Spaces:

Do0rMaMu
/

Factory-POC

Sleeping

App Files Files Community

Factory-POC / flash-attention /csrc /cutlass /python /docs /cutlass.op.html

Do0rMaMu

Upload folder using huggingface_hub

e45d058 verified 8 months ago

raw

history blame contribute delete

75.9 kB

	<!doctype html>
	<html class="no-js" lang="en">
	<head><meta charset="utf-8"/>
	<meta name="viewport" content="width=device-width,initial-scale=1"/>
	<meta name="color-scheme" content="light dark"><meta name="generator" content="Docutils 0.19: https://docutils.sourceforge.io/" />
	<link rel="index" title="Index" href="genindex.html" /><link rel="search" title="Search" href="search.html" /><link rel="next" title="Utilities" href="cutlass.utils.html" /><link rel="prev" title="Emitters" href="cutlass.emit.html" />
	<link rel="canonical" href="docs/cutlass.op.html" />

	<!-- Generated with Sphinx 6.1.3 and Furo 2023.03.27 -->
	<title>Operations - CUTLASS Python</title>
	<link rel="stylesheet" type="text/css" href="_static/pygments.css" />
	<link rel="stylesheet" type="text/css" href="_static/styles/furo.css?digest=fad236701ea90a88636c2a8c73b44ae642ed2a53" />
	<link rel="stylesheet" type="text/css" href="_static/copybutton.css" />
	<link rel="stylesheet" type="text/css" href="_static/tabs.css" />
	<link rel="stylesheet" type="text/css" href="_static/styles/furo-extensions.css?digest=30d1aed668e5c3a91c3e3bf6a60b675221979f0e" />




	<style>
	body {
	--color-code-background: #eeffcc;
	--color-code-foreground: black;
	--color-brand-primary: #76B900;
	--color-brand-content: #76B900;

	}
	@media not print {
	body[data-theme="dark"] {
	--color-code-background: #272822;
	--color-code-foreground: #f8f8f2;
	--color-brand-primary: #76B900;
	--color-brand-content: #76B900;

	}
	@media (prefers-color-scheme: dark) {
	body:not([data-theme="light"]) {
	--color-code-background: #272822;
	--color-code-foreground: #f8f8f2;
	--color-brand-primary: #76B900;
	--color-brand-content: #76B900;

	}
	}
	}
	</style></head>
	<body>

	<script>
	document.body.dataset.theme = localStorage.getItem("theme") \|\| "auto";
	</script>


	<svg xmlns="http://www.w3.org/2000/svg" style="display: none;">
	<symbol id="svg-toc" viewBox="0 0 24 24">
	<title>Contents</title>
	<svg stroke="currentColor" fill="currentColor" stroke-width="0" viewBox="0 0 1024 1024">
	<path d="M408 442h480c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8H408c-4.4 0-8 3.6-8 8v56c0 4.4 3.6 8 8 8zm-8 204c0 4.4 3.6 8 8 8h480c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8H408c-4.4 0-8 3.6-8 8v56zm504-486H120c-4.4 0-8 3.6-8 8v56c0 4.4 3.6 8 8 8h784c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8zm0 632H120c-4.4 0-8 3.6-8 8v56c0 4.4 3.6 8 8 8h784c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8zM115.4 518.9L271.7 642c5.8 4.6 14.4.5 14.4-6.9V388.9c0-7.4-8.5-11.5-14.4-6.9L115.4 505.1a8.74 8.74 0 0 0 0 13.8z"/>
	</svg>
	</symbol>
	<symbol id="svg-menu" viewBox="0 0 24 24">
	<title>Menu</title>
	<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
	stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather-menu">
	<line x1="3" y1="12" x2="21" y2="12"></line>
	<line x1="3" y1="6" x2="21" y2="6"></line>
	<line x1="3" y1="18" x2="21" y2="18"></line>
	</svg>
	</symbol>
	<symbol id="svg-arrow-right" viewBox="0 0 24 24">
	<title>Expand</title>
	<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
	stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather-chevron-right">
	<polyline points="9 18 15 12 9 6"></polyline>
	</svg>
	</symbol>
	<symbol id="svg-sun" viewBox="0 0 24 24">
	<title>Light mode</title>
	<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
	stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="feather-sun">
	<circle cx="12" cy="12" r="5"></circle>
	<line x1="12" y1="1" x2="12" y2="3"></line>
	<line x1="12" y1="21" x2="12" y2="23"></line>
	<line x1="4.22" y1="4.22" x2="5.64" y2="5.64"></line>
	<line x1="18.36" y1="18.36" x2="19.78" y2="19.78"></line>
	<line x1="1" y1="12" x2="3" y2="12"></line>
	<line x1="21" y1="12" x2="23" y2="12"></line>
	<line x1="4.22" y1="19.78" x2="5.64" y2="18.36"></line>
	<line x1="18.36" y1="5.64" x2="19.78" y2="4.22"></line>
	</svg>
	</symbol>
	<symbol id="svg-moon" viewBox="0 0 24 24">
	<title>Dark mode</title>
	<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
	stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-moon">
	<path stroke="none" d="M0 0h24v24H0z" fill="none" />
	<path d="M12 3c.132 0 .263 0 .393 0a7.5 7.5 0 0 0 7.92 12.446a9 9 0 1 1 -8.313 -12.454z" />
	</svg>
	</symbol>
	<symbol id="svg-sun-half" viewBox="0 0 24 24">
	<title>Auto light/dark mode</title>
	<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
	stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-shadow">
	<path stroke="none" d="M0 0h24v24H0z" fill="none"/>
	<circle cx="12" cy="12" r="9" />
	<path d="M13 12h5" />
	<path d="M13 15h4" />
	<path d="M13 18h1" />
	<path d="M13 9h4" />
	<path d="M13 6h1" />
	</svg>
	</symbol>
	</svg>

	<input type="checkbox" class="sidebar-toggle" name="__navigation" id="__navigation">
	<input type="checkbox" class="sidebar-toggle" name="__toc" id="__toc">
	<label class="overlay sidebar-overlay" for="__navigation">
	<div class="visually-hidden">Hide navigation sidebar</div>
	</label>
	<label class="overlay toc-overlay" for="__toc">
	<div class="visually-hidden">Hide table of contents sidebar</div>
	</label>



