| |
| |
| |
| |
| |
| |
| <!DOCTYPE html> |
| <html class="writer-html5" lang="en" > |
| <head> |
| <meta charset="utf-8"> |
| |
| <meta name="viewport" content="width=device-width, initial-scale=1.0"> |
| |
| <title>Deploy a Framework-prequantized Model with TVM — tvm 0.18.dev0 documentation</title> |
| |
| |
| |
| <link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/css/bootstrap.min.css" integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous"> |
| <link rel="stylesheet" href="../../_static/css/theme.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/pygments.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/css/theme.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/sg_gallery.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/sg_gallery-binder.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/sg_gallery-dataframe.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/sg_gallery-rendered-html.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/pygments.css" type="text/css" /> |
| <link rel="stylesheet" href="../../_static/css/tlcpack_theme.css" type="text/css" /> |
| |
| |
| |
| <link rel="shortcut icon" href="../../_static/tvm-logo-square.png"/> |
| |
| |
| |
| |
| |
| |
| |
| <script type="text/javascript" id="documentation_options" data-url_root="../../" src="../../_static/documentation_options.js"></script> |
| <script data-url_root="../../" id="documentation_options" src="../../_static/documentation_options.js"></script> |
| <script src="../../_static/jquery.js"></script> |
| <script src="../../_static/underscore.js"></script> |
| <script src="../../_static/doctools.js"></script> |
| |
| <script type="text/javascript" src="../../_static/js/theme.js"></script> |
| |
| |
| <script type="text/javascript" src="../../_static/js/tlcpack_theme.js"></script> |
| <link rel="index" title="Index" href="../../genindex.html" /> |
| <link rel="search" title="Search" href="../../search.html" /> |
| <link rel="next" title="Deploy a Framework-prequantized Model with TVM - Part 3 (TFLite)" href="deploy_prequantized_tflite.html" /> |
| <link rel="prev" title="Compile PyTorch Object Detection Models" href="deploy_object_detection_pytorch.html" /> |
| </head> |
| |
| <body class="wy-body-for-nav"> |
| |
| |
| <div class="wy-grid-for-nav"> |
| |
| |
| <header class="header"> |
| <div class="innercontainer"> |
| <div class="headerInner d-flex justify-content-between align-items-center"> |
| <div class="headerLogo"> |
| <a href="https://tvm.apache.org/"><img src=https://tvm.apache.org/assets/images/logo.svg alt="logo"></a> |
| </div> |
| |
| <div id="headMenu" class="headerNav"> |
| <button type="button" id="closeHeadMenu" class="navCloseBtn"><img src="../../_static/img/close-icon.svg" alt="Close"></button> |
| <ul class="nav"> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://tvm.apache.org/community>Community</a> |
| </li> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://tvm.apache.org/download>Download</a> |
| </li> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://tvm.apache.org/vta>VTA</a> |
| </li> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://tvm.apache.org/blog>Blog</a> |
| </li> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://tvm.apache.org/docs>Docs</a> |
| </li> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://tvmconf.org>Conference</a> |
| </li> |
| <li class="nav-item"> |
| <a class="nav-link" href=https://github.com/apache/tvm/>Github</a> |
| </li> |
| </ul> |
| <div class="responsivetlcdropdown"> |
| <button type="button" class="btn-link"> |
| ASF |
| </button> |
| <ul> |
| <li> |
| <a href=https://apache.org/>Apache Homepage</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/licenses/>License</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/foundation/sponsorship.html>Sponsorship</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/security/>Security</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/foundation/thanks.html>Thanks</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/events/current-event>Events</a> |
| </li> |
| </ul> |
| </div> |
| </div> |
| <div class="responsiveMenuIcon"> |
| <button type="button" id="menuBtn" class="btn-menu"><img src="../../_static/img/menu-icon.svg" alt="Menu Icon"></button> |
| </div> |
| |
| <div class="tlcDropdown"> |
| <div class="dropdown"> |
| <button type="button" class="btn-link dropdown-toggle" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false"> |
| ASF |
| </button> |
| <div class="dropdown-menu dropdown-menu-right"> |
| <ul> |
| <li> |
| <a href=https://apache.org/>Apache Homepage</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/licenses/>License</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/foundation/sponsorship.html>Sponsorship</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/security/>Security</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/foundation/thanks.html>Thanks</a> |
| </li> |
| <li> |
| <a href=https://www.apache.org/events/current-event>Events</a> |
| </li> |
| </ul> |
| </div> |
| </div> |
| </div> |
| </div> |
| </div> |
| </header> |
| |
| <nav data-toggle="wy-nav-shift" class="wy-nav-side fixed"> |
| <div class="wy-side-scroll"> |
| <div class="wy-side-nav-search" > |
| |
| |
| |
| <a href="../../index.html"> |
| |
| |
| |
| |
| <img src="../../_static/tvm-logo-small.png" class="logo" alt="Logo"/> |
| |
| </a> |
| |
| |
| |
| |
| <input type="checkbox" class="version-toggle-box" hidden id="version-toggle"> |
