| <!DOCTYPE html> |
| |
| <html lang="en"> |
| <head> |
| <meta charset="utf-8"/> |
| <meta content="IE=edge" http-equiv="X-UA-Compatible"/> |
| <meta content="width=device-width, initial-scale=1" name="viewport"/> |
| <title>Run MXNet on Multiple CPU/GPUs with Data Parallelism — mxnet documentation</title> |
| <link crossorigin="anonymous" href="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.6/css/bootstrap.min.css" integrity="sha384-1q8mTJOASx8j1Au+a5WDVnPi2lkFfwwEAa8hDDdjZlpLegxhjVME1fgjWPGmkzs7" rel="stylesheet"/> |
| <link href="https://maxcdn.bootstrapcdn.com/font-awesome/4.5.0/css/font-awesome.min.css" rel="stylesheet"/> |
| <link href="../_static/basic.css" rel="stylesheet" type="text/css"> |
| <link href="../_static/pygments.css" rel="stylesheet" type="text/css"> |
| <link href="../_static/mxnet.css" rel="stylesheet" type="text/css"/> |
| <script type="text/javascript"> |
| var DOCUMENTATION_OPTIONS = { |
| URL_ROOT: '../', |
| VERSION: '', |
| COLLAPSE_INDEX: false, |
| FILE_SUFFIX: '.html', |
| HAS_SOURCE: true, |
| SOURCELINK_SUFFIX: '' |
| }; |
| </script> |
| <script src="../_static/jquery-1.11.1.js" type="text/javascript"></script> |
| <script src="../_static/underscore.js" type="text/javascript"></script> |
| <script src="../_static/searchtools_custom.js" type="text/javascript"></script> |
| <script src="../_static/doctools.js" type="text/javascript"></script> |
| <script src="../_static/selectlang.js" type="text/javascript"></script> |
| <script src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML" type="text/javascript"></script> |
| <script type="text/javascript"> jQuery(function() { Search.loadIndex("/searchindex.js"); Search.init();}); </script> |
| <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-96378503-1', 'auto'); |
| ga('send', 'pageview'); |
| |
| </script> |
| <!-- --> |
| <!-- <script type="text/javascript" src="../_static/jquery.js"></script> --> |
| <!-- --> |
| <!-- <script type="text/javascript" src="../_static/underscore.js"></script> --> |
| <!-- --> |
| <!-- <script type="text/javascript" src="../_static/doctools.js"></script> --> |
| <!-- --> |
| <!-- <script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script> --> |
| <!-- --> |
| <link href="https://raw.githubusercontent.com/dmlc/web-data/master/mxnet/image/mxnet-icon.png" rel="icon" type="image/png"/> |
| </link></link></head> |
| <body role="document"><!-- Previous Navbar Layout |
| <div class="navbar navbar-default navbar-fixed-top"> |
| <div class="container"> |
| <div class="navbar-header"> |
| <button type="button" class="navbar-toggle collapsed" data-toggle="collapse" data-target="#navbar" aria-expanded="false" aria-controls="navbar"> |
| <span class="sr-only">Toggle navigation</span> |
| <span class="icon-bar"></span> |
| <span class="icon-bar"></span> |
| <span class="icon-bar"></span> |
| </button> |
| <a href="../" class="navbar-brand"> |
| <img src="http://data.mxnet.io/theme/mxnet.png"> |
| </a> |
| </div> |
| <div id="navbar" class="navbar-collapse collapse"> |
| <ul id="navbar" class="navbar navbar-left"> |
| |
| <li> <a href="../get_started/index.html">Get Started</a> </li> |
| |
| <li> <a href="../tutorials/index.html">Tutorials</a> </li> |
| |
| <li> <a href="../how_to/index.html">How To</a> </li> |
| |
| |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown" role="button" aria-haspopup="true" aria-expanded="true">Packages <span class="caret"></span></a> |
| <ul class="dropdown-menu"> |
| |
| <li><a href="../packages/python/index.html"> |
| Python |
| </a></li> |
| |
| <li><a href="../packages/r/index.html"> |
| R |
| </a></li> |
| |
| <li><a href="../packages/julia/index.html"> |
| Julia |
| </a></li> |
| |
| <li><a href="../packages/c++/index.html"> |
| C++ |
| </a></li> |
| |
| <li><a href="../packages/scala/index.html"> |
| Scala |
| </a></li> |
| |
| <li><a href="../packages/perl/index.html"> |
| Perl |
| </a></li> |
| |
| </ul> |
| </li> |
| |
| <li> <a href="../system/index.html">System</a> </li> |
| <li> |
| <form class="" role="search" action="../search.html" method="get" autocomplete="off"> |
| <div class="form-group inner-addon left-addon"> |
| <i class="glyphicon glyphicon-search"></i> |
| <input type="text" name="q" class="form-control" placeholder="Search"> |