	<div class="page">
	<header class="mobile-header">
	<div class="header-left">
	<label class="nav-overlay-icon" for="__navigation">
	<div class="visually-hidden">Toggle site navigation sidebar</div>
	<i class="icon"><svg><use href="#svg-menu"></use></svg></i>
	</label>
	</div>
	<div class="header-center">
	<a href="index.html"><div class="brand">CUTLASS Python</div></a>
	</div>
	<div class="header-right">
	<div class="theme-toggle-container theme-toggle-header">
	<button class="theme-toggle">
	<div class="visually-hidden">Toggle Light / Dark / Auto color theme</div>
	<svg class="theme-icon-when-auto"><use href="#svg-sun-half"></use></svg>
	<svg class="theme-icon-when-dark"><use href="#svg-moon"></use></svg>
	<svg class="theme-icon-when-light"><use href="#svg-sun"></use></svg>
	</button>
	</div>
	<label class="toc-overlay-icon toc-header-icon" for="__toc">
	<div class="visually-hidden">Toggle table of contents sidebar</div>
	<i class="icon"><svg><use href="#svg-toc"></use></svg></i>
	</label>
	</div>
	</header>
	<aside class="sidebar-drawer">
	<div class="sidebar-container">

	<div class="sidebar-sticky"><a class="sidebar-brand" href="index.html">

	<div class="sidebar-logo-container">
	<img class="sidebar-logo only-light" src="_static/cutlass-logo-small.png" alt="Light Logo"/>
	<img class="sidebar-logo only-dark" src="_static/cutlass-logo-small.png" alt="Dark Logo"/>
	</div>

	<span class="sidebar-brand-text">CUTLASS Python</span>

	</a><form class="sidebar-search-container" method="get" action="search.html" role="search">
	<input class="sidebar-search" placeholder="Search" name="q" aria-label="Search">
	<input type="hidden" name="check_keywords" value="yes">
	<input type="hidden" name="area" value="default">
	</form>
	<div id="searchbox"></div><div class="sidebar-scroll"><div class="sidebar-tree">
	<ul>
	<li class="toctree-l1"><a class="reference internal" href="index.html">Home</a></li>
	</ul>
	<p class="caption" role="heading"><span class="caption-text">Getting Started:</span></p>
	<ul>
	<li class="toctree-l1"><a class="reference internal" href="install.html">Installation</a></li>
	<li class="toctree-l1"><a class="reference internal" href="externals/00_basic_gemm.html">Getting Started</a></li>
	<li class="toctree-l1"><a class="reference internal" href="contribute.html">Contributing</a></li>
	</ul>
	<p class="caption" role="heading"><span class="caption-text">Python Documentation:</span></p>
	<ul class="current">
	<li class="toctree-l1 current has-children"><a class="reference internal" href="modules.html">CUTLASS Python API</a><input checked="" class="toctree-checkbox" id="toctree-checkbox-1" name="toctree-checkbox-1" role="switch" type="checkbox"/><label for="toctree-checkbox-1"><div class="visually-hidden">Toggle child pages in navigation</div><i class="icon"><svg><use href="#svg-arrow-right"></use></svg></i></label><ul class="current">
	<li class="toctree-l2 current has-children"><a class="reference internal" href="cutlass.html">CUTLASS</a><input checked="" class="toctree-checkbox" id="toctree-checkbox-2" name="toctree-checkbox-2" role="switch" type="checkbox"/><label for="toctree-checkbox-2"><div class="visually-hidden">Toggle child pages in navigation</div><i class="icon"><svg><use href="#svg-arrow-right"></use></svg></i></label><ul class="current">
	<li class="toctree-l3"><a class="reference internal" href="cutlass.emit.html">Emitters</a></li>
	<li class="toctree-l3 current current-page"><a class="current reference internal" href="#">Operations</a></li>
	<li class="toctree-l3"><a class="reference internal" href="cutlass.utils.html">Utilities</a></li>
	</ul>
	</li>
	</ul>
	</li>
	</ul>
	<p class="caption" role="heading"><span class="caption-text">Examples and Tutorials:</span></p>
	<ul>
	<li class="toctree-l1 has-children"><a class="reference internal" href="examples.html">Examples</a><input class="toctree-checkbox" id="toctree-checkbox-3" name="toctree-checkbox-3" role="switch" type="checkbox"/><label for="toctree-checkbox-3"><div class="visually-hidden">Toggle child pages in navigation</div><i class="icon"><svg><use href="#svg-arrow-right"></use></svg></i></label><ul>
	<li class="toctree-l2"><a class="reference internal" href="externals/00_basic_gemm.html">Basic GEMM</a></li>
	<li class="toctree-l2"><a class="reference internal" href="externals/01_epilogue.html">Epilogue</a></li>
	<li class="toctree-l2"><a class="reference internal" href="externals/02_pytorch_extension_grouped_gemm.html">PyTorch Extension</a></li>
	</ul>
	</li>
	</ul>
	<p class="caption" role="heading"><span class="caption-text">Reference:</span></p>
	<ul>
	<li class="toctree-l1"><a class="reference external" href="https://github.com/NVIDIA/cutlass">Github</a></li>
	</ul>

	</div>
	</div>

	</div>

	</div>
	</aside>
	<div class="main">
	<div class="content">
	<div class="article-container">
	<a href="#" class="back-to-top muted-link">
	<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24">
	<path d="M13 20h-2V8l-5.5 5.5-1.42-1.42L12 4.16l7.92 7.92-1.42 1.42L13 8v12z"></path>
	</svg>
	<span>Back to top</span>
	</a>
	<div class="content-icon-container">