| <label for="version-toggle" class="version-toggle-label"> |
| <div tabindex="0" class="version version-selector version-selector-show"> |
| 0.18.dev0 <span class="chevron versions-hidden"><svg fill="none" height="24" viewBox="0 0 24 24" width="24" xmlns="http://www.w3.org/2000/svg"><path d="m8 4 8 8-8 8" stroke="#000" stroke-linecap="round" stroke-linejoin="round" stroke-width="2"/></svg></span><span class="chevron versions-shown"><svg fill="none" height="24" viewBox="0 0 24 24" width="24" xmlns="http://www.w3.org/2000/svg"><path d="m4 8 8 8 8-8" stroke="#000" stroke-linecap="round" stroke-linejoin="round" stroke-width="2"/></svg></span> |
| </div> |
| </label> |
| <div class="version-details wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation"> |
| <p class="caption" role="heading"><span class="caption-text">Versions</span></p> |
| <ol style="text-align: left"> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="/">0.18.dev0 (main)</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.8.0/">v0.8.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.9.0/">v0.9.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.10.0/">v0.10.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.11.0/">v0.11.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.12.0/">v0.12.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.13.0/">v0.13.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.14.0/">v0.14.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.15.0/">v0.15.0</a></div></li> |
| |
| |
| |
| |
| <li><div class="version"><a style="font-size: 0.8em; padding: 4px" href="v0.16.0/">v0.16.0</a></div></li> |
| |
| </ol> |
| </div> |
| |
| |
| |
| |
| <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="main navigation"> |
| |
| |
| |
| |
| |
| |
| <p class="caption" role="heading"><span class="caption-text">Getting Started</span></p> |
| <ul> |
| <li class="toctree-l1"><a class="reference internal" href="../../install/index.html">Installing TVM</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../contribute/index.html">Contributor Guide</a></li> |
| </ul> |
| <p class="caption" role="heading"><span class="caption-text">User Guide</span></p> |
| <ul class="current"> |
| <li class="toctree-l1"><a class="reference internal" href="../../tutorial/index.html">User Tutorial</a></li> |
| <li class="toctree-l1 current"><a class="reference internal" href="../index.html">How To Guides</a><ul class="current"> |
| <li class="toctree-l2"><a class="reference internal" href="../compile_models/index.html">Compile Deep Learning Models</a></li> |
| <li class="toctree-l2 current"><a class="reference internal" href="../deploy/index.html">Deploy Models and Integrate TVM</a><ul class="current"> |
| <li class="toctree-l3"><a class="reference internal" href="../deploy/index.html#build-the-tvm-runtime-library">Build the TVM runtime library</a></li> |
| <li class="toctree-l3"><a class="reference internal" href="../deploy/index.html#cross-compile-the-tvm-runtime-for-other-architectures">Cross compile the TVM runtime for other architectures</a></li> |
| <li class="toctree-l3"><a class="reference internal" href="../deploy/index.html#optimize-and-tune-models-for-target-devices">Optimize and tune models for target devices</a></li> |
| <li class="toctree-l3"><a class="reference internal" href="../deploy/index.html#deploy-optimized-model-on-target-devices">Deploy optimized model on target devices</a></li> |
| <li class="toctree-l3 current"><a class="reference internal" href="../deploy/index.html#additional-deployment-how-tos">Additional Deployment How-Tos</a><ul class="current"> |
| <li class="toctree-l4 current"><a class="reference internal" href="index.html">Deploy Deep Learning Models</a></li> |
| </ul> |
| </li> |
| </ul> |
| </li> |
| <li class="toctree-l2"><a class="reference internal" href="../work_with_relay/index.html">Work With Relay</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../work_with_schedules/index.html">Work With Tensor Expression and Schedules</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../optimize_operators/index.html">Optimize Tensor Operators</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../tune_with_autotvm/index.html">Auto-Tune with Templates and AutoTVM</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../tune_with_autoscheduler/index.html">Use AutoScheduler for Template-Free Scheduling</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../work_with_microtvm/index.html">Work With microTVM</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../extend_tvm/index.html">Extend TVM</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../profile/index.html">Profile Models</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../../errors.html">Handle TVM Errors</a></li> |
| <li class="toctree-l2"><a class="reference internal" href="../../faq.html">Frequently Asked Questions</a></li> |
| </ul> |
| </li> |
| </ul> |
| <p class="caption" role="heading"><span class="caption-text">Developer Guide</span></p> |
| <ul> |
| <li class="toctree-l1"><a class="reference internal" href="../../dev/tutorial/index.html">Developer Tutorial</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../dev/how_to/how_to.html">Developer How-To Guide</a></li> |
| </ul> |
| <p class="caption" role="heading"><span class="caption-text">Architecture Guide</span></p> |
| <ul> |
| <li class="toctree-l1"><a class="reference internal" href="../../arch/index.html">Design and Architecture</a></li> |
