Hardware</a></li>
<li class="toctree-l2"><a class="reference internal" href="../deployment-guide/deployment-guide-for-llama3.3-70b-on-trtllm.html">Deployment Guide for Llama3.3 70B on TensorRT LLM - Blackwell &amp; Hopper Hardware</a></li>
<li class="toctree-l2"><a class="reference internal" href="../deployment-guide/deployment-guide-for-llama4-scout-on-trtllm.html">Deployment Guide for Llama4 Scout 17B on TensorRT LLM - Blackwell &amp; Hopper Hardware</a></li>
<li class="toctree-l2"><a class="reference internal" href="../deployment-guide/deployment-guide-for-gpt-oss-on-trtllm.html">Deployment Guide for GPT-OSS on TensorRT-LLM - Blackwell Hardware</a></li>
<li class="toctree-l2"><a class="reference internal" href="../deployment-guide/deployment-guide-for-qwen3-next-on-trtllm.html">Deployment Guide for Qwen3 Next on TensorRT LLM - Blackwell &amp; Hopper Hardware</a></li>
</ul>
</details></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Models</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../models/supported-models.html">Supported Models</a></li>
<li class="toctree-l1"><a class="reference internal" href="../models/adding-new-model.html">Adding a New Model</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">CLI Reference</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../commands/trtllm-bench.html">trtllm-bench</a></li>
<li class="toctree-l1"><a class="reference internal" href="../commands/trtllm-eval.html">trtllm-eval</a></li>
<li class="toctree-l1 has-children"><a class="reference internal" href="../commands/trtllm-serve/index.html">trtllm-serve</a><details><summary><span class="toctree-toggle" role="presentation"><i class="fa-solid fa-chevron-down"></i></span></summary><ul>
<li class="toctree-l2"><a class="reference internal" href="../commands/trtllm-serve/trtllm-serve.html">trtllm-serve</a></li>
<li class="toctree-l2"><a class="reference internal" href="../commands/trtllm-serve/run-benchmark-with-trtllm-serve.html">Run benchmarking with <code class="docutils literal notranslate"><span class="pre">trtllm-serve</span></code></a></li>
</ul>
</details></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">API Reference</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../llm-api/index.html">LLM API Introduction</a></li>
<li class="toctree-l1"><a class="reference internal" href="../llm-api/reference.html">API Reference</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Features</span></p>
<ul class="current nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="feature-combination-matrix.html">Feature Combination Matrix</a></li>
<li class="toctree-l1"><a class="reference internal" href="attention.html">Multi-Head, Multi-Query, and Group-Query Attention</a></li>
<li class="toctree-l1"><a class="reference internal" href="disagg-serving.html">Disaggregated Serving</a></li>
<li class="toctree-l1"><a class="reference internal" href="kvcache.html">KV Cache System</a></li>
<li class="toctree-l1"><a class="reference internal" href="long-sequence.html">Long Sequences</a></li>
<li class="toctree-l1"><a class="reference internal" href="lora.html">LoRA (Low-Rank Adaptation)</a></li>
<li class="toctree-l1"><a class="reference internal" href="multi-modality.html">Multimodal Support in TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="overlap-scheduler.html">Overlap Scheduler</a></li>
<li class="toctree-l1"><a class="reference internal" href="paged-attention-ifb-scheduler.html">Paged Attention, IFB, and Request Scheduling</a></li>
<li class="toctree-l1 current active"><a class="current reference internal" href="#">Parallelism in TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="quantization.html">Quantization</a></li>
<li class="toctree-l1"><a class="reference internal" href="sampling.html">Sampling</a></li>
<li class="toctree-l1"><a class="reference internal" href="additional-outputs.html">Additional Outputs</a></li>
<li class="toctree-l1"><a class="reference internal" href="speculative-decoding.html">Speculative Decoding</a></li>
<li class="toctree-l1"><a class="reference internal" href="checkpoint-loading.html">Checkpoint Loading</a></li>
<li class="toctree-l1"><a class="reference internal" href="auto_deploy/auto-deploy.html">AutoDeploy (Prototype)</a></li>
