kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-30 15:43:19 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity
f873055b02
TensorRT-LLMs/performance/perf-overview.html
石晓伟 f873055b02
update gh-pages (#2168)
2024-08-30 13:09:14 +08:00

1210 lines
48 KiB
HTML
Raw Blame History

<!DOCTYPE html>
<html class="writer-html5" lang="en" data-content_root="../">
<head>
<meta charset="utf-8" /><meta name="viewport" content="width=device-width, initial-scale=1" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>Overview &mdash; tensorrt_llm documentation</title>
<link rel="stylesheet" type="text/css" href="../_static/pygments.css?v=80d5e7a1" />
<link rel="stylesheet" type="text/css" href="../_static/css/theme.css?v=19f00094" />
<!--[if lt IE 9]>
<script src="../_static/js/html5shiv.min.js"></script>
<![endif]-->
<script src="../_static/jquery.js?v=5d32c60e"></script>
<script src="../_static/_sphinx_javascript_frameworks_compat.js?v=2cd50e6c"></script>
<script src="../_static/documentation_options.js?v=5929fcd5"></script>
<script src="../_static/doctools.js?v=888ff710"></script>
<script src="../_static/sphinx_highlight.js?v=dc90522c"></script>
<script src="../_static/js/theme.js"></script>
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="Best Practices for Tuning the Performance of TensorRT-LLM" href="perf-best-practices.html" />
<link rel="prev" title="Expert Parallelism in TensorRT-LLM" href="../advanced/expert-parallelism.html" />
</head>
<body class="wy-body-for-nav">
<div class="wy-grid-for-nav">
<nav data-toggle="wy-nav-shift" class="wy-nav-side">
<div class="wy-side-scroll">
<div class="wy-side-nav-search" >
<a href="../index.html" class="icon icon-home">
tensorrt_llm
</a>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="../search.html" method="get">
<input type="text" name="q" placeholder="Search docs" aria-label="Search docs" />
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
</div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
<p class="caption" role="heading"><span class="caption-text">Getting Started</span></p>
<ul>
<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"><a class="reference internal" href="../key-features.html">Key Features</a></li>
<li class="toctree-l1"><a class="reference internal" href="../release-notes.html">Release Notes</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Installation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../installation/linux.html">Installing on Linux</a></li>
<li class="toctree-l1"><a class="reference internal" href="../installation/build-from-source-linux.html">Building from Source Code on Linux</a></li>
<li class="toctree-l1"><a class="reference internal" href="../installation/windows.html">Installing on Windows</a></li>
<li class="toctree-l1"><a class="reference internal" href="../installation/build-from-source-windows.html">Building from Source Code on Windows</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">LLM API Examples</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../llm-api-examples/index.html">LLM Examples Introduction</a></li>
<li class="toctree-l1"><a class="reference internal" href="../llm-api-examples/customization.html">Common Customizations</a></li>
<li class="toctree-l1"><a class="reference internal" href="../llm-api-examples/llm_api_examples.html">Examples</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">LLM API</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../llm-api/index.html">API Reference</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Model Definition API</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../python-api/tensorrt_llm.layers.html">Layers</a></li>
<li class="toctree-l1"><a class="reference internal" href="../python-api/tensorrt_llm.functional.html">Functionals</a></li>
<li class="toctree-l1"><a class="reference internal" href="../python-api/tensorrt_llm.models.html">Models</a></li>
<li class="toctree-l1"><a class="reference internal" href="../python-api/tensorrt_llm.plugin.html">Plugin</a></li>
<li class="toctree-l1"><a class="reference internal" href="../python-api/tensorrt_llm.quantization.html">Quantization</a></li>
<li class="toctree-l1"><a class="reference internal" href="../python-api/tensorrt_llm.runtime.html">Runtime</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">C++ API</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_gen/executor.html">Executor</a></li>
<li class="toctree-l1"><a class="reference internal" href="../_cpp_gen/runtime.html">Runtime</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Architecture</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../architecture/overview.html">TensorRT-LLM Architecture</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/core-concepts.html">Model Definition</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/core-concepts.html#compilation">Compilation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/core-concepts.html#runtime">Runtime</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/core-concepts.html#multi-gpu-and-multi-node-support">Multi-GPU and Multi-Node Support</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/checkpoint.html">TensorRT-LLM Checkpoint</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/workflow.html">TensorRT-LLM Build Workflow</a></li>