	<div class="theme-toggle-container theme-toggle-content">
	<button class="theme-toggle">
	<div class="visually-hidden">Toggle Light / Dark / Auto color theme</div>
	<svg class="theme-icon-when-auto"><use href="#svg-sun-half"></use></svg>
	<svg class="theme-icon-when-dark"><use href="#svg-moon"></use></svg>
	<svg class="theme-icon-when-light"><use href="#svg-sun"></use></svg>
	</button>
	</div>
	<label class="toc-overlay-icon toc-content-icon" for="__toc">
	<div class="visually-hidden">Toggle table of contents sidebar</div>
	<i class="icon"><svg><use href="#svg-toc"></use></svg></i>
	</label>
	</div>
	<article role="main">
	<section id="operations">
	<h1>Operations<a class="headerlink" href="#operations" title="Permalink to this heading">#</a></h1>
	<section id="module-cutlass.op.gemm">
	<span id="gemm"></span><h2>GEMM<a class="headerlink" href="#module-cutlass.op.gemm" title="Permalink to this heading">#</a></h2>
	<p>Ease-of-use interface for constructing, compiling, and running GEMMs.</p>
	<p>The <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> interface is meant to allow one to easily instantiate, compile, and run
	GEMM operations in CUTLASS via Python, without specifying many configuration parameters.
	Under the hood, the interface will select sensible default parameters for the many template
	parameters for CUTLASS GEMMs.</p>
	<p>Note: optimal performance is not to be expected from this interface. To achieve optimal
	performance, one should specify and tune each configuration parameter.</p>
	<p>The simplest example of using this interface is the following:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="c1"># A, B, C, and D are torch/numpy/cupy tensor objects</span>
	<span class="n">plan</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">op</span><span class="o">.</span><span class="n">Gemm</span><span class="p">(</span><span class="n">A</span><span class="p">,</span> <span class="n">B</span><span class="p">,</span> <span class="n">C</span><span class="p">,</span> <span class="n">D</span><span class="p">)</span>
	<span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>
	</pre></div>
	</div>
	<p>One can also use the interface by specifying data types of operands at construction
	and using different tensor objects with these data types at runtime:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="c1"># The following is shorthand for:</span>
	<span class="c1"># cutlass.op.Gemm(element_A=torch.float32, element_B=torch.float32,</span>
	<span class="c1"># element_C=torch.float32, element_D=torch.float32,</span>
	<span class="c1"># element_accumulator=torch.float32,</span>
	<span class="c1"># layout=cutlass.LayoutType.RowMajor)</span>
	<span class="n">plan</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">op</span><span class="o">.</span><span class="n">Gemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>

	<span class="n">A0</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">((</span><span class="mi">128</span><span class="p">,</span> <span class="mi">256</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">B0</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">((</span><span class="mi">256</span><span class="p">,</span> <span class="mi">64</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">C0</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="mi">128</span><span class="p">,</span> <span class="mi">64</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">D0</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="mi">128</span><span class="p">,</span> <span class="mi">64</span><span class="p">),</span> <span class="n">device</span><span class="o">.</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">(</span><span class="n">A0</span><span class="p">,</span> <span class="n">B0</span><span class="p">,</span> <span class="n">C0</span><span class="p">,</span> <span class="n">D0</span><span class="p">)</span>

	<span class="n">A</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">((</span><span class="mi">32</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">B</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">((</span><span class="mi">128</span><span class="p">,</span> <span class="mi">256</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">C</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="mi">32</span><span class="p">,</span> <span class="mi">256</span><span class="p">),</span> <span class="n">device</span><span class="o">=</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">D</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">((</span><span class="mi">32</span><span class="p">,</span> <span class="mi">256</span><span class="p">),</span> <span class="n">device</span><span class="o">.</span><span class="s1">'cuda'</span><span class="p">)</span>
	<span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">(</span><span class="n">A1</span><span class="p">,</span> <span class="n">B1</span><span class="p">,</span> <span class="n">C1</span><span class="p">,</span> <span class="n">D1</span><span class="p">)</span>
	</pre></div>
	</div>
	<p>The interface additionally enables one to decouple the compilation of the underlying CUTLASS
	kernel from its execution:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">plan</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">op</span><span class="o">.</span><span class="n">Gemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>
	<span class="n">plan</span><span class="o">.</span><span class="n">compile</span><span class="p">()</span>

	<span class="c1"># Do other work...</span>

	<span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">(</span><span class="n">A0</span><span class="p">,</span> <span class="n">B0</span><span class="p">,</span> <span class="n">C0</span><span class="p">,</span> <span class="n">D0</span><span class="p">)</span>

	<span class="c1"># Do other work...</span>

	<span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">(</span><span class="n">A1</span><span class="p">,</span> <span class="n">B1</span><span class="p">,</span> <span class="n">C1</span><span class="p">,</span> <span class="n">D1</span><span class="p">)</span>
	</pre></div>
	</div>
	<p>Elementwise activation functions are easily fused to the GEMM via the interface:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">plan</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">op</span><span class="o">.</span><span class="n">Gemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>
	<span class="n">plan</span><span class="o">.</span><span class="n">activation</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">epilogue</span><span class="o">.</span><span class="n">relu</span>
	</pre></div>
	</div>
	<p>Operations can also be run asynchronously:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">plan</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">op</span><span class="o">.</span><span class="n">Gemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>
	<span class="n">args</span> <span class="o">=</span> <span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">()</span>

	<span class="c1"># Do other work...</span>

	<span class="n">args</span><span class="o">.</span><span class="n">sync</span><span class="p">()</span>
	</pre></div>
	</div>
	<dl class="py class">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm">
	<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">cutlass.op.gemm.</span></span><span class="sig-name descname"><span class="pre">Gemm</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">D</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">beta</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_accumulator</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_D</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cc</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">kernel_cc</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm.html#Gemm"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm.Gemm" title="Permalink to this definition">#</a></dt>
	<dd><p>Bases: <a class="reference internal" href="#cutlass.op.op.OperationBase" title="cutlass.op.op.OperationBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">OperationBase</span></code></a></p>
	<p>Constructs a <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> object.</p>
	<p>The data types and layouts of operands A, B, and C, along with the data type of output D
	and that used for accumulation, are bound to the <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> object throughout its lifetime –
	these are not to be changed after a <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> has been constructed.</p>
	<p>The constructor has optional parameters for flexibly setting these parameters. The following
	constructors are equivalent:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="c1"># Use F32 for A, B, C, D, and accumulation. All operands are row major.</span>

	<span class="c1"># Use the generic ``element`` and ``layout`` parameters to concisely set all data types and layouts</span>
	<span class="c1"># for operands to the same values.</span>
	<span class="n">Gemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>