| </div> |
| <input type="hidden" name="check_keywords" value="yes" /> |
| <input type="hidden" name="area" value="default" /> |
| |
| </form> </li> |
| </ul> |
| <ul id="navbar" class="navbar navbar-right"> |
| <li> <a href="../index.html"><span class="flag-icon flag-icon-us"></span></a> </li> |
| <li> <a href="..//zh/index.html"><span class="flag-icon flag-icon-cn"></span></a> </li> |
| </ul> |
| </div> |
| </div> |
| </div> |
| Previous Navbar Layout End --> |
| <div class="navbar navbar-fixed-top"> |
| <div class="container" id="navContainer"> |
| <div class="innder" id="header-inner"> |
| <h1 id="logo-wrap"> |
| <a href="../" id="logo"><img src="http://data.mxnet.io/theme/mxnet.png"/></a> |
| </h1> |
| <nav class="nav-bar" id="main-nav"> |
| <a class="main-nav-link" href="../get_started/install.html">Install</a> |
| <a class="main-nav-link" href="../tutorials/index.html">Tutorials</a> |
| <a class="main-nav-link" href="../how_to/index.html">How To</a> |
| <span id="dropdown-menu-position-anchor"> |
| <a aria-expanded="true" aria-haspopup="true" class="main-nav-link dropdown-toggle" data-toggle="dropdown" href="#" role="button">API <span class="caret"></span></a> |
| <ul class="dropdown-menu" id="package-dropdown-menu"> |
| <li><a class="main-nav-link" href="../api/python/index.html">Python</a></li> |
| <li><a class="main-nav-link" href="../api/scala/index.html">Scala</a></li> |
| <li><a class="main-nav-link" href="../api/r/index.html">R</a></li> |
| <li><a class="main-nav-link" href="../api/julia/index.html">Julia</a></li> |
| <li><a class="main-nav-link" href="../api/c++/index.html">C++</a></li> |
| <li><a class="main-nav-link" href="../api/perl/index.html">Perl</a></li> |
| </ul> |
| </span> |
| <a class="main-nav-link" href="../architecture/index.html">Architecture</a> |
| <!-- <a class="main-nav-link" href="../community/index.html">Community</a> --> |
| <a class="main-nav-link" href="https://github.com/dmlc/mxnet">Github</a> |
| <span id="dropdown-menu-position-anchor-version" style="position: relative"><a href="#" class="main-nav-link dropdown-toggle" data-toggle="dropdown" role="button" aria-haspopup="true" aria-expanded="true">Versions(master)<span class="caret"></span></a><ul id="package-dropdown-menu" class="dropdown-menu"><li><a class="main-nav-link" href=http://mxnet.incubator.apache.org/test/>v0.10.14</a></li><li><a class="main-nav-link" href=http://mxnet.incubator.apache.org/test/versions/0.10/index.html>0.10</a></li><li><a class="main-nav-link" href=http://mxnet.incubator.apache.org/test/versions/master/index.html>master</a></li></ul></span></nav> |
| <script> function getRootPath(){ return "../" } </script> |
| <div class="burgerIcon dropdown"> |
| <a class="dropdown-toggle" data-toggle="dropdown" href="#" role="button">☰</a> |
| <ul class="dropdown-menu dropdown-menu-right" id="burgerMenu"> |
| <li><a href="../get_started/install.html">Install</a></li> |
| <li><a href="../tutorials/index.html">Tutorials</a></li> |
| <li><a href="../how_to/index.html">How To</a></li> |
| <li class="dropdown-submenu"> |
| <a href="#" tabindex="-1">API</a> |
| <ul class="dropdown-menu"> |
| <li><a href="../api/python/index.html" tabindex="-1">Python</a> |
| </li> |
| <li><a href="../api/scala/index.html" tabindex="-1">Scala</a> |
| </li> |
| <li><a href="../api/r/index.html" tabindex="-1">R</a> |
| </li> |
| <li><a href="../api/julia/index.html" tabindex="-1">Julia</a> |
| </li> |
| <li><a href="../api/c++/index.html" tabindex="-1">C++</a> |
| </li> |
| <li><a href="../api/perl/index.html" tabindex="-1">Perl</a> |
| </li> |
| </ul> |
| </li> |
| <li><a href="../architecture/index.html">Architecture</a></li> |
| <li><a class="main-nav-link" href="https://github.com/dmlc/mxnet">Github</a></li> |
| <li id="dropdown-menu-position-anchor-version-mobile" class="dropdown-submenu" style="position: relative"><a href="#" tabindex="-1">Versions(master)</a><ul class="dropdown-menu"><li><a tabindex="-1" href=http://mxnet.incubator.apache.org/test/>v0.10.14</a></li><li><a tabindex="-1" href=http://mxnet.incubator.apache.org/test/versions/0.10/index.html>0.10</a></li><li><a tabindex="-1" href=http://mxnet.incubator.apache.org/test/versions/master/index.html>master</a></li></ul></li></ul> |
