<!DOCTYPE html><html lang="en"><head><meta charSet="utf-8"/><meta http-equiv="X-UA-Compatible" content="IE=edge"/><title>Train CNN over Cifar-10 · Apache SINGA</title><meta name="viewport" content="width=device-width"/><meta name="generator" content="Docusaurus"/><meta name="description" content="&lt;!--- Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements.  See the NOTICE file distributed with this work for additional information regarding copyright ownership.  The ASF licenses this file to you under the Apache License, Version 2.0 (the &quot;License&quot;); you may not use this file except in compliance with the License.  You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an &quot;AS IS&quot; BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.  See the License for the specific language governing permissions and limitations under the License.  --&gt;"/><meta name="docsearch:version" content="next"/><meta name="docsearch:language" content="en"/><meta property="og:title" content="Train CNN over Cifar-10 · Apache SINGA"/><meta property="og:type" content="website"/><meta property="og:url" content="https://feynmandna.github.io/"/><meta property="og:description" content="&lt;!--- Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements.  See the NOTICE file distributed with this work for additional information regarding copyright ownership.  The ASF licenses this file to you under the Apache License, Version 2.0 (the &quot;License&quot;); you may not use this file except in compliance with the License.  You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an &quot;AS IS&quot; BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.  See the License for the specific language governing permissions and limitations under the License.  --&gt;"/><meta property="og:image" content="https://feynmandna.github.io/img/singa_twitter_banner.jpeg"/><meta name="twitter:card" content="summary"/><meta name="twitter:image" content="https://feynmandna.github.io/img/singa_twitter_banner.jpeg"/><link rel="shortcut icon" href="/img/favicon.ico"/><link rel="stylesheet" href="//cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/atom-one-dark.min.css"/><link rel="alternate" type="application/atom+xml" href="https://feynmandna.github.io/blog/atom.xml" title="Apache SINGA Blog ATOM Feed"/><link rel="alternate" type="application/rss+xml" href="https://feynmandna.github.io/blog/feed.xml" title="Apache SINGA Blog RSS Feed"/><script type="text/javascript" src="https://buttons.github.io/buttons.js"></script><script src="https://unpkg.com/vanilla-back-to-top@7.1.14/dist/vanilla-back-to-top.min.js"></script><script>
        document.addEventListener('DOMContentLoaded', function() {
          addBackToTop(
            {"zIndex":100}
          )
        });
        </script><script src="/js/scrollSpy.js"></script><link rel="stylesheet" href="/css/main.css"/><script src="/js/codetabs.js"></script></head><body class="sideNavVisible separateOnPageNav"><div class="fixedHeaderContainer"><div class="headerWrapper wrapper"><header><a href="/"><img class="logo" src="/img/singa.png" alt="Apache SINGA"/></a><a href="/versions"><h3>next</h3></a><div class="navigationWrapper navigationSlider"><nav class="slidingNav"><ul class="nav-site nav-site-internal"><li class="siteNavGroupActive"><a href="/docs/next/installation" target="_self">Docs</a></li><li class=""><a href="/docs/next/source-repository" target="_self">Community</a></li><li class=""><a href="/blog/" target="_self">News</a></li><li class=""><a target="_self"></a></li><li class=""><a href="https://github.com/apache/singa-doc" target="_self">GitHub</a></li></ul></nav></div></header></div></div><div class="navPusher"><div class="docMainWrapper wrapper"><div class="docsNavContainer" id="docsNav"><nav class="toc"><div class="toggleNav"><section class="navWrapper