<li class="toctree-l1"><a class="reference internal" href="ray-orchestrator.html">Ray Orchestrator (Prototype)</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_compile_and_piecewise_cuda_graph.html">Torch Compile &amp; Piecewise CUDA Graph</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Developer Guide</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/overview.html">Architecture Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/perf-analysis.html">Performance Analysis</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/perf-benchmarking.html">TensorRT LLM Benchmarking</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/ci-overview.html">Continuous Integration Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/dev-containers.html">Using Dev Containers</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/api-change.html">LLM API Change Guide</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developer-guide/kv-transfer.html">Introduction to KV Cache Transmission</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Blogs</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog10_ADP_Balance_Strategy.html">ADP Balance Strategy</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog11_GPT_OSS_Eagle3.html">Running GPT-OSS-120B with Eagle3 Speculative Decoding on GB200/B200 (TensorRT LLM)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog12_Combining_Guided_Decoding_and_Speculative_Decoding.html">Combining Guided Decoding and Speculative Decoding: Making CPU and GPU Cooperate Seamlessly</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog13_Inference_Time_Compute_Implementation_in_TensorRT-LLM.html">Inference Time Compute Implementation in TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog14_Scaling_Expert_Parallelism_in_TensorRT-LLM_part3.html">Scaling Expert Parallelism in TensorRT LLM (Part 3: Pushing the Performance Boundary)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog1_Pushing_Latency_Boundaries_Optimizing_DeepSeek-R1_Performance_on_NVIDIA_B200_GPUs.html">Pushing Latency Boundaries: Optimizing DeepSeek-R1 Performance on NVIDIA B200 GPUs</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog2_DeepSeek_R1_MTP_Implementation_and_Optimization.html">DeepSeek R1 MTP Implementation and Optimization</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog3_Optimizing_DeepSeek_R1_Throughput_on_NVIDIA_Blackwell_GPUs.html">Optimizing DeepSeek R1 Throughput on NVIDIA Blackwell GPUs: A Deep Dive for Developers</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog4_Scaling_Expert_Parallelism_in_TensorRT-LLM.html">Scaling Expert Parallelism in TensorRT LLM (Part 1: Design and Implementation of Large-scale EP)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog5_Disaggregated_Serving_in_TensorRT-LLM.html">Disaggregated Serving in TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog6_Llama4_maverick_eagle_guide.html">How to launch Llama4 Maverick + Eagle3 TensorRT LLM server</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog7_NGram_performance_Analysis_And_Auto_Enablement.html">N-Gram Speculative Decoding in TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog8_Scaling_Expert_Parallelism_in_TensorRT-LLM_part2.html">Scaling Expert Parallelism in TensorRT LLM (Part 2: Performance Status and Optimization)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/tech_blog/blog9_Deploying_GPT_OSS_on_TRTLLM.html">Running a High Performance GPT-OSS-120B Inference Server with TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/Best_perf_practice_on_DeepSeek-R1_in_TensorRT-LLM.html">How to get best performance on DeepSeek-R1 in TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/H200launch.html">H200 achieves nearly 12,000 tokens/sec on Llama2-13B with TensorRT LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/XQA-kernel.html">New XQA-kernel provides 2.4x more Llama-70B throughput within the same latency budget</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/H100vsA100.html">H100 has 4.6x A100 Performance in TensorRT LLM, achieving 10,000 tok/s at 100ms to first token</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Quick Links</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM/releases">Releases</a></li>
<li class="toctree-l1"><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM">Github Code</a></li>
<li class="toctree-l1"><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM/issues?q=is%3Aissue%20state%3Aopen%20label%3Aroadmap">Roadmap</a></li>
</ul>
<p aria-level="2" class="caption" role="heading"><span class="caption-text">Use TensorRT Engine</span></p>
<ul class="nav bd-sidenav">