| </ul> |
| <p class="caption" role="heading"><span class="caption-text">Topic Guides</span></p> |
| <ul> |
| <li class="toctree-l1"><a class="reference internal" href="../../topic/microtvm/index.html">microTVM: TVM on bare-metal</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../topic/vta/index.html">VTA: Versatile Tensor Accelerator</a></li> |
| </ul> |
| <p class="caption" role="heading"><span class="caption-text">Reference Guide</span></p> |
| <ul> |
| <li class="toctree-l1"><a class="reference internal" href="../../reference/langref/index.html">Language Reference</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../reference/api/python/index.html">Python API</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../reference/api/links.html">Other APIs</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../reference/publications.html">Publications</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../../genindex.html">Index</a></li> |
| </ul> |
| |
| |
| |
| </div> |
| |
| </div> |
| </nav> |
| |
| <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"> |
| |
| <nav class="wy-nav-top" aria-label="top navigation" data-toggle="wy-nav-top"> |
| |
| <div class="togglemenu"> |
| |
| </div> |
| <div class="nav-content"> |
| <!-- tvm --> |
| Table of Contents |
| </div> |
| |
| </nav> |
| |
| |
| <div class="wy-nav-content"> |
| |
| <div class="rst-content"> |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| <div role="navigation" aria-label="breadcrumbs navigation"> |
| |
| <ul class="wy-breadcrumbs"> |
| |
| <li><a href="../../index.html">Docs</a> <span class="br-arrow">></span></li> |
| |
| <li><a href="../index.html">How To Guides</a> <span class="br-arrow">></span></li> |
| |
| <li><a href="../deploy/index.html">Deploy Models and Integrate TVM</a> <span class="br-arrow">></span></li> |
| |
| <li><a href="index.html">Deploy Deep Learning Models</a> <span class="br-arrow">></span></li> |
| |
| <li>Deploy a Framework-prequantized Model with TVM</li> |
| |
| |
| |
| |
| |
| |
| |
| |
| <li class="wy-breadcrumbs-aside"> |
| |
| |
| |
| <a href="https://github.com/apache/tvm/edit/main/docs/how_to/deploy_models/deploy_prequantized.rst" class="fa fa-github"> Edit on GitHub</a> |
| |
| |
| |
| </li> |
| |
| </ul> |
| |
| |
| <hr/> |
| </div> |
| <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article"> |
| <div itemprop="articleBody"> |
| |
| <div class="sphx-glr-download-link-note admonition note"> |
| <p class="admonition-title">Note</p> |
| <p>This tutorial can be used interactively with Google Colab! You can also click |
| <a class="reference internal" href="#sphx-glr-download-how-to-deploy-models-deploy-prequantized-py"><span class="std std-ref">here</span></a> to run the Jupyter notebook locally.</p> |
| <a class="reference external image-reference" href="https://colab.research.google.com/github/apache/tvm-site/blob/asf-site/docs/_downloads/c20f81a94729f461f33b52cc110fd9d6/deploy_prequantized.ipynb"><img alt="https://raw.githubusercontent.com/tlc-pack/web-data/main/images/utilities/colab_button.svg" class="align-center" src="https://raw.githubusercontent.com/tlc-pack/web-data/main/images/utilities/colab_button.svg" width="300px" /></a> |
| </div> |
| <div class="sphx-glr-example-title section" id="deploy-a-framework-prequantized-model-with-tvm"> |
| <span id="sphx-glr-how-to-deploy-models-deploy-prequantized-py"></span><h1>Deploy a Framework-prequantized Model with TVM<a class="headerlink" href="#deploy-a-framework-prequantized-model-with-tvm" title="Permalink to this headline">¶</a></h1> |
| <p><strong>Author</strong>: <a class="reference external" href="https://github.com/masahi">Masahiro Masuda</a></p> |
| <p>This is a tutorial on loading models quantized by deep learning frameworks into TVM. |
| Pre-quantized model import is one of the quantization support we have in TVM. More details on |
| the quantization story in TVM can be found |
| <a class="reference external" href="https://discuss.tvm.apache.org/t/quantization-story/3920">here</a>.</p> |
| <p>Here, we demonstrate how to load and run models quantized by PyTorch, MXNet, and TFLite. |
| Once loaded, we can run compiled, quantized models on any hardware TVM supports.</p> |
| <p>First, necessary imports</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">PIL</span> <span class="kn">import</span> <span class="n">Image</span> |
| |
| <span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span> |
| |
| <span class="kn">import</span> <span class="nn">torch</span> |
| <span class="kn">from</span> <span class="nn">torchvision.models.quantization</span> <span class="kn">import</span> <span class="n">mobilenet</span> <span class="k">as</span> <span class="n">qmobilenet</span> |
| |
| <span class="kn">import</span> <span class="nn">tvm</span> |
| <span class="kn">from</span> <span class="nn">tvm</span> <span class="kn">import</span> <span class="n">relay</span> |
| <span class="kn">from</span> <span class="nn">tvm.contrib.download</span> <span class="kn">import</span> <span class="n">download_testdata</span> |
| </pre></div> |
| </div> |
| <p>Helper functions to run the demo</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">get_transform</span><span class="p">():</span> |
| <span class="kn">import</span> <span class="nn">torchvision.transforms</span> <span class="k">as</span> <span class="nn">transforms</span> |
| |