<li class="toctree-l1"><a class="reference internal" href="../architecture/add-model.html">Adding a Model</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Advanced</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../advanced/gpt-attention.html">Multi-Head, Multi-Query, and Group-Query Attention</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/gpt-runtime.html">C++ GPT Runtime</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/graph-rewriting.html">Graph Rewriting Module</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/batch-manager.html">The Batch Manager in TensorRT-LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/inference-request.html">Inference Request</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/inference-request.html#responses">Responses</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/lora.html">Run gpt-2b + LoRA using GptManager / cpp runtime</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/expert-parallelism.html">Expert Parallelism in TensorRT-LLM</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Performance</span></p>
<ul class="current">
<li class="toctree-l1 current"><a class="current reference internal" href="#">Overview</a><ul>
<li class="toctree-l2"><a class="reference internal" href="#known-issues">Known Issues</a><ul>
<li class="toctree-l3"><a class="reference internal" href="#unexpected-extra-gpu-memory-allocation-when-enabling-multiple-profiles">Unexpected extra GPU memory allocation when enabling <code class="docutils literal notranslate"><span class="pre">--multiple_profiles</span></code></a></li>
<li class="toctree-l3"><a class="reference internal" href="#fused-matmul-gated-silu-llama">Fused Matmul + Gated-SiLU (LLaMA)</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="#throughput-measurements">Throughput Measurements</a></li>
<li class="toctree-l2"><a class="reference internal" href="#reproducing-benchmarked-results">Reproducing Benchmarked Results</a><ul>
<li class="toctree-l3"><a class="reference internal" href="#commands">Commands</a></li>
<li class="toctree-l3"><a class="reference internal" href="#variables">Variables</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="#preparing-a-dataset">Preparing a Dataset</a></li>
<li class="toctree-l2"><a class="reference internal" href="#engine-building">Engine Building</a></li>
<li class="toctree-l2"><a class="reference internal" href="#running-the-benchmark">Running the Benchmark</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="perf-best-practices.html">Best Practices for Tuning the Performance of TensorRT-LLM</a></li>
<li class="toctree-l1"><a class="reference internal" href="perf-analysis.html">Performance Analysis</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Reference</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../reference/troubleshooting.html">Troubleshooting</a></li>
<li class="toctree-l1"><a class="reference internal" href="../reference/support-matrix.html">Support Matrix</a></li>
<li class="toctree-l1"><a class="reference internal" href="../reference/precision.html">Numerical Precision</a></li>
<li class="toctree-l1"><a class="reference internal" href="../reference/memory.html">Memory Usage of TensorRT-LLM</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Blogs</span></p>
<ul>
<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>
<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/Falcon180B-H200.html">Falcon-180B on a single H200 GPU with INT4 AWQ, and 6.7x faster Llama-70B over A100</a></li>
<li class="toctree-l1"><a class="reference internal" href="../blogs/quantization-in-TRT-LLM.html">Speed up inference with SOTA quantization techniques in TRT-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>
</ul>
</div>
</div>
</nav>
<section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
<i data-toggle="wy-nav-top" class="fa fa-bars"></i>
<a href="../index.html">tensorrt_llm</a>
</nav>
<div class="wy-nav-content">
<div class="rst-content">
<div role="navigation" aria-label="Page navigation">
<ul class="wy-breadcrumbs">
<li><a href="../index.html" class="icon icon-home" aria-label="Home"></a></li>
<li class="breadcrumb-item active">Overview</li>
<li class="wy-breadcrumbs-aside">
<a href="../_sources/performance/perf-overview.md.txt" rel="nofollow"> View page source</a>
</li>
</ul>
<hr/>
</div>
<div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
<div itemprop="articleBody">
<blockquote id="perf-overview">
<div><p>[!IMPORTANT]
As of TensorRT-LLM v0.10, these performance benchmarks have changed methodology to utilize in-flight batching and
no longer utilize static benchmarking. These numbers are initial measurements and are expected to improve in future
releases.</p>
</div></blockquote>
<section id="overview">
<h1>Overview<a class="headerlink" href="#overview" title="Link to this heading"></a></h1>
<p>This document summarizes performance measurements of TensorRT-LLM on H100
(Hopper), L40S (Ada) and A100 (Ampere) GPUs for a few key models.</p>
<p>The data in the following tables is provided as a reference point to help users
validate observed performance. It should not be considered as the peak
performance that can be delivered by TensorRT-LLM.</p>
<section id="known-issues">
<h2>Known Issues<a class="headerlink" href="#known-issues" title="Link to this heading"></a></h2>
<p>The following issues are being addressed to improve the efficiency of TensorRT-LLM.</p>
<section id="unexpected-extra-gpu-memory-allocation-when-enabling-multiple-profiles">
<h3>Unexpected extra GPU memory allocation when enabling <code class="docutils literal notranslate"><span class="pre">--multiple_profiles</span></code><a class="headerlink" href="#unexpected-extra-gpu-memory-allocation-when-enabling-multiple-profiles" title="Link to this heading"></a></h3>
<p>We observed that enabling multiple profiles can lead to extra
unexpected GPU memory usage on some cases starting from v0.11.