<li class="toctree-l1"><a class="reference internal" href="../legacy/tensorrt_quickstart.html">LLM API with TensorRT Engine</a></li>
</ul>
</div>
</nav></div>
</div>
<div class="sidebar-primary-items__end sidebar-primary__section">
</div>
</div>
<main id="main-content" class="bd-main" role="main">
<div class="bd-content">
<div class="bd-article-container">
<div class="bd-header-article d-print-none">
<div class="header-article-items header-article__inner">
<div class="header-article-items__start">
<div class="header-article-item">
<nav aria-label="Breadcrumb" class="d-print-none">
<ul class="bd-breadcrumbs">
<li class="breadcrumb-item breadcrumb-home">
<a href="../index.html" class="nav-link" aria-label="Home">
<i class="fa-solid fa-home"></i>
</a>
</li>
<li class="breadcrumb-item active" aria-current="page"><span class="ellipsis">Parallelism in TensorRT LLM</span></li>
</ul>
</nav>
</div>
</div>
</div>
</div>
<div id="searchbox"></div>
<article class="bd-article">
<section class="tex2jax_ignore mathjax_ignore" id="parallelism-in-tensorrt-llm">
<h1>Parallelism in TensorRT LLM<a class="headerlink" href="#parallelism-in-tensorrt-llm" title="Link to this heading">#</a></h1>
<p>Parallelism across multiple GPUs becomes necessary when either</p>
<ul class="simple">
<li><p>the model cannot fit in a single GPU’s memory, or</p></li>
<li><p>a single GPU cannot deliver the desired performance.</p></li>
</ul>
<p>TensorRT LLM supports multiple parallelism strategies for deployment on both single and multiple nodes:</p>
<ul class="simple">
<li><p><strong>Tensor Parallel (TP)</strong> - Shards model weights across GPUs</p></li>
<li><p><strong>Pipeline Parallel (PP)</strong> - Distributes model layers across GPUs</p></li>
<li><p><strong>Data Parallel (DP)</strong> - Replicates model across GPUs for different requests</p></li>
<li><p><strong>Expert Parallel (EP)</strong> - Distributes experts across GPUs for MoE models</p></li>
<li><p><strong>Context Parallel (CP)</strong> - Distributes context processing across GPUs</p></li>
<li><p><strong>Wide Expert Parallel (Wide-EP)</strong> - Advanced EP with load balancing for large-scale MoE models</p></li>
</ul>
<section id="overview-of-parallelism-strategies">
<h2>Overview of Parallelism Strategies<a class="headerlink" href="#overview-of-parallelism-strategies" title="Link to this heading">#</a></h2>
<section id="tensor-parallelism-tp">
<h3>Tensor Parallelism (TP)<a class="headerlink" href="#tensor-parallelism-tp" title="Link to this heading">#</a></h3>
<p>Tensor parallelism splits the model weights across multiple GPUs. Each GPU holds a portion of the weights and processes the same input tokens, with results combined through communication.</p>
<p><strong>Best for:</strong> Small batch sizes, memory-constrained scenarios</p>
</section>
<section id="pipeline-parallelism-pp">
<h3>Pipeline Parallelism (PP)<a class="headerlink" href="#pipeline-parallelism-pp" title="Link to this heading">#</a></h3>
<p>Pipeline parallelism distributes different layers of the model across multiple GPUs. Each GPU processes a subset of layers, with activations passed between GPUs.</p>
<p><strong>Best for:</strong> Large models that don’t fit in single GPU memory</p>
</section>
<section id="data-parallelism-dp">
<h3>Data Parallelism (DP)<a class="headerlink" href="#data-parallelism-dp" title="Link to this heading">#</a></h3>
<p>Data parallelism replicates the entire model across multiple GPUs. Each GPU processes different requests independently.</p>
<p><strong>Best for:</strong> Large batch sizes, high throughput scenarios</p>
</section>
<section id="expert-parallelism-ep">
<h3>Expert Parallelism (EP)<a class="headerlink" href="#expert-parallelism-ep" title="Link to this heading">#</a></h3>
<p>Expert parallelism is specifically designed for Mixture of Experts (MoE) models, where different experts are distributed across GPUs.</p>
<p><strong>Best for:</strong> MoE models with high expert count</p>
</section>
<section id="context-parallelism-cp">
<h3>Context Parallelism (CP)<a class="headerlink" href="#context-parallelism-cp" title="Link to this heading">#</a></h3>
<p>Context parallelism distributes the processing of long sequences across multiple GPUs.</p>
<p><strong>Best for:</strong> Long context scenarios</p>
</section>
<section id="wide-expert-parallelism-wide-ep">
<h3>Wide Expert Parallelism (Wide-EP)<a class="headerlink" href="#wide-expert-parallelism-wide-ep" title="Link to this heading">#</a></h3>