	<span class="c1"># Explicitly specify the data types to use for A, B, C, and D. Use the generic ``layout``.</span>
	<span class="n">Gemm</span><span class="p">(</span><span class="n">element_A</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">element_B</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">element_C</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span>
	<span class="n">element_D</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>

	<span class="c1"># Set the data types and elements from existing tensors. Note that one can use different tensors when</span>
	<span class="c1"># executing GEMM via the ``run()`` method than passed in here (though those passed in to ``run()`` must</span>
	<span class="c1"># have the same data type and layout as those passed in here).</span>
	<span class="c1"># A, B, C, and D are row-major torch.Tensor objects of type torch.float32</span>
	<span class="n">Gemm</span><span class="p">(</span><span class="n">A</span><span class="o">=</span><span class="n">A</span><span class="p">,</span> <span class="n">B</span><span class="o">=</span><span class="n">B</span><span class="p">,</span> <span class="n">C</span><span class="o">=</span><span class="n">C</span><span class="p">,</span> <span class="n">D</span><span class="o">=</span><span class="n">D</span><span class="p">)</span>

	<span class="c1"># Use the generic ``element`` and explicitly specify the layouts to use for A, B, and C (layout of D is</span>
	<span class="c1"># the same as that for D, at present)</span>
	<span class="n">Gemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">layout_A</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">,</span>
	<span class="n">layout_B</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">,</span> <span class="n">layout_C</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>

	<span class="c1"># Explicitly specify the data type and layout for only some of A, B, C, and D. Unspecified data types</span>
	<span class="c1"># and layouts will inherit those passed in via the generic ``element`` and ``layout``</span>
	<span class="n">Gemm</span><span class="p">(</span><span class="n">element_A</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">layout_B</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">,</span>
	<span class="n">element</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f32</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>
	</pre></div>
	</div>
	<dl class="simple">
	<dt>The order of precedence for the setting of the data type and layout for a given operand/output is as follows:</dt><dd><ol class="arabic simple">
	<li><p>If the tensor type is specified (e.g., <code class="docutils literal notranslate"><span class="pre">A</span></code>), use the data type and layout inferred from this tensor</p></li>
	<li><p>Otherwise, if the data type/layout (e.g., <code class="docutils literal notranslate"><span class="pre">element_A</span></code>, <code class="docutils literal notranslate"><span class="pre">layout_A</span></code>) is specified, use those</p></li>
	<li><p>Otherwise, use the generic values (e.g., <code class="docutils literal notranslate"><span class="pre">element</span></code>, <code class="docutils literal notranslate"><span class="pre">layout</span></code>)</p></li>
	</ol>
	</dd>
	</dl>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>cc</strong> (<em>int</em>) – compute capability of device for which kernels should be compiled. For example, if running on H100, this should be set to 90</p></li>
	<li><p><strong>kernel_cc</strong> (<em>int</em>) – compute capability of kernels to generate. For example, if running on SM90, but desiring to use a CUTLASS 2.x-style Ampere kernel, this should be set to 80</p></li>
	<li><p><strong>A</strong> – tensor representing data type and layout of operand A</p></li>
	<li><p><strong>B</strong> – tensor representing data type and layout of operand B</p></li>
	<li><p><strong>C</strong> – tensor representing data type and layout of operand C</p></li>
	<li><p><strong>D</strong> – tensor representing data type and layout of operand D</p></li>
	<li><p><strong>alpha</strong> – scalar paramter alpha from GEMM computation that scales the product of operands A and B</p></li>
	<li><p><strong>beta</strong> – scalar parameter beta from GEMM operation that scales operand C</p></li>
	<li><p><strong>element_accumulator</strong> (<em>cutlass.DataType</em>) – data type to be used in accumulation of the product of operands A and B</p></li>
	<li><p><strong>element</strong> (<em>cutlass.DataType</em>) – generic data type to be used for operands A, B, C, D, as well as the accumulation data type</p></li>
	<li><p><strong>layout</strong> (<em>cutlass.LayoutType</em>) – generic layout type to be used for operands A, B, C, and D</p></li>
	<li><p><strong>element_A</strong> (<em>cutlass.DataType</em>) – data type to be used for operand A</p></li>
	<li><p><strong>element_B</strong> (<em>cutlass.DataType</em>) – data type to be used for operand B</p></li>
	<li><p><strong>element_C</strong> (<em>cutlass.DataType</em>) – data type to be used for operand C</p></li>
	<li><p><strong>element_D</strong> (<em>cutlass.DataType</em>) – data type to be used for operand D</p></li>
	<li><p><strong>layout_A</strong> (<em>layout</em><em> of </em><em>operand A</em>) – cutlass.LayoutType</p></li>
	<li><p><strong>layout_B</strong> (<em>layout</em><em> of </em><em>operand B</em>) – cutlass.LayoutType</p></li>
	<li><p><strong>layout_C</strong> (<em>layout</em><em> of </em><em>operand C</em>) – cutlass.LayoutType</p></li>
	<li><p><strong>layout_D</strong> (<em>layout</em><em> of </em><em>operand D</em>) – cutlass.LayoutType</p></li>
	</ul>
	</dd>
	</dl>
	<dl class="py property">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.activation">
	<em class="property"><span class="pre">property</span><span class="w"> </span></em><span class="sig-name descname"><span class="pre">activation</span></span><a class="headerlink" href="#cutlass.op.gemm.Gemm.activation" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns the type of the current activation function used</p>
	</dd></dl>