The issue will be addressed in future releases.</p>
</section>
<section id="fused-matmul-gated-silu-llama">
<h3>Fused Matmul + Gated-SiLU (LLaMA)<a class="headerlink" href="#fused-matmul-gated-silu-llama" title="Link to this heading"></a></h3>
<p>The current implementation combines two Matmul operations into one Matmul followed by
a separate SwiGLU kernel (when <code class="docutils literal notranslate"><span class="pre">--use_fused_mlp</span></code> is enabled). There is also a more
efficient implementation that runs single Matmul + SwiGLU fused kernel for FP8 on Hopper
(when <code class="docutils literal notranslate"><span class="pre">--use_fused_mlp</span> <span class="pre">--gemm_swiglu_plugin</span> <span class="pre">fp8</span></code> is enabled). The gemm_swiglu_plugin
will support more data types and GPU architectures in the future release.</p>
</section>
</section>
<section id="throughput-measurements">
<h2>Throughput Measurements<a class="headerlink" href="#throughput-measurements" title="Link to this heading"></a></h2>
<p>The below table shows performance data where a local inference client is fed requests at an infinite rate (no delay between messages),
and shows the throughput client-server scenario under maximum load.</p>
<p>The performance numbers below were collected using the steps described in this document.</p>
<p><strong>All data in the table below was generated using version 0.12.0 and presents token throughput in tokens/second.</strong></p>
<table class="docutils align-default">
<thead>
<tr class="row-odd"><th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
<th class="head"><p></p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p><strong>GPU</strong></p></td>
<td><p>H200 141GB HBM3</p></td>
<td><p>GH200 120GB</p></td>
<td><p>H100 80GB HBM3</p></td>
<td><p>H100 80GB HBM3</p></td>
<td><p>A100-SXM4-80GB</p></td>
<td><p>L40S</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p><strong>Precision</strong></p></td>
<td><p>FP8</p></td>
<td><p>FP8</p></td>
<td><p>FP8</p></td>
<td><p>Mixed</p></td>
<td><p>Mixed</p></td>
<td><p>FP8</p></td>
</tr>
<tr class="row-even"><td><p><strong>Model</strong></p></td>
<td><p><strong>Input/Output Lengths</strong></p></td>
<td><p><strong>TP</strong></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p>GPTJ 6B</p></td>
<td><p>128/128</p></td>
<td><p>1</p></td>
<td><p>24834.76</p></td>
<td><p>22454.79</p></td>
<td><p>24429.55</p></td>
<td><p>13085.91</p></td>
<td><p>5864.81</p></td>
<td><p>7647.24</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>1</p></td>
<td><p>8348.93</p></td>
<td><p>6656.25</p></td>
<td><p>7831.38</p></td>
<td><p>3882.21</p></td>
<td><p>2194.57</p></td>
<td><p>1843.91</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>1</p></td>
<td><p>5062.80</p></td>
<td><p>3678.91</p></td>
<td><p>3968.98</p></td>
<td><p>2046.53</p></td>
<td><p>1118.22</p></td>
<td><p>980.67</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>1</p></td>
<td><p>2776.53</p></td>
<td><p>2491.03</p></td>
<td><p>2724.38</p></td>
<td><p>1488.56</p></td>
<td><p>657.01</p></td>
<td><p>741.06</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>1</p></td>
<td><p>3631.54</p></td>
<td><p>2994.81</p></td>
<td><p>3004.17</p></td>
<td><p>1280.54</p></td>
<td><p>854.37</p></td>
<td><p>754.16</p></td>
</tr>
<tr class="row-even"><td><p>LLaMA v2 7B</p></td>
<td><p>128/128</p></td>
<td><p>1</p></td>
<td><p>19706.35</p></td>
<td><p>17803.58</p></td>
<td><p>19068.99</p></td>
<td><p>11393.48</p></td>
<td><p>5272.39</p></td>
<td><p>6345.72</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>1</p></td>
<td><p>7651.12</p></td>
<td><p>5472.34</p></td>
<td><p>6610.03</p></td>
<td><p>2964.65</p></td>
<td><p>1785.79</p></td>
<td><p>1551.37</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>1</p></td>
<td><p>4424.90</p></td>
<td><p>3271.61</p></td>
<td><p>3649.38</p></td>
<td><p>1596.87</p></td>
<td><p>957.12</p></td>
<td><p>817.24</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>1</p></td>
<td><p>2385.54</p></td>
<td><p>2035.42</p></td>
<td><p>2271.63</p></td>
<td><p>1189.06</p></td>
<td><p>564.77</p></td>
<td><p>625.09</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>1</p></td>
<td><p>3191.34</p></td>
<td><p>2726.29</p></td>
<td><p>2802.41</p></td>
<td><p>1243.96</p></td>
<td><p>735.19</p></td>
<td><p>641.56</p></td>
</tr>
<tr class="row-odd"><td><p>LLaMA v3 8B</p></td>
<td><p>128/128</p></td>
<td><p>1</p></td>
<td><p>28288.75</p></td>
<td><p>25420.52</p></td>
<td><p>27399.75</p></td>
<td><p>15567.44</p></td>
<td><p>6586.88</p></td>
<td><p>8745.80</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>1</p></td>
<td><p>23230.62</p></td>
<td><p>16426.68</p></td>
<td><p>19198.73</p></td>
<td><p>8817.39</p></td>
<td><p>4882.13</p></td>
<td><p>5084.49</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>1</p></td>
<td><p>16144.44</p></td>
<td><p>9832.66</p></td>
<td><p>12084.97</p></td>
<td><p>5352.37</p></td>