| <span class="n">normalize</span> <span class="o">=</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Normalize</span><span class="p">(</span><span class="n">mean</span><span class="o">=</span><span class="p">[</span><span class="mf">0.485</span><span class="p">,</span> <span class="mf">0.456</span><span class="p">,</span> <span class="mf">0.406</span><span class="p">],</span> <span class="n">std</span><span class="o">=</span><span class="p">[</span><span class="mf">0.229</span><span class="p">,</span> <span class="mf">0.224</span><span class="p">,</span> <span class="mf">0.225</span><span class="p">])</span> |
| <span class="k">return</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">(</span> |
| <span class="p">[</span> |
| <span class="n">transforms</span><span class="o">.</span><span class="n">Resize</span><span class="p">(</span><span class="mi">256</span><span class="p">),</span> |
| <span class="n">transforms</span><span class="o">.</span><span class="n">CenterCrop</span><span class="p">(</span><span class="mi">224</span><span class="p">),</span> |
| <span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span> |
| <span class="n">normalize</span><span class="p">,</span> |
| <span class="p">]</span> |
| <span class="p">)</span> |
| |
| |
| <span class="k">def</span> <span class="nf">get_real_image</span><span class="p">(</span><span class="n">im_height</span><span class="p">,</span> <span class="n">im_width</span><span class="p">):</span> |
| <span class="n">img_url</span> <span class="o">=</span> <span class="s2">"https://github.com/dmlc/mxnet.js/blob/main/data/cat.png?raw=true"</span> |
| <span class="n">img_path</span> <span class="o">=</span> <span class="n">download_testdata</span><span class="p">(</span><span class="n">img_url</span><span class="p">,</span> <span class="s2">"cat.png"</span><span class="p">,</span> <span class="n">module</span><span class="o">=</span><span class="s2">"data"</span><span class="p">)</span> |
| <span class="k">return</span> <span class="n">Image</span><span class="o">.</span><span class="n">open</span><span class="p">(</span><span class="n">img_path</span><span class="p">)</span><span class="o">.</span><span class="n">resize</span><span class="p">((</span><span class="n">im_height</span><span class="p">,</span> <span class="n">im_width</span><span class="p">))</span> |
| |
| |
| <span class="k">def</span> <span class="nf">get_imagenet_input</span><span class="p">():</span> |
| <span class="n">im</span> <span class="o">=</span> <span class="n">get_real_image</span><span class="p">(</span><span class="mi">224</span><span class="p">,</span> <span class="mi">224</span><span class="p">)</span> |
| <span class="n">preprocess</span> <span class="o">=</span> <span class="n">get_transform</span><span class="p">()</span> |
| <span class="n">pt_tensor</span> <span class="o">=</span> <span class="n">preprocess</span><span class="p">(</span><span class="n">im</span><span class="p">)</span> |
| <span class="k">return</span> <span class="n">np</span><span class="o">.</span><span class="n">expand_dims</span><span class="p">(</span><span class="n">pt_tensor</span><span class="o">.</span><span class="n">numpy</span><span class="p">(),</span> <span class="mi">0</span><span class="p">)</span> |
| |
| |
| <span class="k">def</span> <span class="nf">get_synset</span><span class="p">():</span> |
| <span class="n">synset_url</span> <span class="o">=</span> <span class="s2">""</span><span class="o">.</span><span class="n">join</span><span class="p">(</span> |
| <span class="p">[</span> |
| <span class="s2">"https://gist.githubusercontent.com/zhreshold/"</span><span class="p">,</span> |
| <span class="s2">"4d0b62f3d01426887599d4f7ede23ee5/raw/"</span><span class="p">,</span> |
| <span class="s2">"596b27d23537e5a1b5751d2b0481ef172f58b539/"</span><span class="p">,</span> |
| <span class="s2">"imagenet1000_clsid_to_human.txt"</span><span class="p">,</span> |
| <span class="p">]</span> |
| <span class="p">)</span> |
| <span class="n">synset_name</span> <span class="o">=</span> <span class="s2">"imagenet1000_clsid_to_human.txt"</span> |
| <span class="n">synset_path</span> <span class="o">=</span> <span class="n">download_testdata</span><span class="p">(</span><span class="n">synset_url</span><span class="p">,</span> <span class="n">synset_name</span><span class="p">,</span> <span class="n">module</span><span class="o">=</span><span class="s2">"data"</span><span class="p">)</span> |
| <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">synset_path</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span> |
| <span class="k">return</span> <span class="nb">eval</span><span class="p">(</span><span class="n">f</span><span class="o">.</span><span class="n">read</span><span class="p">())</span> |
| |
| |
| <span class="k">def</span> <span class="nf">run_tvm_model</span><span class="p">(</span><span class="n">mod</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_name</span></a><span class="p">,</span> <span class="n">inp</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a><span class="o">=</span><span class="s2">"llvm"</span><span class="p">):</span> |
| <span class="k">with</span> <a href="../../reference/api/python/ir.html#tvm.transform.PassContext" title="tvm.transform.PassContext" class="sphx-glr-backref-module-tvm-transform sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">tvm</span><span class="o">.</span><span class="n">transform</span><span class="o">.</span><span class="n">PassContext</span></a><span class="p">(</span><span class="n">opt_level</span><span class="o">=</span><span class="mi">3</span><span class="p">):</span> |