	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.compile">
	<span class="sig-name descname"><span class="pre">compile</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tile_description</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">print_module</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm.html#Gemm.compile"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm.Gemm.compile" title="Permalink to this definition">#</a></dt>
	<dd><p>Emits and compiles the kernel currently specified. If <code class="docutils literal notranslate"><span class="pre">tile_description</span></code> and any
	of the <code class="docutils literal notranslate"><span class="pre">alignment</span></code> parameters are set, the kernel will be chosen using this
	tile description and alignments. Otherwise, a default tile description and alignment
	will be used.</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>tile_description</strong> (<em>cutlass.backend.TileDescription</em>) – tile description specifying shapes and operand types to use in the kernel</p></li>
	<li><p><strong>alignment_A</strong> (<em>int</em>) – alignment of operand A</p></li>
	<li><p><strong>alignment_B</strong> (<em>int</em>) – alignment of operand B</p></li>
	<li><p><strong>alignment_C</strong> (<em>int</em>) – alignment of operand C</p></li>
	<li><p><strong>print_module</strong> (<em>bool</em>) – whether to print the emitted C++ code</p></li>
	</ul>
	</dd>
	<dt class="field-even">Returns<span class="colon">:</span></dt>
	<dd class="field-even"><p>operation that was compiled</p>
	</dd>
	<dt class="field-odd">Return type<span class="colon">:</span></dt>
	<dd class="field-odd"><p>cutlass.backend.GemmOperationUniversal</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.construct">
	<span class="sig-name descname"><span class="pre">construct</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tile_description</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm.html#Gemm.construct"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm.Gemm.construct" title="Permalink to this definition">#</a></dt>
	<dd><p>Constructs a <code class="docutils literal notranslate"><span class="pre">cutlass.backend.GemmUniversalOperation</span></code> based on the input parameters and current
	kernel specification of the <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> object.</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>tile_description</strong> (<em>cutlass.backend.TileDescription</em>) – tile description specifying shapes and operand types to use in the kernel</p></li>
	<li><p><strong>alignment_A</strong> (<em>int</em>) – alignment of operand A</p></li>
	<li><p><strong>alignment_B</strong> (<em>int</em>) – alignment of operand B</p></li>
	<li><p><strong>alignment_C</strong> (<em>int</em>) – alignment of operand C</p></li>
	</ul>
	</dd>
	<dt class="field-even">Returns<span class="colon">:</span></dt>
	<dd class="field-even"><p>operation that was constructed</p>
	</dd>
	<dt class="field-odd">Return type<span class="colon">:</span></dt>
	<dd class="field-odd"><p>cutlass.backend.GemmOperationUniversal</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py property">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.opclass">
	<em class="property"><span class="pre">property</span><span class="w"> </span></em><span class="sig-name descname"><span class="pre">opclass</span></span><em class="property"><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="pre">OpcodeClass</span></em><a class="headerlink" href="#cutlass.op.gemm.Gemm.opclass" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns the opcode class currently in use by the GEMM</p>
	<dl class="field-list simple">
	<dt class="field-odd">Returns<span class="colon">:</span></dt>
	<dd class="field-odd"><p>opcode class currently in use</p>
	</dd>
	<dt class="field-even">Return type<span class="colon">:</span></dt>
	<dd class="field-even"><p>cutlass.OpcodeClass</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.run">
	<span class="sig-name descname"><span class="pre">run</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">D</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">beta</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_count</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">sync</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">print_module</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm.html#Gemm.run"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm.Gemm.run" title="Permalink to this definition">#</a></dt>
	<dd><p>Runs the kernel currently specified. If it has not already been, the kernel is emitted and
	compiled. Tensors holding operands and outputs of the kernel are sourced either from the
	<code class="docutils literal notranslate"><span class="pre">A</span></code>, <code class="docutils literal notranslate"><span class="pre">B</span></code>, <code class="docutils literal notranslate"><span class="pre">C</span></code>, <code class="docutils literal notranslate"><span class="pre">D</span></code>, <code class="docutils literal notranslate"><span class="pre">alpha</span></code>, and <code class="docutils literal notranslate"><span class="pre">beta</span></code>
	parameters provided in this call, or from those
	passed in on the construction of this object – one of the two must be specified.</p>
	<p>By default, this call returns only once the kernel has completed. To launch the kernel
	and immediately return, set <code class="docutils literal notranslate"><span class="pre">sync=False</span></code>. In this case, it is the responsibility of the
	caller to syncrhonize the results of the kernel before attempting to access outputs
	by calling <code class="docutils literal notranslate"><span class="pre">sync()</span></code> on the arguments returned from this call.</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>A</strong> – tensor representing data type and layout of operand A</p></li>
	<li><p><strong>B</strong> – tensor representing data type and layout of operand B</p></li>
	<li><p><strong>C</strong> – tensor representing data type and layout of operand C</p></li>
	<li><p><strong>D</strong> – tensor representing data type and layout of operand D</p></li>
	<li><p><strong>alpha</strong> – scalar paramter alpha from GEMM computation that scales the product of operands A and B</p></li>
	<li><p><strong>beta</strong> – scalar parameter beta from GEMM operation that scales operand C</p></li>
	<li><p><strong>batch_count</strong> (<em>int</em>) – number of GEMMs in the batch</p></li>
	<li><p><strong>sync</strong> (<em>bool</em>) – whether the call should wait for the kernel to complete before returning</p></li>
	<li><p><strong>print_module</strong> (<em>bool</em>) – whether to print the emitted C++ code</p></li>
	</ul>
	</dd>
	<dt class="field-even">Returns<span class="colon">:</span></dt>
	<dd class="field-even"><p>arguments passed in to the kernel</p>
	</dd>
	<dt class="field-odd">Return type<span class="colon">:</span></dt>
	<dd class="field-odd"><p>cutlass.backend.GemmArguments</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py property">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.swizzling_functor">
	<em class="property"><span class="pre">property</span><span class="w"> </span></em><span class="sig-name descname"><span class="pre">swizzling_functor</span></span><a class="headerlink" href="#cutlass.op.gemm.Gemm.swizzling_functor" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns the type of the swizzling functor currently being used by the GEMM</p>
	<dl class="field-list simple">
	<dt class="field-odd">Returns<span class="colon">:</span></dt>
	<dd class="field-odd"><p>swizzing functor type</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.gemm.Gemm.tile_descriptions">
	<span class="sig-name descname"><span class="pre">tile_descriptions</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm.html#Gemm.tile_descriptions"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm.Gemm.tile_descriptions" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns a list of valid tile descriptions for the operations</p>
	<dl class="field-list simple">
	<dt class="field-odd">Returns<span class="colon">:</span></dt>
	<dd class="field-odd"><p>list of valid tile descriptions for the operations</p>
	</dd>
	<dt class="field-even">Return type<span class="colon">:</span></dt>
	<dd class="field-even"><p>list</p>
	</dd>
	</dl>
	</dd></dl>