<td><p>3079.90</p></td>
<td><p>2755.13</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>1</p></td>
<td><p>3623.79</p></td>
<td><p>3290.22</p></td>
<td><p>3463.26</p></td>
<td><p>1852.48</p></td>
<td><p>781.63</p></td>
<td><p>980.86</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>1</p></td>
<td><p>11093.62</p></td>
<td><p>7573.35</p></td>
<td><p>8894.11</p></td>
<td><p>3986.83</p></td>
<td><p>2268.13</p></td>
<td><p>2051.79</p></td>
</tr>
<tr class="row-even"><td><p>Mistral 7B</p></td>
<td><p>128/128</p></td>
<td><p>1</p></td>
<td><p>30223.01</p></td>
<td><p>27696.90</p></td>
<td><p>29788.46</p></td>
<td><p>16319.25</p></td>
<td><p>6807.02</p></td>
<td><p>9612.58</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>1</p></td>
<td><p>24989.54</p></td>
<td><p>17942.29</p></td>
<td><p>20509.72</p></td>
<td><p>9982.01</p></td>
<td><p>5296.02</p></td>
<td><p>5444.89</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>1</p></td>
<td><p>17036.14</p></td>
<td><p>10846.03</p></td>
<td><p>12807.80</p></td>
<td><p>5718.89</p></td>
<td><p>3241.33</p></td>
<td><p>2931.17</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>1</p></td>
<td><p>3678.80</p></td>
<td><p>3294.02</p></td>
<td><p>3521.71</p></td>
<td><p>1887.75</p></td>
<td><p>786.43</p></td>
<td><p>1002.49</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>1</p></td>
<td><p>11510.54</p></td>
<td><p>8357.75</p></td>
<td><p>9214.61</p></td>
<td><p>4284.82</p></td>
<td><p>2363.25</p></td>
<td><p>2154.26</p></td>
</tr>
<tr class="row-odd"><td><p>Mixtral 8x7B</p></td>
<td><p>128/128</p></td>
<td><p>2</p></td>
<td><p>24895.03</p></td>
<td><p>8785.80</p></td>
<td><p>24394.71</p></td>
<td><p>15529.86</p></td>
<td><p>5921.41</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>42014.24</p></td>
<td><p>38828.53</p></td>
<td><p>40197.42</p></td>
<td><p>28132.17</p></td>
<td><p>11414.95</p></td>
<td><p>6820.26</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>2</p></td>
<td><p>29389.21</p></td>
<td><p>5474.69</p></td>
<td><p>20873.02</p></td>
<td><p>7066.02</p></td>
<td><p>4306.98</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>52348.10</p></td>
<td><p>41573.66</p></td>
<td><p>40588.05</p></td>
<td><p>21285.72</p></td>
<td><p>10974.83</p></td>
<td><p>7467.15</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>2</p></td>
<td><p>21480.27</p></td>
<td><p>2277.66</p></td>
<td><p>12838.28</p></td>
<td><p>3986.01</p></td>
<td><p>2400.11</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>39182.04</p></td>
<td><p>28626.55</p></td>
<td><p>28337.31</p></td>
<td><p>12447.13</p></td>
<td><p>7278.89</p></td>
<td><p>5233.43</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>2</p></td>
<td><p>2934.44</p></td>
<td><p>1003.51</p></td>
<td><p>2898.27</p></td>
<td><p>1834.77</p></td>
<td><p>693.51</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>5152.40</p></td>
<td><p>4724.01</p></td>
<td><p>5028.61</p></td>
<td><p>3393.18</p></td>
<td><p>1362.93</p></td>
<td><p>805.49</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>2</p></td>
<td><p>14029.17</p></td>
<td><p>2671.88</p></td>
<td><p>10479.45</p></td>
<td><p>3531.31</p></td>
<td><p>1945.88</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>25436.05</p></td>
<td><p>20302.56</p></td>
<td><p>19971.72</p></td>
<td><p>9622.66</p></td>
<td><p>5221.74</p></td>
<td><p>3616.30</p></td>
</tr>
<tr class="row-odd"><td><p>LLaMA v3 70B</p></td>
<td><p>128/128</p></td>
<td><p>2</p></td>
<td><p>5386.88</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>2959.22</p></td>
<td><p>1301.14</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>8944.26</p></td>
<td><p>8587.01</p></td>
<td><p>8642.05</p></td>
<td><p>5966.47</p></td>
<td><p>2413.95</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>16125.20</p></td>
<td><p></p></td>
<td><p>15397.47</p></td>
<td><p>10406.55</p></td>
<td><p>4548.32</p></td>
<td><p>1364.08</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>2</p></td>
<td><p>7007.27</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>720.73</p></td>
<td><p>500.83</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>12906.75</p></td>
<td><p>10761.53</p></td>
<td><p>8978.95</p></td>
<td><p>4736.61</p></td>
<td><p>2380.02</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>19417.37</p></td>
<td><p></p></td>
<td><p>14822.93</p></td>
<td><p>6672.14</p></td>
<td><p>3815.08</p></td>
<td><p>1809.40</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>2</p></td>
<td><p>6183.85</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>369.29</p></td>
<td><p>251.24</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>8859.54</p></td>
<td><p>7270.77</p></td>
<td><p>6073.48</p></td>
<td><p>2969.99</p></td>
<td><p>1634.82</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>13969.95</p></td>
<td><p></p></td>
<td><p>10094.57</p></td>
<td><p>4358.77</p></td>
<td><p>2847.54</p></td>
<td><p>1313.78</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>2</p></td>