<p>Wide-EP is an advanced form of expert parallelism that addresses the inherent workload imbalance in large-scale MoE models through intelligent load balancing and expert replication.</p>
<p><strong>Best for:</strong> Large-scale MoE models like DeepSeek-V3/R1, LLaMA4, Qwen3</p>
</section>
</section>
<section id="module-level-parallelism-guide">
<h2>Module-level Parallelism Guide<a class="headerlink" href="#module-level-parallelism-guide" title="Link to this heading">#</a></h2>
<section id="attention-module">
<h3>Attention Module<a class="headerlink" href="#attention-module" title="Link to this heading">#</a></h3>
<p>TensorRT LLM supports two strategies for attention modules:</p>
<ul class="simple">
<li><p><strong>Tensor Parallelism (TP)</strong> — best for small batch sizes</p></li>
<li><p><strong>Data Parallelism (DP)</strong> — best for large batch sizes</p></li>
</ul>
<section id="id1">
<h4>Tensor Parallelism (TP)<a class="headerlink" href="#id1" title="Link to this heading">#</a></h4>
<ul class="simple">
<li><p>The GEMM weights before and after the attention kernel are evenly sharded across GPUs, as are the attention <code class="docutils literal notranslate"><span class="pre">num_heads</span></code>.</p></li>
<li><p>Exceptions:</p>
<ol class="arabic simple">
<li><p><strong>DeepSeek-R1</strong>: the <code class="docutils literal notranslate"><span class="pre">fused_A</span></code> GEMM is <em>not</em> sharded.</p></li>
<li><p><strong>GQA / MQA / MLA</strong>: if <code class="docutils literal notranslate"><span class="pre">num_heads</span> <span class="pre">&lt;</span> <span class="pre">tensor_parallel_size</span></code>, the KV-cache is replicated on every GPU.</p></li>
</ol>
</li>
</ul>
</section>
<section id="id2">
<h4>Data Parallelism (DP)<a class="headerlink" href="#id2" title="Link to this heading">#</a></h4>
<ul class="simple">
<li><p>All GEMM weights are <strong>replicated</strong> on every GPU.</p></li>
<li><p>The KV-cache is <strong>partitioned</strong>, because different user requests are routed to different DP ranks.</p></li>
</ul>
</section>
<section id="how-to-enable-attention-parallelism">
<h4>How to Enable Attention Parallelism<a class="headerlink" href="#how-to-enable-attention-parallelism" title="Link to this heading">#</a></h4>
<p>To deploy a model with the above parallel strategies using <code class="docutils literal notranslate"><span class="pre">trtllm-serve</span></code> or run benchmarking with <code class="docutils literal notranslate"><span class="pre">trtllm-bench</span></code>, create a YAML configuration file named <code class="docutils literal notranslate"><span class="pre">parallel_config.yaml</span></code>:</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>cat<span class="w"> </span><span class="s">&lt;&lt;EOF &gt; parallel_config.yaml</span>
<span class="s"># TP-8</span>
<span class="s">tensor_parallel_size: 8</span>
<span class="s">enable_attention_dp: false # default</span>
<span class="s"># DP-8</span>
<span class="s">tensor_parallel_size: 8</span>
<span class="s">enable_attention_dp: true</span>
<span class="s">EOF</span>
</pre></div>
</div>
</section>
</section>
<section id="ffn-module">
<h3>FFN Module<a class="headerlink" href="#ffn-module" title="Link to this heading">#</a></h3>
<section id="dense-models">
<h4>Dense Models<a class="headerlink" href="#dense-models" title="Link to this heading">#</a></h4>
<p>Tensor Parallelism is supported for the FFN layers of dense models.</p>
</section>
<section id="mixture-of-experts-moe">
<h4>Mixture of Experts (MoE)<a class="headerlink" href="#mixture-of-experts-moe" title="Link to this heading">#</a></h4>
<p>MoE replaces a single FFN with multiple experts. A router selects the top-k experts for each token and dispatches the corresponding hidden states.</p>
<p>TensorRT LLM supports three execution patterns for MoE:</p>
<ul class="simple">
<li><p><strong>TP</strong> - Every expert’s weight matrix is sliced across all GPUs. Each GPU sees all tokens.</p></li>
<li><p><strong>EP</strong> - Full weights of each expert reside on a single GPU. Each GPU only sees tokens routed to its local experts.</p></li>
<li><p><strong>Hybrid ETP</strong> - Each GPU stores a subset of experts (EP) and shards those weights further (TP), balancing workload and kernel efficiency.</p></li>
</ul>
</section>
<section id="how-to-enable-moe-parallelism">