| </div> |
| <div class="plusIcon dropdown"> |
| <a class="dropdown-toggle" data-toggle="dropdown" href="#" role="button"><span aria-hidden="true" class="glyphicon glyphicon-plus"></span></a> |
| <ul class="dropdown-menu dropdown-menu-right" id="plusMenu"></ul> |
| </div> |
| <div id="search-input-wrap"> |
| <form action="../search.html" autocomplete="off" class="" method="get" role="search"> |
| <div class="form-group inner-addon left-addon"> |
| <i class="glyphicon glyphicon-search"></i> |
| <input class="form-control" name="q" placeholder="Search" type="text"/> |
| </div> |
| <input name="check_keywords" type="hidden" value="yes"> |
| <input name="area" type="hidden" value="default"/> |
| </input></form> |
| <div id="search-preview"></div> |
| </div> |
| <div id="searchIcon"> |
| <span aria-hidden="true" class="glyphicon glyphicon-search"></span> |
| </div> |
| <!-- <div id="lang-select-wrap"> --> |
| <!-- <label id="lang-select-label"> --> |
| <!-- <\!-- <i class="fa fa-globe"></i> -\-> --> |
| <!-- <span></span> --> |
| <!-- </label> --> |
| <!-- <select id="lang-select"> --> |
| <!-- <option value="en">Eng</option> --> |
| <!-- <option value="zh">中文</option> --> |
| <!-- </select> --> |
| <!-- </div> --> |
| <!-- <a id="mobile-nav-toggle"> |
| <span class="mobile-nav-toggle-bar"></span> |
| <span class="mobile-nav-toggle-bar"></span> |
| <span class="mobile-nav-toggle-bar"></span> |
| </a> --> |
| </div> |
| </div> |
| </div> |
| <div class="container"> |
| <div class="row"> |
| <div aria-label="main navigation" class="sphinxsidebar leftsidebar" role="navigation"> |
| <div class="sphinxsidebarwrapper"> |
| <ul> |
| <li class="toctree-l1"><a class="reference internal" href="../api/python/index.html">Python Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../api/r/index.html">R Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../api/julia/index.html">Julia Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../api/c++/index.html">C++ Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../api/scala/index.html">Scala Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../api/perl/index.html">Perl Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="index.html">HowTo Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../architecture/index.html">System Documents</a></li> |
| <li class="toctree-l1"><a class="reference internal" href="../tutorials/index.html">Tutorials</a></li> |
| </ul> |
| </div> |
| </div> |
| <div class="content"> |
| <div class="section" id="run-mxnet-on-multiple-cpu-gpus-with-data-parallelism"> |
| <span id="run-mxnet-on-multiple-cpu-gpus-with-data-parallelism"></span><h1>Run MXNet on Multiple CPU/GPUs with Data Parallelism<a class="headerlink" href="#run-mxnet-on-multiple-cpu-gpus-with-data-parallelism" title="Permalink to this headline">¶</a></h1> |
| <p><em>MXNet</em> supports training with multiple CPUs and GPUs, which may be located on different physical machines.</p> |
| <div class="section" id="data-parallelism-vs-model-parallelism"> |
| <span id="data-parallelism-vs-model-parallelism"></span><h2>Data Parallelism vs Model Parallelism<a class="headerlink" href="#data-parallelism-vs-model-parallelism" title="Permalink to this headline">¶</a></h2> |
| <p>By default, <em>MXNet</em> uses data parallelism to partition the workload over multiple |
| devices. |
| Assume there are <em>n</em> devices. |
| Then each one will receive a copy of the complete model |
| and train it on <em>1/n</em> of the data. |
| The results such as gradients and |
| updated model are communicated across these devices.</p> |
| <p>MXNet also supports model parallelism. |
| In this approach, each device holds onto only part of the model. |
| This proves useful when the model is too large to fit onto a single device. |
| As an example, see the following <a class="reference internal" href="model_parallel_lstm.html"><em>tutorial</em></a> |
| which shows how to use model parallelism for training a multi-layer LSTM model. |
| In this tutorial, we’ll focus on data parallelism.</p> |
| </div> |
| <div class="section" id="multiple-gpus-within-a-single-machine"> |
| <span id="multiple-gpus-within-a-single-machine"></span><h2>Multiple GPUs within a Single Machine<a class="headerlink" href="#multiple-gpus-within-a-single-machine" title="Permalink to this headline">¶</a></h2> |