	</dd></dl>

	</section>
	<section id="module-cutlass.op.gemm_grouped">
	<span id="grouped-gemm"></span><h2>Grouped GEMM<a class="headerlink" href="#module-cutlass.op.gemm_grouped" title="Permalink to this heading">#</a></h2>
	<p>Ease-of-use interface for constructing, compiling, and running GEMMs.</p>
	<p>The <code class="docutils literal notranslate"><span class="pre">GroupedGemm</span></code> interface is meant to allow one to easily instantiate, compile, and run
	grouped GEMM operations in CUTLASS via Python, without specifying many configuration parameters.
	Under the hood, the interface will select sensible default parameters for the many template
	parameters for CUTLASS grouped GEMMs.</p>
	<p>Note: optimal performance is not to be expected from this interface. To achieve optimal
	performance, one should specify and tune each configuration parameter.</p>
	<p>The simplest example of using this interface is the following:</p>
	<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="c1"># As, Bs, Cs, and Ds are torch/numpy/cupy tensor objects</span>
	<span class="n">plan</span> <span class="o">=</span> <span class="n">cutlass</span><span class="o">.</span><span class="n">op</span><span class="o">.</span><span class="n">GroupedGemm</span><span class="p">(</span><span class="n">element</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">DataType</span><span class="o">.</span><span class="n">f16</span><span class="p">,</span> <span class="n">layout</span><span class="o">=</span><span class="n">cutlass</span><span class="o">.</span><span class="n">LayoutType</span><span class="o">.</span><span class="n">RowMajor</span><span class="p">)</span>
	<span class="n">plan</span><span class="o">.</span><span class="n">run</span><span class="p">([</span><span class="n">A0</span><span class="p">,</span> <span class="n">A1</span><span class="p">],</span> <span class="p">[</span><span class="n">B0</span><span class="p">,</span> <span class="n">B1</span><span class="p">],</span> <span class="p">[</span><span class="n">C0</span><span class="p">,</span> <span class="n">C1</span><span class="p">],</span> <span class="p">[</span><span class="n">D0</span><span class="p">,</span> <span class="n">D1</span><span class="p">])</span>
	</pre></div>
	</div>
	<dl class="py class">
	<dt class="sig sig-object py" id="cutlass.op.gemm_grouped.GroupedGemm">
	<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">cutlass.op.gemm_grouped.</span></span><span class="sig-name descname"><span class="pre">GroupedGemm</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">D</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">beta</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_accumulator</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">element_D</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">layout_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cc</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm_grouped.html#GroupedGemm"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm_grouped.GroupedGemm" title="Permalink to this definition">#</a></dt>
	<dd><p>Bases: <a class="reference internal" href="#cutlass.op.gemm.Gemm" title="cutlass.op.gemm.Gemm"><code class="xref py py-class docutils literal notranslate"><span class="pre">Gemm</span></code></a></p>
	<p>Constructs a <code class="docutils literal notranslate"><span class="pre">GroupedGemm</span></code> object.</p>
	<p>The data types and layouts of operands A, B, and C, along with the data type of output D
	and that used for accumulation, are bound to the <code class="docutils literal notranslate"><span class="pre">GroupedGemm</span></code> object throughout its lifetime –
	these are not to be changed after a <code class="docutils literal notranslate"><span class="pre">GroupedGemm</span></code> has been constructed.</p>
	<p>The constructor has optional parameters for flexibly setting these parameters. Please see the constructor
	for <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> for examples of these.</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>cc</strong> (<em>int</em>) – compute capability of device to generate kernels for</p></li>
	<li><p><strong>A</strong> – tensor representing data type and layout of operands A</p></li>
	<li><p><strong>B</strong> – tensor representing data type and layout of operands B</p></li>
	<li><p><strong>C</strong> – tensor representing data type and layout of operands C</p></li>
	<li><p><strong>D</strong> – tensor representing data type and layout of operands D</p></li>
	<li><p><strong>alpha</strong> – scalar paramter alpha from GEMM computation that scales the product of operands A and B</p></li>
	<li><p><strong>beta</strong> – scalar parameter beta from GEMM operation that scales operand C</p></li>
	<li><p><strong>element_accumulator</strong> (<em>cutlass.DataType</em>) – data type to be used in accumulation of the product of operands A and B</p></li>
	<li><p><strong>element</strong> (<em>cutlass.DataType</em>) – generic data type to be used for operands A, B, C, D, as well as the accumulation data type</p></li>
	<li><p><strong>layout</strong> (<em>cutlass.LayoutType</em>) – generic layout type to be used for operands A, B, C, and D</p></li>
	<li><p><strong>element_A</strong> (<em>cutlass.DataType</em>) – data type to be used for operand A</p></li>
	<li><p><strong>element_B</strong> (<em>cutlass.DataType</em>) – data type to be used for operand B</p></li>
	<li><p><strong>element_C</strong> (<em>cutlass.DataType</em>) – data type to be used for operand C</p></li>
	<li><p><strong>element_D</strong> (<em>cutlass.DataType</em>) – data type to be used for operand D</p></li>
	<li><p><strong>layout_A</strong> (<em>layout</em><em> of </em><em>operand A</em>) – cutlass.LayoutType</p></li>
	<li><p><strong>layout_B</strong> (<em>layout</em><em> of </em><em>operand B</em>) – cutlass.LayoutType</p></li>
	<li><p><strong>layout_C</strong> (<em>layout</em><em> of </em><em>operand C</em>) – cutlass.LayoutType</p></li>
	<li><p><strong>layout_D</strong> (<em>layout</em><em> of </em><em>operand D</em>) – cutlass.LayoutType</p></li>
	</ul>
	</dd>
	</dl>
	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.gemm_grouped.GroupedGemm.construct">
	<span class="sig-name descname"><span class="pre">construct</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tile_description</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_A</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_B</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alignment_C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm_grouped.html#GroupedGemm.construct"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm_grouped.GroupedGemm.construct" title="Permalink to this definition">#</a></dt>
	<dd><p>Constructs a <code class="docutils literal notranslate"><span class="pre">cutlass.backend.GemmOperationGrouped</span></code> based on the input parameters and current
	kernel specification of the <code class="docutils literal notranslate"><span class="pre">Gemm</span></code> object.</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>tile_description</strong> (<em>cutlass.backend.TileDescription</em>) – tile description specifying shapes and operand types to use in the kernel</p></li>
	<li><p><strong>alignment_A</strong> (<em>int</em>) – alignment of operand A</p></li>
	<li><p><strong>alignment_B</strong> (<em>int</em>) – alignment of operand B</p></li>
	<li><p><strong>alignment_C</strong> (<em>int</em>) – alignment of operand C</p></li>
	</ul>
	</dd>
	<dt class="field-even">Returns<span class="colon">:</span></dt>
	<dd class="field-even"><p>operation that was constructed</p>
	</dd>
	<dt class="field-odd">Return type<span class="colon">:</span></dt>
	<dd class="field-odd"><p>cutlass.backend.GemmOperationGrouped</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.gemm_grouped.GroupedGemm.run">
	<span class="sig-name descname"><span class="pre">run</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">A</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">B</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">C</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">D</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">beta</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">sync</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">print_module</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/gemm_grouped.html#GroupedGemm.run"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.gemm_grouped.GroupedGemm.run" title="Permalink to this definition">#</a></dt>
	<dd><p>Runs the kernel currently specified.</p>
	<p>By default, this call returns only once the kernel has completed. To launch the kernel
	and immediately return, set <code class="docutils literal notranslate"><span class="pre">sync=False</span></code>. In this case, it is the responsibility of the
	caller to syncrhonize the results of the kernel before attempting to access outputs
	by calling <code class="docutils literal notranslate"><span class="pre">sync()</span></code> on the arguments returned from this call.</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>A</strong> (<em>list</em>) – list of tensors representing data type and layout of operand A</p></li>
	<li><p><strong>B</strong> (<em>list</em>) – list of tensors representing data type and layout of operand B</p></li>
	<li><p><strong>C</strong> (<em>list</em>) – list of tensors representing data type and layout of operand C</p></li>
	<li><p><strong>D</strong> (<em>list</em>) – list of tensors representing data type and layout of operand D</p></li>
	<li><p><strong>alpha</strong> – scalar paramter alpha from GEMM computation that scales the product of operands A and B</p></li>
	<li><p><strong>beta</strong> – scalar parameter beta from GEMM operation that scales operand C</p></li>
	<li><p><strong>sync</strong> (<em>bool</em>) – whether the call should wait for the kernel to complete before returning</p></li>
	<li><p><strong>print_module</strong> (<em>bool</em>) – whether to print the emitted C++ code</p></li>
	</ul>
	</dd>
	<dt class="field-even">Returns<span class="colon">:</span></dt>
	<dd class="field-even"><p>arguments passed in to the kernel</p>
	</dd>
	<dt class="field-odd">Return type<span class="colon">:</span></dt>
	<dd class="field-odd"><p>cutlass.backend.GemmGroupedArguments</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py property">
	<dt class="sig sig-object py" id="cutlass.op.gemm_grouped.GroupedGemm.swizzling_functor">
	<em class="property"><span class="pre">property</span><span class="w"> </span></em><span class="sig-name descname"><span class="pre">swizzling_functor</span></span><a class="headerlink" href="#cutlass.op.gemm_grouped.GroupedGemm.swizzling_functor" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns the type of the swizzling functor currently being used by the GEMM</p>
	<dl class="field-list simple">
	<dt class="field-odd">Returns<span class="colon">:</span></dt>
	<dd class="field-odd"><p>swizzing functor type</p>
	</dd>
	</dl>
	</dd></dl>