<h4>How to Enable MoE Parallelism<a class="headerlink" href="#how-to-enable-moe-parallelism" title="Link to this heading">#</a></h4>
<p>To deploy a model with the above parallel strategies using <code class="docutils literal notranslate"><span class="pre">trtllm-serve</span></code> or run benchmarking with <code class="docutils literal notranslate"><span class="pre">trtllm-bench</span></code>, create a YAML configuration file named <code class="docutils literal notranslate"><span class="pre">parallel_config.yaml</span></code> as follows:</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>cat<span class="w"> </span><span class="s">&lt;&lt;EOF &gt; parallel_config.yaml</span>
<span class="s"># TP only</span>
<span class="s">tensor_parallel_size: 8</span>
<span class="s">moe_tensor_parallel_size: 8</span>
<span class="s"># EP only</span>
<span class="s">tensor_parallel_size: 8</span>
<span class="s">moe_expert_parallel_size: 8</span>
<span class="s"># Hybrid (TP-4 × EP-2)</span>
<span class="s">tensor_parallel_size: 8 # 4 × 2</span>
<span class="s">moe_tensor_parallel_size: 4</span>
<span class="s">moe_expert_parallel_size: 2</span>
<span class="s">EOF</span>
</pre></div>
</div>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>The product of <code class="docutils literal notranslate"><span class="pre">moe_tensor_parallel_size</span></code> and <code class="docutils literal notranslate"><span class="pre">moe_expert_parallel_size</span></code> must equal <code class="docutils literal notranslate"><span class="pre">tensor_parallel_size</span></code>.</p>
</div>
</section>
</section>
</section>
<section id="id3">
<h2>Wide Expert Parallelism (Wide-EP)<a class="headerlink" href="#id3" title="Link to this heading">#</a></h2>
<p>Wide Expert Parallelism (Wide-EP) is TensorRT LLM’s advanced solution for large-scale MoE model inference. It addresses the challenges of traditional expert parallelism through intelligent load balancing and expert replication strategies.</p>
<section id="motivation-for-wide-ep">
<h3>Motivation for Wide-EP<a class="headerlink" href="#motivation-for-wide-ep" title="Link to this heading">#</a></h3>
<p>Large-scale MoE models like DeepSeek-V3/R1, LLaMA4, and Qwen3 use fine-grained expert designs that introduce new challenges:</p>
<ul class="simple">
<li><p><strong>High memory demands</strong> for expert weights</p></li>
<li><p><strong>Inherent expert-level workload imbalance</strong> due to sparse execution patterns</p></li>
<li><p><strong>Communication overhead</strong> in distributed expert parallelism</p></li>
<li><p><strong>Hot expert problem</strong> where certain experts receive significantly more tokens than others</p></li>
</ul>
</section>
<section id="key-features-of-wide-ep">
<h3>Key Features of Wide-EP<a class="headerlink" href="#key-features-of-wide-ep" title="Link to this heading">#</a></h3>
<section id="expert-replication-and-load-balancing">
<h4>1. Expert Replication and Load Balancing<a class="headerlink" href="#expert-replication-and-load-balancing" title="Link to this heading">#</a></h4>
<p>Wide-EP introduces the concept of <strong>expert slots</strong> that are decoupled from specific experts. This allows:</p>
<ul class="simple">
<li><p>Multiple replicas of hot experts across different GPUs</p></li>
<li><p>Dynamic expert placement based on workload patterns</p></li>
<li><p>Both offline and online load balancing strategies</p></li>
</ul>
</section>
<section id="custom-ep-communication-kernels">
<h4>2. Custom EP Communication Kernels<a class="headerlink" href="#custom-ep-communication-kernels" title="Link to this heading">#</a></h4>
<ul class="simple">
<li><p>Optimized for NVIDIA GB200 Multi-Node NVLink (MNNVL)</p></li>
<li><p>Efficient all-to-all communication for expert dispatch and combine</p></li>
<li><p>Reduced communication overhead compared to traditional EP</p></li>
</ul>
</section>
<section id="expert-parallelism-load-balancer-eplb">
<h4>3. Expert Parallelism Load Balancer (EPLB)<a class="headerlink" href="#expert-parallelism-load-balancer-eplb" title="Link to this heading">#</a></h4>
<ul class="simple">
<li><p><strong>Offline EPLB</strong>: Pre-computed expert placement based on historical workload statistics</p></li>
<li><p><strong>Online EPLB</strong>: Dynamic expert placement that adapts to real-time traffic patterns</p></li>
<li><p>Layer-wise weight redistribution to minimize inference disruption</p></li>
</ul>
</section>
</section>
<section id="architecture-overview">