| <span class="n">lib</span> <span class="o">=</span> <span class="n">relay</span><span class="o">.</span><span class="n">build</span><span class="p">(</span><span class="n">mod</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a><span class="o">=</span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a><span class="o">=</span><a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a><span class="p">)</span> |
| |
| <span class="n">runtime</span> <span class="o">=</span> <a href="../../reference/api/python/graph_executor.html#tvm.contrib.graph_executor.GraphModule" title="tvm.contrib.graph_executor.GraphModule" class="sphx-glr-backref-module-tvm-contrib-graph_executor sphx-glr-backref-type-py-class"><span class="n">tvm</span><span class="o">.</span><span class="n">contrib</span><span class="o">.</span><span class="n">graph_executor</span><span class="o">.</span><span class="n">GraphModule</span></a><span class="p">(</span><span class="n">lib</span><span class="p">[</span><span class="s2">"default"</span><span class="p">](</span><span class="n">tvm</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a><span class="p">,</span> <span class="mi">0</span><span class="p">)))</span> |
| |
| <span class="n">runtime</span><span class="o">.</span><span class="n">set_input</span><span class="p">(</span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_name</span></a><span class="p">,</span> <span class="n">inp</span><span class="p">)</span> |
| <span class="n">runtime</span><span class="o">.</span><span class="n">run</span><span class="p">()</span> |
| <span class="k">return</span> <span class="n">runtime</span><span class="o">.</span><span class="n">get_output</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">numpy</span><span class="p">(),</span> <span class="n">runtime</span> |
| </pre></div> |
| </div> |
| <p>A mapping from label to class name, to verify that the outputs from models below |
| are reasonable</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">synset</span></a> <span class="o">=</span> <span class="n">get_synset</span><span class="p">()</span> |
| </pre></div> |
| </div> |
| <p>Everyone’s favorite cat image for demonstration</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">inp</span> <span class="o">=</span> <span class="n">get_imagenet_input</span><span class="p">()</span> |
| </pre></div> |
| </div> |
| <div class="section" id="deploy-a-quantized-pytorch-model"> |
| <h2>Deploy a quantized PyTorch Model<a class="headerlink" href="#deploy-a-quantized-pytorch-model" title="Permalink to this headline">¶</a></h2> |
| <p>First, we demonstrate how to load deep learning models quantized by PyTorch, |
| using our PyTorch frontend.</p> |
| <p>Please refer to the PyTorch static quantization tutorial below to learn about |
| their quantization workflow. |
| <a class="reference external" href="https://pytorch.org/tutorials/advanced/static_quantization_tutorial.html">https://pytorch.org/tutorials/advanced/static_quantization_tutorial.html</a></p> |
| <p>We use this function to quantize PyTorch models. |
| In short, this function takes a floating point model and converts it to uint8. |
| The model is per-channel quantized.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">quantize_model</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">inp</span><span class="p">):</span> |
| <span class="n">model</span><span class="o">.</span><span class="n">fuse_model</span><span class="p">()</span> |
| <span class="n">model</span><span class="o">.</span><span class="n">qconfig</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">get_default_qconfig</span><span class="p">(</span><span class="s2">"fbgemm"</span><span class="p">)</span> |
| <span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">prepare</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span> |
| <span class="c1"># Dummy calibration</span> |
| <span class="n">model</span><span class="p">(</span><span class="n">inp</span><span class="p">)</span> |
| <span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">convert</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span> |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="load-quantization-ready-pretrained-mobilenet-v2-model-from-torchvision"> |
| <h2>Load quantization-ready, pretrained Mobilenet v2 model from torchvision<a class="headerlink" href="#load-quantization-ready-pretrained-mobilenet-v2-model-from-torchvision" title="Permalink to this headline">¶</a></h2> |
| <p>We choose mobilenet v2 because this model was trained with quantization aware |
| training. Other models require a full post training calibration.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">qmodel</span> <span class="o">=</span> <span class="n">qmobilenet</span><span class="o">.</span><span class="n">mobilenet_v2</span><span class="p">(</span><span class="n">pretrained</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span> |
| </pre></div> |
| </div> |
| <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/venv/apache-tvm-py3.8/lib/python3.8/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead. |
| warnings.warn( |
| /venv/apache-tvm-py3.8/lib/python3.8/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=MobileNet_V2_Weights.IMAGENET1K_V1`. You can also use `weights=MobileNet_V2_Weights.DEFAULT` to get the most up-to-date weights. |