	</dd></dl>

	</section>
	<section id="module-cutlass.op.op">
	<span id="operation"></span><h2>Operation<a class="headerlink" href="#module-cutlass.op.op" title="Permalink to this heading">#</a></h2>
	<p>Base operation used for defining high-level CUTLASS operations (e.g., GEMM, Conv2d)</p>
	<dl class="py class">
	<dt class="sig sig-object py" id="cutlass.op.op.OperationBase">
	<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">cutlass.op.op.</span></span><span class="sig-name descname"><span class="pre">OperationBase</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">cc</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">kernel_cc</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/op.html#OperationBase"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.op.OperationBase" title="Permalink to this definition">#</a></dt>
	<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
	<p>Base operation used for defining high-level CUTLASS operations (e.g., GEMM, Conv2d)</p>
	<dl class="field-list simple">
	<dt class="field-odd">Parameters<span class="colon">:</span></dt>
	<dd class="field-odd"><ul class="simple">
	<li><p><strong>cc</strong> (<em>int</em>) – </p></li>
	<li><p><strong>kernel_cc</strong> (<em>int</em>) – </p></li>
	</ul>
	</dd>
	</dl>
	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.op.OperationBase.activations">
	<span class="sig-name descname"><span class="pre">activations</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/op.html#OperationBase.activations"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.op.OperationBase.activations" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns possible activation functions that can be used</p>
	<dl class="field-list simple">
	<dt class="field-odd">Returns<span class="colon">:</span></dt>
	<dd class="field-odd"><p>list of activation functions that can be used</p>
	</dd>
	<dt class="field-even">Return type<span class="colon">:</span></dt>
	<dd class="field-even"><p>list</p>
	</dd>
	</dl>
	</dd></dl>

	<dl class="py method">
	<dt class="sig sig-object py" id="cutlass.op.op.OperationBase.swizzling_functors">
	<span class="sig-name descname"><span class="pre">swizzling_functors</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/cutlass/op/op.html#OperationBase.swizzling_functors"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#cutlass.op.op.OperationBase.swizzling_functors" title="Permalink to this definition">#</a></dt>
	<dd><p>Returns possible swizzling functions that can be used</p>
	<dl class="field-list simple">
	<dt class="field-odd">Returns<span class="colon">:</span></dt>
	<dd class="field-odd"><p>list of swizzling functions that can be used</p>
	</dd>
	<dt class="field-even">Return type<span class="colon">:</span></dt>
	<dd class="field-even"><p>list</p>
	</dd>
	</dl>
	</dd></dl>