<h3>Architecture Overview<a class="headerlink" href="#architecture-overview" title="Link to this heading">#</a></h3>
<p>Wide-EP separates the concepts of <strong>experts</strong> and <strong>slots</strong>:</p>
<ul class="simple">
<li><p><strong>Expert</strong>: The concept from the model’s perspective (e.g., Expert 0, Expert 1, etc.)</p></li>
<li><p><strong>Slot</strong>: The concept from the model engine’s perspective (e.g., Slot 0, Slot 1, etc.)</p></li>
</ul>
<p>The system maintains a routing table that maps Expert IDs to Slot IDs, which can be updated by the load balancing policy.</p>
</section>
<section id="best-practices">
<h3>Best Practices<a class="headerlink" href="#best-practices" title="Link to this heading">#</a></h3>
<ol class="arabic simple">
<li><p><strong>Start with offline EPLB</strong> for production deployments with known workload patterns</p></li>
<li><p><strong>Use online EPLB</strong> for dynamic workloads or when traffic patterns change frequently</p></li>
<li><p><strong>Monitor expert statistics</strong> to understand workload distribution</p></li>
<li><p><strong>Tune max_num_tokens</strong> based on your memory constraints and EP size</p></li>
<li><p><strong>Test with representative datasets</strong> to validate load balancing effectiveness</p></li>
</ol>
</section>
<section id="references">
<h3>References<a class="headerlink" href="#references" title="Link to this heading">#</a></h3>
<ul class="simple">
<li><p><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM/blob/main/docs/source/blogs/tech_blog/blog4_Scaling_Expert_Parallelism_in_TensorRT-LLM.md">Technical Blog: Scaling Expert Parallelism in TensorRT LLM</a></p></li>
<li><p><a class="reference external" href="https://arxiv.org/abs/2412.19437">DeepSeek-V3 Paper</a></p></li>
<li><p><a class="reference external" href="https://github.com/deepseek-ai/EPLB">EPLB Implementation</a></p></li>
</ul>
<p>For detailed implementation examples and advanced usage, see:</p>
<ul class="simple">
<li><p><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/wide_ep/"><code class="docutils literal notranslate"><span class="pre">examples/wide_ep/</span></code></a>: Complete Wide-EP examples</p></li>
<li><p><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/wide_ep/ep_load_balancer/"><code class="docutils literal notranslate"><span class="pre">examples/wide_ep/ep_load_balancer/</span></code></a>: Load balancing tools</p></li>
<li><p><a class="reference external" href="https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/wide_ep/slurm_scripts/"><code class="docutils literal notranslate"><span class="pre">examples/wide_ep/slurm_scripts/</span></code></a>: Cluster deployment scripts</p></li>
</ul>
</section>
</section>
</section>
</article>
<footer class="prev-next-footer d-print-none">
<div class="prev-next-area">
<a class="left-prev"
href="paged-attention-ifb-scheduler.html"
title="previous page">
<i class="fa-solid fa-angle-left"></i>
<div class="prev-next-info">
<p class="prev-next-subtitle">previous</p>
<p class="prev-next-title">Paged Attention, IFB, and Request Scheduling</p>
</div>
</a>
<a class="right-next"
href="quantization.html"
title="next page">
<div class="prev-next-info">
<p class="prev-next-subtitle">next</p>
<p class="prev-next-title">Quantization</p>
</div>
<i class="fa-solid fa-angle-right"></i>
</a>
</div>
</footer>
</div>
<dialog id="pst-secondary-sidebar-modal"></dialog>
<div id="pst-secondary-sidebar" class="bd-sidebar-secondary bd-toc"><div class="sidebar-secondary-items sidebar-secondary__inner">
<div class="sidebar-secondary-item">
<div
id="pst-page-navigation-heading-2"
class="page-toc tocsection onthispage">
<i class="fa-solid fa-list"></i> On this page
</div>
<nav class="bd-toc-nav page-toc" aria-labelledby="pst-page-navigation-heading-2">
<ul class="visible nav section-nav flex-column">
<li class="toc-h2 nav-item toc-entry"><a class="reference internal nav-link" href="#overview-of-parallelism-strategies">Overview of Parallelism Strategies</a><ul class="nav section-nav flex-column">
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#tensor-parallelism-tp">Tensor Parallelism (TP)</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#pipeline-parallelism-pp">Pipeline Parallelism (PP)</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#data-parallelism-dp">Data Parallelism (DP)</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#expert-parallelism-ep">Expert Parallelism (EP)</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#context-parallelism-cp">Context Parallelism (CP)</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#wide-expert-parallelism-wide-ep">Wide Expert Parallelism (Wide-EP)</a></li>