<td><p>696.59</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>301.46</p></td>
<td><p>140.88</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>1044.35</p></td>
<td><p>1000.55</p></td>
<td><p>1022.06</p></td>
<td><p>681.72</p></td>
<td><p>278.76</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>2018.47</p></td>
<td><p></p></td>
<td><p>1933.15</p></td>
<td><p>1279.46</p></td>
<td><p>543.73</p></td>
<td><p>163.36</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>2</p></td>
<td><p>3525.18</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>87.54</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>6550.76</p></td>
<td><p>4859.38</p></td>
<td><p>4870.26</p></td>
<td><p>2379.66</p></td>
<td><p>1209.69</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>9706.95</p></td>
<td><p></p></td>
<td><p>7670.04</p></td>
<td><p>3692.41</p></td>
<td><p>2192.28</p></td>
<td><p>895.23</p></td>
</tr>
<tr class="row-even"><td><p>LLaMA v2 70B</p></td>
<td><p>128/128</p></td>
<td><p>2</p></td>
<td><p>6355.16</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>2927.71</p></td>
<td><p>1374.05</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>10818.97</p></td>
<td><p>10819.19</p></td>
<td><p>10754.99</p></td>
<td><p>6603.10</p></td>
<td><p>2765.94</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>16667.25</p></td>
<td><p></p></td>
<td><p>16074.84</p></td>
<td><p>11369.11</p></td>
<td><p>4796.89</p></td>
<td><p>1402.92</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>2</p></td>
<td><p>6185.77</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>668.52</p></td>
<td><p>445.04</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>12884.76</p></td>
<td><p>11356.48</p></td>
<td><p>8870.71</p></td>
<td><p>5067.06</p></td>
<td><p>2710.53</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>19053.13</p></td>
<td><p></p></td>
<td><p>17534.62</p></td>
<td><p>8805.16</p></td>
<td><p>5665.93</p></td>
<td><p>2203.33</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>2</p></td>
<td><p>4873.24</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>334.10</p></td>
<td><p>215.70</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>8664.90</p></td>
<td><p>6311.85</p></td>
<td><p>7564.99</p></td>
<td><p>3354.02</p></td>
<td><p>1884.46</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>15110.32</p></td>
<td><p></p></td>
<td><p>10584.03</p></td>
<td><p>5373.10</p></td>
<td><p>3672.80</p></td>
<td><p>1787.76</p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>2</p></td>
<td><p>732.09</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>302.49</p></td>
<td><p>141.70</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>1272.90</p></td>
<td><p>1269.58</p></td>
<td><p>1265.80</p></td>
<td><p>774.93</p></td>
<td><p>320.79</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>2015.77</p></td>
<td><p></p></td>
<td><p>1943.96</p></td>
<td><p>1355.78</p></td>
<td><p>569.48</p></td>
<td><p>165.52</p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>2</p></td>
<td><p>3508.50</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p>321.95</p></td>
<td><p>212.97</p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p></p></td>
<td><p>4</p></td>
<td><p>6642.69</p></td>
<td><p>5545.83</p></td>
<td><p>4889.26</p></td>
<td><p>2439.10</p></td>
<td><p>1276.58</p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>10178.71</p></td>
<td><p></p></td>
<td><p>8071.77</p></td>
<td><p>4275.74</p></td>
<td><p>2589.60</p></td>
<td><p>1083.45</p></td>
</tr>
<tr class="row-odd"><td><p>Falcon 180B</p></td>
<td><p>128/128</p></td>
<td><p>4</p></td>
<td><p>5129.55</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>8370.98</p></td>
<td><p></p></td>
<td><p>8268.72</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/2048</p></td>
<td><p>4</p></td>
<td><p>7823.79</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>13278.59</p></td>
<td><p></p></td>
<td><p>13107.48</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>128/4096</p></td>
<td><p>4</p></td>
<td><p>6374.10</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>12660.89</p></td>
<td><p></p></td>
<td><p>10493.79</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/128</p></td>
<td><p>4</p></td>
<td><p>601.67</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>1002.57</p></td>
<td><p></p></td>
<td><p>991.22</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p></p></td>
<td><p>2048/2048</p></td>
<td><p>4</p></td>
<td><p>3869.76</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-even"><td><p></p></td>
<td><p></p></td>
<td><p>8</p></td>
<td><p>7134.33</p></td>
<td><p></p></td>
<td><p>6386.83</p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p><em>TP stands for Tensor Parallelism</em></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
<td><p></p></td>
</tr>
</tbody>
</table>
</section>
<section id="reproducing-benchmarked-results">