| warnings.warn(msg) |
| Downloading: "https://download.pytorch.org/models/mobilenet_v2-b0353104.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2-b0353104.pth |
| |
| 0%| | 0.00/13.6M [00:00<?, ?B/s] |
| 47%|####6 | 6.30M/13.6M [00:00<00:00, 31.8MB/s] |
| 69%|######8 | 9.34M/13.6M [00:00<00:00, 21.2MB/s] |
| 90%|########9 | 12.2M/13.6M [00:00<00:00, 22.4MB/s] |
| 100%|##########| 13.6M/13.6M [00:00<00:00, 25.6MB/s] |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="quantize-trace-and-run-the-pytorch-mobilenet-v2-model"> |
| <h2>Quantize, trace and run the PyTorch Mobilenet v2 model<a class="headerlink" href="#quantize-trace-and-run-the-pytorch-mobilenet-v2-model" title="Permalink to this headline">¶</a></h2> |
| <p>The details are out of scope for this tutorial. Please refer to the tutorials |
| on the PyTorch website to learn about quantization and jit.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pt_inp</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">from_numpy</span><span class="p">(</span><span class="n">inp</span><span class="p">)</span> |
| <span class="n">quantize_model</span><span class="p">(</span><span class="n">qmodel</span><span class="p">,</span> <span class="n">pt_inp</span><span class="p">)</span> |
| <span class="n">script_module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">qmodel</span><span class="p">,</span> <span class="n">pt_inp</span><span class="p">)</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span> |
| |
| <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span> |
| <span class="n">pt_result</span> <span class="o">=</span> <span class="n">script_module</span><span class="p">(</span><span class="n">pt_inp</span><span class="p">)</span><span class="o">.</span><span class="n">numpy</span><span class="p">()</span> |
| </pre></div> |
| </div> |
| <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/venv/apache-tvm-py3.8/lib/python3.8/site-packages/torch/ao/quantization/observer.py:214: UserWarning: Please use quant_min and quant_max to specify the range for observers. reduce_range will be deprecated in a future release of PyTorch. |
| warnings.warn( |
| /venv/apache-tvm-py3.8/lib/python3.8/site-packages/torch/ao/quantization/observer.py:1209: UserWarning: must run observer before calling calculate_qparams. Returning default scale and zero point |
| warnings.warn( |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="convert-quantized-mobilenet-v2-to-relay-qnn-using-the-pytorch-frontend"> |
| <h2>Convert quantized Mobilenet v2 to Relay-QNN using the PyTorch frontend<a class="headerlink" href="#convert-quantized-mobilenet-v2-to-relay-qnn-using-the-pytorch-frontend" title="Permalink to this headline">¶</a></h2> |
| <p>The PyTorch frontend has support for converting a quantized PyTorch model to |
| an equivalent Relay module enriched with quantization-aware operators. |
| We call this representation Relay QNN dialect.</p> |
| <p>You can print the output from the frontend to see how quantized models are |
| represented.</p> |
| <p>You would see operators specific to quantization such as |
| qnn.quantize, qnn.dequantize, qnn.requantize, and qnn.conv2d etc.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_name</span></a> <span class="o">=</span> <span class="s2">"input"</span> <span class="c1"># the input name can be arbitrary for PyTorch frontend.</span> |
| <a href="https://docs.python.org/3/library/stdtypes.html#list" title="builtins.list" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_shapes</span></a> <span class="o">=</span> <span class="p">[(</span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_name</span></a><span class="p">,</span> <span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">224</span><span class="p">,</span> <span class="mi">224</span><span class="p">))]</span> |
| <span class="n">mod</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a> <span class="o">=</span> <a href="../../reference/api/python/relay/frontend.html#tvm.relay.frontend.from_pytorch" title="tvm.relay.frontend.from_pytorch" class="sphx-glr-backref-module-tvm-relay-frontend sphx-glr-backref-type-py-function"><span class="n">relay</span><span class="o">.</span><span class="n">frontend</span><span class="o">.</span><span class="n">from_pytorch</span></a><span class="p">(</span><span class="n">script_module</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#list" title="builtins.list" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_shapes</span></a><span class="p">)</span> |
| <span class="c1"># print(mod) # comment in to see the QNN IR dump</span> |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="compile-and-run-the-relay-module"> |
| <h2>Compile and run the Relay module<a class="headerlink" href="#compile-and-run-the-relay-module" title="Permalink to this headline">¶</a></h2> |
| <p>Once we obtained the quantized Relay module, the rest of the workflow |
| is the same as running floating point models. Please refer to other |
| tutorials for more details.</p> |
| <p>Under the hood, quantization specific operators are lowered to a sequence of |
| standard Relay operators before compilation.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a> <span class="o">=</span> <span class="s2">"llvm"</span> |