wrapper"><div class="navBreadcrumb wrapper"><div class="navToggle" id="navToggler"><div class="hamburger-menu"><div class="line1"></div><div class="line2"></div><div class="line3"></div></div></div><h2><i>›</i><span>Model Zoo</span></h2><div class="tocToggler" id="tocToggler"><i class="icon-toc"></i></div></div><div class="navGroups"><div class="navGroup"><h3 class="navGroupCategoryTitle">Getting Started</h3><ul class=""><li class="navListItem"><a class="navItem" href="/docs/next/installation">Installation</a></li><li class="navListItem"><a class="navItem" href="/docs/next/software-stack">Software Stack</a></li><li class="navListItem"><a class="navItem" href="/docs/next/benchmark-train">Benchmark for Distributed Training</a></li></ul></div><div class="navGroup"><h3 class="navGroupCategoryTitle">Guides</h3><ul class=""><li class="navListItem"><a class="navItem" href="/docs/next/device">Device</a></li><li class="navListItem"><a class="navItem" href="/docs/next/tensor">Tensor</a></li><li class="navListItem"><a class="navItem" href="/docs/next/autograd">Autograd in SINGA</a></li><li class="navListItem"><a class="navItem" href="/docs/next/graph">Computational Graph</a></li><li class="navListItem"><a class="navItem" href="/docs/next/dist-train">Distributed Training</a></li></ul></div><div class="navGroup"><h3 class="navGroupCategoryTitle">Model Zoo</h3><ul class=""><li class="navListItem navListItemActive"><a class="navItem" href="/docs/next/model-zoo-cnn-cifar10">Train CNN over Cifar-10</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-char-rnn">Train Char-RNN over plain text</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-rbm-mnist">Train a RBM model against MNIST dataset</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-imagenet-alexnet">Train AlexNet over ImageNet</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-imagenet-densenet">Image Classification using DenseNet</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-imagenet-googlenet">Image Classification using GoogleNet</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-imagenet-inception">Image Classification using Inception V4</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-imagenet-resnet">Image Classification using Residual Networks</a></li><li class="navListItem"><a class="navItem" href="/docs/next/model-zoo-imagenet-vgg">Image Classification using VGG</a></li></ul></div><div class="navGroup"><h3 class="navGroupCategoryTitle">Development</h3><ul class=""><li class="navListItem"><a class="navItem" href="/docs/next/download-singa">Download SINGA</a></li><li class="navListItem"><a class="navItem" href="/docs/next/build">Build SINGA from Source</a></li><li class="navListItem"><a class="navItem" href="/docs/next/contribute-code">How to Contribute Code</a></li><li class="navListItem"><a class="navItem" href="/docs/next/contribute-docs">How to Contribute to Documentation</a></li><li class="navListItem"><a class="navItem" href="/docs/next/how-to-release">How to Prepare a Release</a></li><li class="navListItem"><a class="navItem" href="/docs/next/git-workflow">Git Workflow</a></li></ul></div></div></section></div><script>
            var coll = document.getElementsByClassName('collapsible');
            var checkActiveCategory = true;
            for (var i = 0; i < coll.length; i++) {
              var links = coll[i].nextElementSibling.getElementsByTagName('*');
              if (checkActiveCategory){
                for (var j = 0; j < links.length; j++) {
                  if (links[j].classList.contains('navListItemActive')){
                    coll[i].nextElementSibling.classList.toggle('hide');
                    coll[i].childNodes[1].classList.toggle('rotate');
                    checkActiveCategory = false;
                    break;
                  }
                }
              }