<h2>Reproducing Benchmarked Results<a class="headerlink" href="#reproducing-benchmarked-results" title="Link to this heading"></a></h2>
<blockquote>
<div><p>[!NOTE] The only models supported in this workflow are those listed in the table above.</p>
</div></blockquote>
<p>The following tables are references for commands that are used as part of the benchmarking process. For a more detailed
description of this benchmarking workflow, see the <span class="xref myst">Benchmarking Suite README</span>.</p>
<section id="commands">
<h3>Commands<a class="headerlink" href="#commands" title="Link to this heading"></a></h3>
<table class="docutils align-default">
<thead>
<tr class="row-odd"><th class="head text-left"><p>Stage</p></th>
<th class="head"><p>Description</p></th>
<th class="head"><p>Command</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td class="text-left"><p><a class="reference internal" href="#preparing-a-dataset"><span class="xref myst">Dataset</span></a></p></td>
<td><p>Create a synthetic dataset</p></td>
<td><p><code class="docutils literal notranslate"><span class="pre">python</span> <span class="pre">benchmarks/cpp/prepare_dataset.py</span> <span class="pre">--tokenizer=$model_name</span> <span class="pre">--stdout</span> <span class="pre">token-norm-dist</span> <span class="pre">--num-requests=$num_requests</span> <span class="pre">--input-mean=$isl</span> <span class="pre">--output-mean=$osl</span> <span class="pre">--input-stdev=0</span> <span class="pre">--output-stdev=0</span> <span class="pre">&gt;</span> <span class="pre">$dataset_file</span></code></p></td>
</tr>
<tr class="row-odd"><td class="text-left"><p><a class="reference internal" href="#engine-building"><span class="xref myst">Build</span></a></p></td>
<td><p>Build a TensorRT-LLM engine</p></td>
<td><p><code class="docutils literal notranslate"><span class="pre">trtllm-bench</span> <span class="pre">--model</span> <span class="pre">$model_name</span> <span class="pre">build</span> <span class="pre">--tp_size</span> <span class="pre">$tp_size</span> <span class="pre">--quantization</span> <span class="pre">FP8</span> <span class="pre">--dataset</span> <span class="pre">$dataset_file</span></code></p></td>
</tr>
<tr class="row-even"><td class="text-left"><p><a class="reference internal" href="#running-the-benchmark"><span class="xref myst">Run</span></a></p></td>
<td><p>Run a benchmark with a dataset</p></td>
<td><p><code class="docutils literal notranslate"><span class="pre">trtllm-bench</span> <span class="pre">--model</span> <span class="pre">$model_name</span> <span class="pre">throughput</span> <span class="pre">--dataset</span> <span class="pre">$dataset_file</span> <span class="pre">--engine_dir</span> <span class="pre">$engine_dir</span></code></p></td>
</tr>
</tbody>
</table>
</section>
<section id="variables">
<h3>Variables<a class="headerlink" href="#variables" title="Link to this heading"></a></h3>
<table class="docutils align-default">
<thead>
<tr class="row-odd"><th class="head text-left"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$isl</span></code></p></td>
<td><p>Benchmark input sequence length.</p></td>
</tr>
<tr class="row-odd"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$osl</span></code></p></td>
<td><p>Benchmark output sequence length.</p></td>
</tr>
<tr class="row-even"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$tp_size</span></code></p></td>
<td><p>Number of GPUs to run the benchmark with</p></td>
</tr>
<tr class="row-odd"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$engine_dir</span></code></p></td>
<td><p>Location to store built engine file (can be deleted after running benchmarks).</p></td>
</tr>
<tr class="row-even"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$model_name</span></code></p></td>
<td><p>HuggingFace model name eg. meta-llama/Llama-2-7b-hf or use the path to a local weights directory</p></td>
</tr>
<tr class="row-odd"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$dataset_file</span></code></p></td>
<td><p>Location of the dataset file generated by <code class="docutils literal notranslate"><span class="pre">prepare_dataset.py</span></code></p></td>
</tr>
<tr class="row-even"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$num_requests</span></code></p></td>
<td><p>The number of requests to generate for dataset generation</p></td>
</tr>
<tr class="row-odd"><td class="text-left"><p><code class="docutils literal notranslate"><span class="pre">$seq_len</span></code></p></td>
<td><p>A sequence length of ISL + OSL</p></td>
</tr>
</tbody>
</table>
</section>
</section>
<section id="preparing-a-dataset">
<h2>Preparing a Dataset<a class="headerlink" href="#preparing-a-dataset" title="Link to this heading"></a></h2>
<p>In order to prepare a dataset, you can use the provided <a class="reference download internal" download="" href="../_downloads/ea8faa5e98124e92f96b66dc586fb429/prepare_dataset.py"><span class="xref download myst">script</span></a>.