| <span class="n">tvm_result</span><span class="p">,</span> <a href="../../reference/api/python/graph_executor.html#tvm.contrib.graph_executor.GraphModule" title="tvm.contrib.graph_executor.GraphModule" class="sphx-glr-backref-module-tvm-contrib-graph_executor sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">rt_mod</span></a> <span class="o">=</span> <span class="n">run_tvm_model</span><span class="p">(</span><span class="n">mod</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">input_name</span></a><span class="p">,</span> <span class="n">inp</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a><span class="o">=</span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">target</span></a><span class="p">)</span> |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="compare-the-output-labels"> |
| <h2>Compare the output labels<a class="headerlink" href="#compare-the-output-labels" title="Permalink to this headline">¶</a></h2> |
| <p>We should see identical labels printed.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pt_top3_labels</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">argsort</span><span class="p">(</span><span class="n">pt_result</span><span class="p">[</span><span class="mi">0</span><span class="p">])[::</span><span class="o">-</span><span class="mi">1</span><span class="p">][:</span><span class="mi">3</span><span class="p">]</span> |
| <span class="n">tvm_top3_labels</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">argsort</span><span class="p">(</span><span class="n">tvm_result</span><span class="p">[</span><span class="mi">0</span><span class="p">])[::</span><span class="o">-</span><span class="mi">1</span><span class="p">][:</span><span class="mi">3</span><span class="p">]</span> |
| |
| <span class="nb">print</span><span class="p">(</span><span class="s2">"PyTorch top3 labels:"</span><span class="p">,</span> <span class="p">[</span><a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">synset</span></a><span class="p">[</span><span class="n">label</span><span class="p">]</span> <span class="k">for</span> <span class="n">label</span> <span class="ow">in</span> <span class="n">pt_top3_labels</span><span class="p">])</span> |
| <span class="nb">print</span><span class="p">(</span><span class="s2">"TVM top3 labels:"</span><span class="p">,</span> <span class="p">[</span><a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">synset</span></a><span class="p">[</span><span class="n">label</span><span class="p">]</span> <span class="k">for</span> <span class="n">label</span> <span class="ow">in</span> <span class="n">tvm_top3_labels</span><span class="p">])</span> |
| </pre></div> |
| </div> |
| <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>PyTorch top3 labels: ['tiger cat', 'Egyptian cat', 'lynx, catamount'] |
| TVM top3 labels: ['tiger cat', 'Egyptian cat', 'tabby, tabby cat'] |
| </pre></div> |
| </div> |
| <p>However, due to the difference in numerics, in general the raw floating point |
| outputs are not expected to be identical. Here, we print how many floating point |
| output values are identical out of 1000 outputs from mobilenet v2.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="s2">"</span><span class="si">%d</span><span class="s2"> in 1000 raw floating outputs identical."</span> <span class="o">%</span> <span class="n">np</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">tvm_result</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">==</span> <span class="n">pt_result</span><span class="p">[</span><span class="mi">0</span><span class="p">]))</span> |
| </pre></div> |
| </div> |
| <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>122 in 1000 raw floating outputs identical. |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="measure-performance"> |
| <h2>Measure performance<a class="headerlink" href="#measure-performance" title="Permalink to this headline">¶</a></h2> |
| <p>Here we give an example of how to measure performance of TVM compiled models.</p> |
| <div class="highlight-default notranslate"><div class="highlight"><pre><span></span><a href="https://docs.python.org/3/library/functions.html#int" title="builtins.int" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">n_repeat</span></a> <span class="o">=</span> <span class="mi">100</span> <span class="c1"># should be bigger to make the measurement more accurate</span> |
| <span class="n">dev</span> <span class="o">=</span> <span class="n">tvm</span><span class="o">.</span><span class="n">cpu</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> |
| <span class="nb">print</span><span class="p">(</span><a href="../../reference/api/python/graph_executor.html#tvm.contrib.graph_executor.GraphModule.benchmark" title="tvm.contrib.graph_executor.GraphModule.benchmark" class="sphx-glr-backref-module-tvm-contrib-graph_executor sphx-glr-backref-type-py-method"><span class="n">rt_mod</span><span class="o">.</span><span class="n">benchmark</span></a><span class="p">(</span><span class="n">dev</span><span class="p">,</span> <span class="n">number</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">repeat</span><span class="o">=</span><a href="https://docs.python.org/3/library/functions.html#int" title="builtins.int" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">n_repeat</span></a><span class="p">))</span> |
| </pre></div> |
| </div> |
| <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time summary: |
| mean (ms) median (ms) max (ms) min (ms) std (ms) |
| 87.9379 87.9031 91.7352 87.4494 0.4585 |
| </pre></div> |
| </div> |