| <div class="section" id="workload-partitioning"> |
| <span id="workload-partitioning"></span><h3>Workload Partitioning<a class="headerlink" href="#workload-partitioning" title="Permalink to this headline">¶</a></h3> |
| <p>By default, <em>MXNet</em> partitions a data batch evenly among the available GPUs. |
| Assume a batch size <em>b</em> and assume there are <em>k</em> GPUs, then in one iteration |
| each GPU will perform forward and backward on <em>b/k</em> examples. The |
| gradients are then summed over all GPUs before updating the model.</p> |
| </div> |
| <div class="section" id="how-to-use"> |
| <span id="how-to-use"></span><h3>How to Use<a class="headerlink" href="#how-to-use" title="Permalink to this headline">¶</a></h3> |
| <blockquote> |
| <div>To use GPUs, we need to compile MXNet with GPU support. For |
| example, set <code class="docutils literal"><span class="pre">USE_CUDA=1</span></code> in <code class="docutils literal"><span class="pre">config.mk</span></code> before <code class="docutils literal"><span class="pre">make</span></code>. (see |
| <a class="reference external" href="http://mxnet.io/get_started/install.html">MXNet installation guide</a> for more options).</div></blockquote> |
| <p>If a machine has one or more GPU cards installed, |
| then each card is labeled by a number starting from 0. |
| To use a particular GPU, one can either |
| specify the context <code class="docutils literal"><span class="pre">context</span></code> in code |
| or pass <code class="docutils literal"><span class="pre">--gpus</span></code> at the command line. |
| For example, to use GPU 0 and 2 in python, |
| one can typically create a module with</p> |
| <div class="highlight-python"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">mxnet</span> <span class="kn">as</span> <span class="nn">mx</span> |
| <span class="n">module</span> <span class="o">=</span> <span class="n">mx</span><span class="o">.</span><span class="n">module</span><span class="o">.</span><span class="n">Module</span><span class="p">(</span><span class="n">context</span><span class="o">=</span><span class="p">[</span><span class="n">mx</span><span class="o">.</span><span class="n">gpu</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="n">mx</span><span class="o">.</span><span class="n">gpu</span><span class="p">(</span><span class="mi">2</span><span class="p">)],</span> <span class="o">...</span><span class="p">)</span> |
| </pre></div> |
| </div> |
| <p>while if the program accepts a <code class="docutils literal"><span class="pre">--gpus</span></code> flag (as seen in |
| <a class="reference external" href="https://github.com/dmlc/mxnet/tree/master/example/image-classification">example/image-classification</a>), |
| then we can try</p> |
| <div class="highlight-bash"><div class="highlight"><pre><span></span>python train_mnist.py --gpus <span class="m">0</span>,2 ... |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="advanced-usage"> |
| <span id="advanced-usage"></span><h3>Advanced Usage<a class="headerlink" href="#advanced-usage" title="Permalink to this headline">¶</a></h3> |
| <p>If the available GPUs are not all equally powerful, |
| we can partition the workload accordingly. |
| For example, if GPU 0 is 3 times faster than GPU 2, |
| then we might use the workload option <code class="docutils literal"><span class="pre">work_load_list=[3,</span> <span class="pre">1]</span></code>, |
| see <a class="reference external" href="http://mxnet.io/api/python/module.html#mxnet.module.Module">Module</a> |
| for more details.</p> |
| <p>Training with multiple GPUs should yield the same results |
| as training on a single GPU if all other hyper-parameters are the same. |
| In practice, the results may exhibit small differences, |
| owing to the randomness of I/O (random order or other augmentations), |
| weight initialization with different seeds, and CUDNN.</p> |
| <p>We can control on which devices the gradient is aggregated |
| and on which device the model is updated via <a class="reference external" href="http://mxnet.io/api/python/kvstore.html"><code class="docutils literal"><span class="pre">KVStore</span></code></a>, |