To generate a synthetic dataset, run the following command:</p>
<div class="highlight-shell notranslate"><div class="highlight"><pre><span></span>python<span class="w"> </span>benchmarks/cpp/prepare_dataset.py<span class="w"> </span>--output<span class="o">=</span><span class="nv">$dataset_file</span><span class="w"> </span>--tokenizer<span class="o">=</span><span class="nv">$model_name</span><span class="w"> </span>token-norm-dist<span class="w"> </span>--num-requests<span class="o">=</span><span class="nv">$num_requests</span><span class="w"> </span>--input-mean<span class="o">=</span><span class="nv">$isl</span><span class="w"> </span>--output-mean<span class="o">=</span><span class="nv">$osl</span><span class="w"> </span>--input-stdev<span class="o">=</span><span class="m">0</span><span class="w"> </span>--output-stdev<span class="o">=</span><span class="m">0</span><span class="w"> </span>&gt;<span class="w"> </span><span class="nv">$dataset_file</span>
</pre></div>
</div>
<p>The command will generate a text file located at the path specified <code class="docutils literal notranslate"><span class="pre">$dataset_file</span></code> where all requests are of the same
input/output sequence length combinations. The script works by using the tokenizer to retrieve the vocabulary size and
randomly sample token IDs from it to create entirely random sequences. In the command above, all requests will be uniform
because the standard deviations for both input and output sequences are set to 0.</p>
<p>For each input and output sequence length combination, the table below details the <code class="docutils literal notranslate"><span class="pre">$num_requests</span></code> that were used. For
shorter input and output lengths, a larger number of messages were used to guarantee that the system hit a steady state
because requests enter and exit the system at a much faster rate. For longer input/output sequence lengths, requests
remain in the system longer and therefore require less requests to achieve steady state.</p>
<table class="docutils align-default">
<thead>
<tr class="row-odd"><th class="head"><p>Input Length</p></th>
<th class="head"><p>Output Length</p></th>
<th class="head"><p>$seq_len</p></th>
<th class="head"><p>$num_requests</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>128</p></td>
<td><p>128</p></td>
<td><p>256</p></td>
<td><p>30000</p></td>
</tr>
<tr class="row-odd"><td><p>128</p></td>
<td><p>2048</p></td>
<td><p>2176</p></td>
<td><p>3000</p></td>
</tr>
<tr class="row-even"><td><p>128</p></td>
<td><p>4096</p></td>
<td><p>4224</p></td>
<td><p>1500</p></td>
</tr>
<tr class="row-odd"><td><p>2048</p></td>
<td><p>128</p></td>
<td><p>2176</p></td>
<td><p>3000</p></td>
</tr>
<tr class="row-even"><td><p>2048</p></td>
<td><p>2048</p></td>
<td><p>4096</p></td>
<td><p>1500</p></td>
</tr>
</tbody>
</table>
</section>
<section id="engine-building">
<h2>Engine Building<a class="headerlink" href="#engine-building" title="Link to this heading"></a></h2>
<p>All engines are built using the <code class="docutils literal notranslate"><span class="pre">trtllm-bench</span> <span class="pre">build</span></code> sub-command. The basic command for FP8 quantized engines is as follows:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span>trtllm-bench --model $model_name build --tp_size $tp_size --quantization FP8 --dataset $dataset_file
</pre></div>
</div>
<p>or if you would like to build for a specific sequence length:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span>trtllm-bench --model $model_name build --tp_size $tp_size --quantization FP8 --max_seq_length $seq_len
</pre></div>
</div>
<p>If you would like to build an FP16 engine without any quantization, simply remove the <code class="docutils literal notranslate"><span class="pre">--quantization</span> <span class="pre">FP8</span></code> option.</p>
<blockquote>
<div><p>[!NOTE] If you specify FP8 quantization, the KV cache will automatically be set to FP8 as well!</p>
</div></blockquote>
<p>The <code class="docutils literal notranslate"><span class="pre">trtllm-bench</span> <span class="pre">build</span></code> sub-command will output the path where the engine is located upon a successful build. For example,</p>
<div class="highlight-shell notranslate"><div class="highlight"><pre><span></span><span class="o">===========================================================</span>
ENGINE<span class="w"> </span>SAVED:<span class="w"> </span>/tmp/meta-llama/Llama-2-7b-hf/tp_1_pp_1
<span class="o">===========================================================</span>
</pre></div>
</div>
</section>
<section id="running-the-benchmark">
<h2>Running the Benchmark<a class="headerlink" href="#running-the-benchmark" title="Link to this heading"></a></h2>
<p>To run the benchmark with the generated data set, simply use the <code class="docutils literal notranslate"><span class="pre">trtllm-bench</span> <span class="pre">throughput</span></code> sub-command. The benchmarker will
run an offline maximum throughput scenario such that all requests are queued in rapid succession. You simply need to provide
the patch to the engine from the <a class="reference internal" href="#engine-building"><span class="xref myst">build</span></a> phase and a <a class="reference internal" href="#preparing-a-dataset"><span class="xref myst">generated dataset</span></a>.</p>
<div class="highlight-shell notranslate"><div class="highlight"><pre><span></span>trtllm-bench<span class="w"> </span>--model<span class="w"> </span><span class="nv">$model_name</span><span class="w"> </span>throughput<span class="w"> </span>--dataset<span class="w"> </span><span class="nv">$dataset_file</span><span class="w"> </span>--engine_dir<span class="w"> </span><span class="nv">$engine_dir</span>
</pre></div>
</div>
<p>The results will be printed to the terminal upon benchmark completion. For example,</p>