| <div class="admonition note"> |
| <p class="admonition-title">Note</p> |
| <p>We recommend this method for the following reasons:</p> |
| <blockquote> |
| <div><ul class="simple"> |
| <li><p>Measurements are done in C++, so there is no Python overhead</p></li> |
| <li><p>It includes several warm up runs</p></li> |
| <li><p>The same method can be used to profile on remote devices (android etc.).</p></li> |
| </ul> |
| </div></blockquote> |
| </div> |
| <div class="admonition note"> |
| <p class="admonition-title">Note</p> |
| <p>Unless the hardware has special support for fast 8 bit instructions, quantized models are |
| not expected to be any faster than FP32 models. Without fast 8 bit instructions, TVM does |
| quantized convolution in 16 bit, even if the model itself is 8 bit.</p> |
| <p>For x86, the best performance can be achieved on CPUs with AVX512 instructions set. |
| In this case, TVM utilizes the fastest available 8 bit instructions for the given target. |
| This includes support for the VNNI 8 bit dot product instruction (CascadeLake or newer).</p> |
| <p>Moreover, the following general tips for CPU performance equally applies:</p> |
| <blockquote> |
| <div><ul class="simple"> |
| <li><p>Set the environment variable TVM_NUM_THREADS to the number of physical cores</p></li> |
| <li><p>Choose the best target for your hardware, such as “llvm -mcpu=skylake-avx512” or |
| “llvm -mcpu=cascadelake” (more CPUs with AVX512 would come in the future)</p></li> |
| </ul> |
| </div></blockquote> |
| </div> |
| </div> |
| <div class="section" id="deploy-a-quantized-mxnet-model"> |
| <h2>Deploy a quantized MXNet Model<a class="headerlink" href="#deploy-a-quantized-mxnet-model" title="Permalink to this headline">¶</a></h2> |
| <p>TODO</p> |
| </div> |
| <div class="section" id="deploy-a-quantized-tflite-model"> |
| <h2>Deploy a quantized TFLite Model<a class="headerlink" href="#deploy-a-quantized-tflite-model" title="Permalink to this headline">¶</a></h2> |
| <p>TODO</p> |
| <p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 35.965 seconds)</p> |
| <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-py"> |
| <div class="sphx-glr-download sphx-glr-download-python docutils container"> |
| <p><a class="reference download internal" download="" href="../../_downloads/fb8217c13f4351224c6cf3aacf1a87fc/deploy_prequantized.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">deploy_prequantized.py</span></code></a></p> |
| </div> |
| <div class="sphx-glr-download sphx-glr-download-jupyter docutils container"> |
| <p><a class="reference download internal" download="" href="../../_downloads/c20f81a94729f461f33b52cc110fd9d6/deploy_prequantized.ipynb"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Jupyter</span> <span class="pre">notebook:</span> <span class="pre">deploy_prequantized.ipynb</span></code></a></p> |
| </div> |
| </div> |
| <p class="sphx-glr-signature"><a class="reference external" href="https://sphinx-gallery.github.io">Gallery generated by Sphinx-Gallery</a></p> |
| </div> |
| </div> |
| |
| |
| </div> |
| |
| </div> |
| |
| |
| <footer> |
| |
| <div class="rst-footer-buttons" role="navigation" aria-label="footer navigation"> |
| |
| <a href="deploy_prequantized_tflite.html" class="btn btn-neutral float-right" title="Deploy a Framework-prequantized Model with TVM - Part 3 (TFLite)" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right"></span></a> |
| |
| |
| <a href="deploy_object_detection_pytorch.html" class="btn btn-neutral float-left" title="Compile PyTorch Object Detection Models" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left"></span> Previous</a> |
| |
| </div> |
| |
| <div id="button" class="backtop"><img src="../../_static/img/right.svg" alt="backtop"/> </div> |
| <section class="footerSec"> |
| <div class="footerHeader"> |
| <div class="d-flex align-md-items-center justify-content-between flex-column flex-md-row"> |
| <div class="copywrite d-flex align-items-center"> |
| <h5 id="copy-right-info">© 2023 Apache Software Foundation | All rights reserved</h5> |
| </div> |
| </div> |
| |
| </div> |
| |
| <div> |
| <div class="footernote">Copyright © 2023 The Apache Software Foundation. Apache TVM, Apache, the Apache feather, and the Apache TVM project logo are either trademarks or registered trademarks of the Apache Software Foundation.</div> |
| </div> |
| |
| </section> |
| </footer> |
| </div> |
| </div> |
| |
| </section> |
| |
| </div> |
| |
| |
| <script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.12.9/umd/popper.min.js" integrity="sha384-ApNbgh9B+Y1QKtv3Rn7W3mgPxhU9K/ScQsAP7hUibX39j7fakFPskvXusvfa0b4Q" crossorigin="anonymous"></script> |
| <script src="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/js/bootstrap.min.js" integrity="sha384-JZR6Spejh4U02d8jOt6vLEHfe/JQGiRRSQQxSfFWpi1MquVdAyjUar5+76PVCmYl" crossorigin="anonymous"></script> |
| |
| </body> |
| <script type="text/javascript"> |
| jQuery(function () { |
| SphinxRtdTheme.Navigation.enable(true); |
| }); |
| </script> |
| |
| |
| |
| |
| <!-- Theme Analytics --> |
| <script> |
| (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ |
| (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), |
| m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) |
| })(window,document,'script','https://www.google-analytics.com/analytics.js','ga'); |
| |
| ga('create', 'UA-75982049-2', 'auto'); |
| ga('send', 'pageview'); |
| </script> |
| |
| |
| |
| |
| </body> |
| </html> |