| the <em>MXNet</em> module that supports data communication. |
| One can either use <code class="docutils literal"><span class="pre">mx.kvstore.create(type)</span></code> to get an instance |
| or use the program flag <code class="docutils literal"><span class="pre">--kv-store</span> <span class="pre">type</span></code>.</p> |
| <p>There are two commonly used types,</p> |
| <ul class="simple"> |
| <li><code class="docutils literal"><span class="pre">local</span></code>: all gradients are copied to CPU memory and weights are updated there.</li> |
| <li><code class="docutils literal"><span class="pre">device</span></code>: both gradient aggregation and weight updates are run on GPUs. |
| With this setting, the <code class="docutils literal"><span class="pre">KVStore</span></code> also attempts to use GPU peer-to-peer communication, |
| potentially accelerating the communication. |
| Note that this option may result in higher GPU memory usage.</li> |
| </ul> |
| <p>When using a large number of GPUs, e.g. >=4, we suggest using <code class="docutils literal"><span class="pre">device</span></code> for better performance.</p> |
| </div> |
| </div> |
| <div class="section" id="distributed-training-with-multiple-machines"> |
| <span id="distributed-training-with-multiple-machines"></span><h2>Distributed Training with Multiple Machines<a class="headerlink" href="#distributed-training-with-multiple-machines" title="Permalink to this headline">¶</a></h2> |
| <p><code class="docutils literal"><span class="pre">KVStore</span></code> also supports a number of options for running on multiple machines.</p> |
| <ul class="simple"> |
| <li><code class="docutils literal"><span class="pre">dist_sync</span></code> behaves similarly to <code class="docutils literal"><span class="pre">local</span></code> but exhibits one major difference. |
| With <code class="docutils literal"><span class="pre">dist_sync</span></code>, <code class="docutils literal"><span class="pre">batch-size</span></code> now means the batch size used on each machine. |
| So if there are <em>n</em> machines and we use batch size <em>b</em>, |
| then <code class="docutils literal"><span class="pre">dist_sync</span></code> behaves like <code class="docutils literal"><span class="pre">local</span></code> with batch size <em>n*b</em>.</li> |
| <li><code class="docutils literal"><span class="pre">dist_device_sync</span></code> is similar to <code class="docutils literal"><span class="pre">dist_sync</span></code>. The difference between them is that |
| <code class="docutils literal"><span class="pre">dist_device_sync</span></code> aggregates gradients and updates weight on GPUs |
| while <code class="docutils literal"><span class="pre">dist_sync</span></code> does so on CPU memory.</li> |
| <li><code class="docutils literal"><span class="pre">dist_async</span></code> performs asynchronous updates. |
| The weight is updated whenever gradients are received from any machine. |
| The update is atomic, i.e., no two updates happen on the same weight at the same time. |
| However, the order is not guaranteed.</li> |
| </ul> |
| <div class="section" id="how-to-launch-a-job"> |
| <span id="how-to-launch-a-job"></span><h3>How to Launch a Job<a class="headerlink" href="#how-to-launch-a-job" title="Permalink to this headline">¶</a></h3> |
| <blockquote> |
| <div>To use distributed training, we need to compile with <code class="docutils literal"><span class="pre">USE_DIST_KVSTORE=1</span></code> |
| (see <a class="reference external" href="http://mxnet.io/get_started/install.html">MXNet installation guide</a> for more options).</div></blockquote> |
| <p>Launching a distributed job is a bit different from running on a single |
| machine. MXNet provides |
| <a class="reference external" href="https://github.com/dmlc/mxnet/blob/master/tools/launch.py">tools/launch.py</a> to |
| start a job by using <code class="docutils literal"><span class="pre">ssh</span></code>, <code class="docutils literal"><span class="pre">mpi</span></code>, <code class="docutils literal"><span class="pre">sge</span></code>, or <code class="docutils literal"><span class="pre">yarn</span></code>.</p> |
| <p>An easy way to set up a cluster of EC2 instances for distributed deep learning |