</ul>
</li>
<li class="toc-h2 nav-item toc-entry"><a class="reference internal nav-link" href="#module-level-parallelism-guide">Module-level Parallelism Guide</a><ul class="nav section-nav flex-column">
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#attention-module">Attention Module</a><ul class="nav section-nav flex-column">
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#id1">Tensor Parallelism (TP)</a></li>
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#id2">Data Parallelism (DP)</a></li>
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#how-to-enable-attention-parallelism">How to Enable Attention Parallelism</a></li>
</ul>
</li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#ffn-module">FFN Module</a><ul class="nav section-nav flex-column">
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#dense-models">Dense Models</a></li>
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#mixture-of-experts-moe">Mixture of Experts (MoE)</a></li>
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#how-to-enable-moe-parallelism">How to Enable MoE Parallelism</a></li>
</ul>
</li>
</ul>
</li>
<li class="toc-h2 nav-item toc-entry"><a class="reference internal nav-link" href="#id3">Wide Expert Parallelism (Wide-EP)</a><ul class="nav section-nav flex-column">
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#motivation-for-wide-ep">Motivation for Wide-EP</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#key-features-of-wide-ep">Key Features of Wide-EP</a><ul class="nav section-nav flex-column">
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#expert-replication-and-load-balancing">1. Expert Replication and Load Balancing</a></li>
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#custom-ep-communication-kernels">2. Custom EP Communication Kernels</a></li>
<li class="toc-h4 nav-item toc-entry"><a class="reference internal nav-link" href="#expert-parallelism-load-balancer-eplb">3. Expert Parallelism Load Balancer (EPLB)</a></li>
</ul>
</li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#architecture-overview">Architecture Overview</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#best-practices">Best Practices</a></li>
<li class="toc-h3 nav-item toc-entry"><a class="reference internal nav-link" href="#references">References</a></li>
</ul>
</li>
</ul>
</nav></div>
</div></div>
</div>
<footer class="bd-footer-content">
</footer>
</main>
</div>
</div>
<!-- Scripts loaded after <body> so the DOM is not blocked -->
<script defer src="../_static/scripts/bootstrap.js?digest=8878045cc6db502f8baf"></script>
<script defer src="../_static/scripts/pydata-sphinx-theme.js?digest=8878045cc6db502f8baf"></script>
<footer class="bd-footer">
<div class="bd-footer__inner bd-page-width">
<div class="footer-items__start">
<div class="footer-item">
<a class="footer-brand logo" href="https://www.nvidia.com">
<img src="../_static/nvidia-logo-horiz-rgb-1c-blk-for-screen.svg" class="logo__image only-light" alt="NVIDIA"/>
<img src="../_static/nvidia-logo-horiz-rgb-1c-wht-for-screen.svg" class="logo__image only-dark" alt="NVIDIA"/>
</a></div>
<div class="footer-item">
<div class="footer-links">
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/privacy-policy/">Privacy Policy</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/privacy-center/">Your Privacy Choices</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/terms-of-service/">Terms of Service</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/accessibility/">Accessibility</a>
|
<a class="external" href="https://www.nvidia.com/en-us/about-nvidia/company-policies/">Corporate Policies</a>
|
<a class="external" href="https://www.nvidia.com/en-us/product-security/">Product Security</a>
|
<a class="external" href="https://www.nvidia.com/en-us/contact/">Contact</a>
</div>
</div>
<div class="footer-item">
<p class="copyright">
Copyright © 2025, NVidia.
<br/>
</p>
</div>
<div class="footer-item">
<div class="extra_footer">
<p>Last updated on November 23, 2025.</p>
<p>This page is generated by TensorRT-LLM commit <a href="https://github.com/NVIDIA/TensorRT-LLM/tree/a761585">a761585</a>.</p>
</div></div>
</div>
</div>
</footer>
</body>
</html>
Reference in New Issue View Git Blame Copy Permalink