	</dd></dl>

	</section>
	</section>

	</article>
	</div>
	<footer>

	<div class="related-pages">
	<a class="next-page" href="cutlass.utils.html">
	<div class="page-info">
	<div class="context">
	<span>Next</span>
	</div>
	<div class="title">Utilities</div>
	</div>
	<svg class="furo-related-icon"><use href="#svg-arrow-right"></use></svg>
	</a>
	<a class="prev-page" href="cutlass.emit.html">
	<svg class="furo-related-icon"><use href="#svg-arrow-right"></use></svg>
	<div class="page-info">
	<div class="context">
	<span>Previous</span>
	</div>

	<div class="title">Emitters</div>

	</div>
	</a>
	</div>
	<div class="bottom-of-page">
	<div class="left-details">
	<div class="copyright">
	Copyright © 2023, NVIDIA
	</div>
	Made with <a href="https://www.sphinx-doc.org/">Sphinx</a> and <a class="muted-link" href="https://pradyunsg.me">@pradyunsg</a>'s

	<a href="https://github.com/pradyunsg/furo">Furo</a>

	</div>
	<div class="right-details">
	<div class="icons">
	<a class="muted-link " href="https://github.com/NVIDIA/cutlass" aria-label="GitHub">
	<svg stroke="currentColor" fill="currentColor" stroke-width="0" viewBox="0 0 16 16">
	<path fill-rule="evenodd" d="M8 0C3.58 0 0 3.58 0 8c0 3.54 2.29 6.53 5.47 7.59.4.07.55-.17.55-.38 0-.19-.01-.82-.01-1.49-2.01.37-2.53-.49-2.69-.94-.09-.23-.48-.94-.82-1.13-.28-.15-.68-.52-.01-.53.63-.01 1.08.58 1.23.82.72 1.21 1.87.87 2.33.66.07-.52.28-.87.51-1.07-1.78-.2-3.64-.89-3.64-3.95 0-.87.31-1.59.82-2.15-.08-.2-.36-1.02.08-2.12 0 0 .67-.21 2.2.82.64-.18 1.32-.27 2-.27.68 0 1.36.09 2 .27 1.53-1.04 2.2-.82 2.2-.82.44 1.1.16 1.92.08 2.12.51.56.82 1.27.82 2.15 0 3.07-1.87 3.75-3.65 3.95.29.25.54.73.54 1.48 0 1.07-.01 1.93-.01 2.2 0 .21.15.46.55.38A8.013 8.013 0 0 0 16 8c0-4.42-3.58-8-8-8z"></path>
	</svg>
	</a>

	</div>
	</div>
	</div>

	</footer>
	</div>
	<aside class="toc-drawer">


	<div class="toc-sticky toc-scroll">
	<div class="toc-title-container">
	<span class="toc-title">
	On this page
	</span>
	</div>
	<div class="toc-tree-container">
	<div class="toc-tree">
	<ul>
	<li><a class="reference internal" href="#">Operations</a><ul>
	<li><a class="reference internal" href="#module-cutlass.op.gemm">GEMM</a><ul>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm"><code class="docutils literal notranslate"><span class="pre">Gemm</span></code></a><ul>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.activation"><code class="docutils literal notranslate"><span class="pre">Gemm.activation</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.compile"><code class="docutils literal notranslate"><span class="pre">Gemm.compile()</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.construct"><code class="docutils literal notranslate"><span class="pre">Gemm.construct()</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.opclass"><code class="docutils literal notranslate"><span class="pre">Gemm.opclass</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.run"><code class="docutils literal notranslate"><span class="pre">Gemm.run()</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.swizzling_functor"><code class="docutils literal notranslate"><span class="pre">Gemm.swizzling_functor</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm.Gemm.tile_descriptions"><code class="docutils literal notranslate"><span class="pre">Gemm.tile_descriptions()</span></code></a></li>
	</ul>
	</li>
	</ul>
	</li>
	<li><a class="reference internal" href="#module-cutlass.op.gemm_grouped">Grouped GEMM</a><ul>
	<li><a class="reference internal" href="#cutlass.op.gemm_grouped.GroupedGemm"><code class="docutils literal notranslate"><span class="pre">GroupedGemm</span></code></a><ul>
	<li><a class="reference internal" href="#cutlass.op.gemm_grouped.GroupedGemm.construct"><code class="docutils literal notranslate"><span class="pre">GroupedGemm.construct()</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm_grouped.GroupedGemm.run"><code class="docutils literal notranslate"><span class="pre">GroupedGemm.run()</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.gemm_grouped.GroupedGemm.swizzling_functor"><code class="docutils literal notranslate"><span class="pre">GroupedGemm.swizzling_functor</span></code></a></li>
	</ul>
	</li>
	</ul>
	</li>
	<li><a class="reference internal" href="#module-cutlass.op.op">Operation</a><ul>
	<li><a class="reference internal" href="#cutlass.op.op.OperationBase"><code class="docutils literal notranslate"><span class="pre">OperationBase</span></code></a><ul>
	<li><a class="reference internal" href="#cutlass.op.op.OperationBase.activations"><code class="docutils literal notranslate"><span class="pre">OperationBase.activations()</span></code></a></li>
	<li><a class="reference internal" href="#cutlass.op.op.OperationBase.swizzling_functors"><code class="docutils literal notranslate"><span class="pre">OperationBase.swizzling_functors()</span></code></a></li>
	</ul>
	</li>
	</ul>
	</li>
	</ul>
	</li>
	</ul>

	</div>
	</div>
	</div>


	</aside>
	</div>
	</div><script data-url_root="./" id="documentation_options" src="_static/documentation_options.js"></script>
	<script src="_static/doctools.js"></script>
	<script src="_static/sphinx_highlight.js"></script>
	<script src="_static/scripts/furo.js"></script>
	<script src="_static/clipboard.min.js"></script>
	<script src="_static/copybutton.js"></script>
	<script src="_static/tabs.js"></script>
	<script crossorigin="anonymous" integrity="sha256-Ae2Vz/4ePdIu6ZyI/5ZGsYnb+m0JlOmKPjt6XZ9JJkA=" src="https://cdnjs.cloudflare.com/ajax/libs/require.js/2.3.4/require.min.js"></script>
	</body>
	</html>