| is using an <a class="reference external" href="https://github.com/awslabs/deeplearning-cfn">AWS CloudFormation template</a>. |
| If you do not have a cluster, you can check the repository before you continue.</p> |
| <p>Assume we are at the directory <code class="docutils literal"><span class="pre">mxnet/example/image-classification</span></code> |
| and want to train LeNet to classify MNIST images, as demonstrated here: |
| <a class="reference external" href="https://github.com/dmlc/mxnet/blob/master/example/image-classification/train_mnist.py">train_mnist.py</a>.</p> |
| <p>On a single machine, we can run:</p> |
| <div class="highlight-bash"><div class="highlight"><pre><span></span>python train_mnist.py --network lenet |
| </pre></div> |
| </div> |
| <p>Now, say we are given two ssh-able machines and <em>MXNet</em> is installed on both machines. |
| We want to train LeNet on these two machines. |
| First, we save the IPs (or hostname) of these two machines in file <code class="docutils literal"><span class="pre">hosts</span></code>, e.g.</p> |
| <div class="highlight-bash"><div class="highlight"><pre><span></span>$ cat hosts |
| <span class="m">172</span>.30.0.172 |
| <span class="m">172</span>.30.0.171 |
| </pre></div> |
| </div> |
| <p>Next, if the mxnet folder is accessible from both machines, e.g. on a |
| <a class="reference external" href="https://help.ubuntu.com/lts/serverguide/network-file-system.html">network filesystem</a>, |
| then we can run:</p> |
| <div class="highlight-bash"><div class="highlight"><pre><span></span>python ../../tools/launch.py -n <span class="m">2</span> --launcher ssh -H hosts python train_mnist.py --network lenet --kv-store dist_sync |
| </pre></div> |
| </div> |
| <p>Note that here we</p> |
| <ul class="simple"> |
| <li>use <code class="docutils literal"><span class="pre">launch.py</span></code> to submit the job.</li> |
| <li>provide launcher, <code class="docutils literal"><span class="pre">ssh</span></code> if all machines are ssh-able, <code class="docutils literal"><span class="pre">mpi</span></code> if <code class="docutils literal"><span class="pre">mpirun</span></code> is |
| available, <code class="docutils literal"><span class="pre">sge</span></code> for Sun Grid Engine, and <code class="docutils literal"><span class="pre">yarn</span></code> for Apache Yarn.</li> |
| <li><code class="docutils literal"><span class="pre">-n</span></code> number of worker nodes to run on</li> |
| <li><code class="docutils literal"><span class="pre">-H</span></code> the host file which is required by <code class="docutils literal"><span class="pre">ssh</span></code> and <code class="docutils literal"><span class="pre">mpi</span></code></li> |
| <li><code class="docutils literal"><span class="pre">--kv-store</span></code> use either <code class="docutils literal"><span class="pre">dist_sync</span></code> or <code class="docutils literal"><span class="pre">dist_async</span></code></li> |
| </ul> |
| </div> |
| <div class="section" id="synchronize-directory"> |
| <span id="synchronize-directory"></span><h3>Synchronize Directory<a class="headerlink" href="#synchronize-directory" title="Permalink to this headline">¶</a></h3> |
| <p>Now consider if the mxnet folder is not accessible. |
| We can first copy the <code class="docutils literal"><span class="pre">MXNet</span></code> library to this folder by</p> |
| <div class="highlight-bash"><div class="highlight"><pre><span></span>cp -r ../../python/mxnet . |
| cp -r ../../lib/libmxnet.so mxnet |
| </pre></div> |
| </div> |
| <p>then ask <code class="docutils literal"><span class="pre">launch.py</span></code> to synchronize the current directory to all machines’ |
| <code class="docutils literal"><span class="pre">/tmp/mxnet</span></code> directory with <code class="docutils literal"><span class="pre">--sync-dst-dir</span></code></p> |
| <div class="highlight-bash"><div class="highlight"><pre><span></span>python ../../tools/launch.py -n <span class="m">2</span> -H hosts --sync-dst-dir /tmp/mxnet <span class="se">\</span> |
| python train_mnist.py --network lenet --kv-store dist_sync |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="use-a-particular-network-interface"> |