<div class="highlight-shell notranslate"><div class="highlight"><pre><span></span><span class="o">===========================================================</span>
<span class="o">=</span><span class="w"> </span>ENGINE<span class="w"> </span><span class="nv">DETAILS</span>
<span class="o">===========================================================</span>
Model:<span class="w"> </span>meta-llama/Llama-2-7b-hf
Engine<span class="w"> </span>Directory:<span class="w"> </span>/tmp/meta-llama/Llama-2-7b-hf/tp_1_pp_1
TensorRT-LLM<span class="w"> </span>Version:<span class="w"> </span><span class="m">0</span>.12.0
Dtype:<span class="w"> </span>float16
KV<span class="w"> </span>Cache<span class="w"> </span>Dtype:<span class="w"> </span>FP8
Quantization:<span class="w"> </span>FP8
Max<span class="w"> </span>Input<span class="w"> </span>Length:<span class="w"> </span><span class="m">2048</span>
Max<span class="w"> </span>Sequence<span class="w"> </span>Length:<span class="w"> </span><span class="nv">4098</span>
<span class="o">===========================================================</span>
<span class="o">=</span><span class="w"> </span>WORLD<span class="w"> </span>+<span class="w"> </span>RUNTIME<span class="w"> </span><span class="nv">INFORMATION</span>
<span class="o">===========================================================</span>
TP<span class="w"> </span>Size:<span class="w"> </span><span class="m">1</span>
PP<span class="w"> </span>Size:<span class="w"> </span><span class="m">1</span>
Max<span class="w"> </span>Runtime<span class="w"> </span>Batch<span class="w"> </span>Size:<span class="w"> </span><span class="m">4096</span>
Max<span class="w"> </span>Runtime<span class="w"> </span>Tokens:<span class="w"> </span><span class="m">8192</span>
Scheduling<span class="w"> </span>Policy:<span class="w"> </span>Guaranteed<span class="w"> </span>No<span class="w"> </span>Evict
KV<span class="w"> </span>Memory<span class="w"> </span>Percentage:<span class="w"> </span><span class="m">99</span>.0%
Issue<span class="w"> </span>Rate<span class="w"> </span><span class="o">(</span>req/sec<span class="o">)</span>:<span class="w"> </span><span class="m">3</span>.680275266452667e+18
<span class="o">===========================================================</span>
<span class="o">=</span><span class="w"> </span><span class="nv">STATISTICS</span>
<span class="o">===========================================================</span>
Number<span class="w"> </span>of<span class="w"> </span>requests:<span class="w"> </span><span class="m">3000</span>
Average<span class="w"> </span>Input<span class="w"> </span>Length<span class="w"> </span><span class="o">(</span>tokens<span class="o">)</span>:<span class="w"> </span><span class="m">128</span>.0
Average<span class="w"> </span>Output<span class="w"> </span>Length<span class="w"> </span><span class="o">(</span>tokens<span class="o">)</span>:<span class="w"> </span><span class="m">128</span>.0
Token<span class="w"> </span>Throughput<span class="w"> </span><span class="o">(</span>tokens/sec<span class="o">)</span>:<span class="w"> </span><span class="m">23405</span>.927228471104
Request<span class="w"> </span>Throughput<span class="w"> </span><span class="o">(</span>req/sec<span class="o">)</span>:<span class="w"> </span><span class="m">182</span>.8588064724305
Total<span class="w"> </span>Latency<span class="w"> </span><span class="o">(</span>seconds<span class="o">)</span>:<span class="w"> </span><span class="m">16</span>.406100739
<span class="o">===========================================================</span>
</pre></div>
</div>
<blockquote>
<div><p>[!WARNING] In some cases, the benchmarker may not print anything at all. This behavior usually
means that the benchmark has hit an out of memory issue. Try reducing the KV cache percentage
using the <code class="docutils literal notranslate"><span class="pre">--kv_cache_free_gpu_mem_fraction</span></code> option to lower the percentage of used memory.</p>
</div></blockquote>
</section>
</section>
</div>
</div>
<footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
<a href="../advanced/expert-parallelism.html" class="btn btn-neutral float-left" title="Expert Parallelism in TensorRT-LLM" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
<a href="perf-best-practices.html" class="btn btn-neutral float-right" title="Best Practices for Tuning the Performance of TensorRT-LLM" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
</div>
<hr/>
<div role="contentinfo">
<jinja2.runtime.BlockReference object at 0x7f4253467580>
<div class="footer">
<p>
Copyright © 2024 NVIDIA Corporation
</p>
<p>
<a class="Link" href="https://www.nvidia.com/en-us/about-nvidia/privacy-policy/" target="_blank" rel="noopener"
data-cms-ai="0">Privacy Policy</a> |
<a class="Link" href="https://www.nvidia.com/en-us/about-nvidia/privacy-center/" target="_blank" rel="noopener"
data-cms-ai="0">Manage My Privacy</a> |
<a class="Link" href="https://www.nvidia.com/en-us/preferences/start/" target="_blank" rel="noopener"
data-cms-ai="0">Do Not Sell or Share My Data</a> |
<a class="Link" href="https://www.nvidia.com/en-us/about-nvidia/terms-of-service/" target="_blank"
rel="noopener" data-cms-ai="0">Terms of Service</a> |
<a class="Link" href="https://www.nvidia.com/en-us/about-nvidia/accessibility/" target="_blank" rel="noopener"
data-cms-ai="0">Accessibility</a> |
<a class="Link" href="https://www.nvidia.com/en-us/about-nvidia/company-policies/" target="_blank"
rel="noopener" data-cms-ai="0">Corporate Policies</a> |
<a class="Link" href="https://www.nvidia.com/en-us/product-security/" target="_blank" rel="noopener"
data-cms-ai="0">Product Security</a> |
<a class="Link" href="https://www.nvidia.com/en-us/contact/" target="_blank" rel="noopener"
data-cms-ai="0">Contact</a>
</p>
</div>
</div>
</footer>
</div>
</div>
</section>
</div>
<script>
jQuery(function () {
SphinxRtdTheme.Navigation.enable(true);
});
</script>
</body>
</html>
Reference in New Issue View Git Blame Copy Permalink