| <span id="use-a-particular-network-interface"></span><h3>Use a Particular Network Interface<a class="headerlink" href="#use-a-particular-network-interface" title="Permalink to this headline">¶</a></h3> |
| <p><em>MXNet</em> often chooses the first available network interface. |
| But for machines that have multiple interfaces, |
| we can specify which network interface to use for data |
| communication by the environment variable <code class="docutils literal"><span class="pre">DMLC_INTERFACE</span></code>. |
| For example, to use the interface <code class="docutils literal"><span class="pre">eth0</span></code>, we can</p> |
| <div class="highlight-python"><div class="highlight"><pre><span></span>export DMLC_INTERFACE=eth0; python ../../tools/launch.py ... |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="debug-connection"> |
| <span id="debug-connection"></span><h3>Debug Connection<a class="headerlink" href="#debug-connection" title="Permalink to this headline">¶</a></h3> |
| <p>Set<code class="docutils literal"><span class="pre">PS_VERBOSE=1</span></code> to see the debug logging, e.g</p> |
| <div class="highlight-python"><div class="highlight"><pre><span></span>export PS_VERBOSE=1; python ../../tools/launch.py ... |
| </pre></div> |
| </div> |
| </div> |
| <div class="section" id="more"> |
| <span id="more"></span><h3>More<a class="headerlink" href="#more" title="Permalink to this headline">¶</a></h3> |
| <ul class="simple"> |
| <li>See more launch options by <code class="docutils literal"><span class="pre">python</span> <span class="pre">../../tools/launch.py</span> <span class="pre">-h</span></code></li> |
| <li>See more options of <a class="reference external" href="http://ps-lite.readthedocs.org/en/latest/how_to.html">ps-lite</a></li> |
| </ul> |
| </div> |
| </div> |
| </div> |
| <div class="container"> |
| <div class="footer"> |
| <p> © 2015-2017 DMLC. All rights reserved. </p> |
| </div> |
| </div> |
| </div> |
| <div aria-label="main navigation" class="sphinxsidebar rightsidebar" role="navigation"> |
| <div class="sphinxsidebarwrapper"> |
| <h3><a href="../index.html">Table Of Contents</a></h3> |
| <ul> |
| <li><a class="reference internal" href="#">Run MXNet on Multiple CPU/GPUs with Data Parallelism</a><ul> |
| <li><a class="reference internal" href="#data-parallelism-vs-model-parallelism">Data Parallelism vs Model Parallelism</a></li> |
| <li><a class="reference internal" href="#multiple-gpus-within-a-single-machine">Multiple GPUs within a Single Machine</a><ul> |
| <li><a class="reference internal" href="#workload-partitioning">Workload Partitioning</a></li> |
| <li><a class="reference internal" href="#how-to-use">How to Use</a></li> |
| <li><a class="reference internal" href="#advanced-usage">Advanced Usage</a></li> |
| </ul> |
| </li> |
| <li><a class="reference internal" href="#distributed-training-with-multiple-machines">Distributed Training with Multiple Machines</a><ul> |
| <li><a class="reference internal" href="#how-to-launch-a-job">How to Launch a Job</a></li> |
| <li><a class="reference internal" href="#synchronize-directory">Synchronize Directory</a></li> |
| <li><a class="reference internal" href="#use-a-particular-network-interface">Use a Particular Network Interface</a></li> |
| <li><a class="reference internal" href="#debug-connection">Debug Connection</a></li> |
| <li><a class="reference internal" href="#more">More</a></li> |
| </ul> |
| </li> |
| </ul> |
| </li> |
| </ul> |
| </div> |
| </div> |
| </div> <!-- pagename != index --> |
| <script crossorigin="anonymous" integrity="sha384-0mSbJDEHialfmuBBQP6A4Qrprq5OVfW37PRR3j5ELqxss1yVqOtnepnHVP9aJ7xS" src="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.6/js/bootstrap.min.js"></script> |
| <script src="../_static/js/sidebar.js" type="text/javascript"></script> |
| <script src="../_static/js/search.js" type="text/javascript"></script> |
| <script src="../_static/js/navbar.js" type="text/javascript"></script> |
| <script src="../_static/js/clipboard.min.js" type="text/javascript"></script> |
| <script src="../_static/js/copycode.js" type="text/javascript"></script> |
| <script type="text/javascript"> |
| $('body').ready(function () { |
| $('body').css('visibility', 'visible'); |
| }); |
| </script> |
| </div></body> |
| </html> |