              coll[i].addEventListener('click', function() {
                var arrow = this.childNodes[1];
                arrow.classList.toggle('rotate');
                var content = this.nextElementSibling;
                content.classList.toggle('hide');
              });
            }

            document.addEventListener('DOMContentLoaded', function() {
              createToggler('#navToggler', '#docsNav', 'docsSliderActive');
              createToggler('#tocToggler', 'body', 'tocActive');

              var headings = document.querySelector('.toc-headings');
              headings && headings.addEventListener('click', function(event) {
                var el = event.target;
                while(el !== headings){
                  if (el.tagName === 'A') {
                    document.body.classList.remove('tocActive');
                    break;
                  } else{
                    el = el.parentNode;
                  }
                }
              }, false);

              function createToggler(togglerSelector, targetSelector, className) {
                var toggler = document.querySelector(togglerSelector);
                var target = document.querySelector(targetSelector);

                if (!toggler) {
                  return;
                }

                toggler.onclick = function(event) {
                  event.preventDefault();

                  target.classList.toggle(className);
                };
              }
            });
        </script></nav></div><div class="container mainContainer docsContainer"><div class="wrapper"><div class="post"><header class="postHeader"><a class="edit-page-link button" href="https://github.com/apache/singa-doc/blob/master/docs/model-zoo-cnn-cifar10.md" target="_blank" rel="noreferrer noopener">Edit</a><h1 id="__docusaurus" class="postHeaderTitle">Train CNN over Cifar-10</h1></header><article><div><span><!--- Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements.  See the NOTICE file distributed with this work for additional information regarding copyright ownership.  The ASF licenses this file to you under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License.  You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.  See the License for the specific language governing permissions and limitations under the License.  -->
<p>Convolution neural network (CNN) is a type of feed-forward artificial neural
network widely used for image and video classification. In this example, we will
train three deep CNN models to do image classification for the CIFAR-10 dataset,</p>
<ol>
<li><p><a href="https://code.google.com/p/cuda-convnet/source/browse/trunk/example-layers/layers-18pct.cfg">AlexNet</a>
the best validation accuracy (without data augmentation) we achieved was
about 82%.</p></li>
<li><p><a href="http://torch.ch/blog/2015/07/30/cifar.html">VGGNet</a>, the best validation
accuracy (without data augmentation) we achieved was about 89%.</p></li>
<li><p><a href="https://github.com/facebook/fb.resnet.torch">ResNet</a>, the best validation
accuracy (without data augmentation) we achieved was about 83%.</p></li>
<li><p><a href="https://github.com/BVLC/caffe/tree/master/examples/cifar10">Alexnet from Caffe</a>,
SINGA is able to convert model from Caffe seamlessly.</p></li>
</ol>
<h2><a class="anchor" aria-hidden="true" id="instructions"></a><a href="#instructions" aria-hidden="true" class="hash-link"><svg class="hash-link-icon" aria-hidden="true" height="16" version="1.1" viewBox="0 0 16 16" width="16"><path fill-rule="evenodd" d="M4 9h1v1H4c-1.5 0-3-1.69-3-3.5S2.55 3 4 3h4c1.45 0 3 1.69 3 3.5 0 1.41-.91 2.72-2 3.25V8.59c.58-.45 1-1.27 1-2.09C10 5.22 8.98 4 8 4H4c-.98 0-2 1.22-2 2.5S3 9 4 9zm9-3h-1v1h1c1 0 2 1.22 2 2.5S13.98 12 13 12H9c-.98 0-2-1.22-2-2.5 0-.83.42-1.64 1-2.09V6.25c-1.09.53-2 1.84-2 3.25C6 11.31 7.55 13 9 13h4c1.45 0 3-1.69 3-3.5S14.5 6 13 6z"></path></svg></a>Instructions</h2>
<h3><a class="anchor" aria-hidden="true" id="singa-installation"></a><a href="#singa-installation" aria-hidden="true" class="hash-link"><svg class="hash-link-icon" aria-hidden="true" height="16" version="1.1" viewBox="0 0 16 16" width="16"><path fill-rule="evenodd" d="M4 9h1v1H4c-1.5 0-3-1.69-3-3.5S2.55 3 4 3h4c1.45 0 3 1.69 3 3.5 0 1.41-.91 2.72-2 3.25V8.59c.58-.45 1-1.27 1-2.09C10 5.22 8.98 4 8 4H4c-.98 0-2 1.22-2 2.5S3 9 4 9zm9-3h-1v1h1c1 0 2 1.22 2 2.5S13.98 12 13 12H9c-.98 0-2-1.22-2-2.5 0-.83.42-1.64 1-2.09V6.25c-1.09.53-2 1.84-2 3.25C6 11.31 7.55 13 9 13h4c1.45 0 3-1.69 3-3.5S14.5 6 13 6z"></path></svg></a>SINGA installation</h3>
<p>Users can compile and install SINGA from source or install the Python version.
The code can ran on both CPU and GPU. For GPU training, CUDA and CUDNN (V4 or
V5) are required. Please refer to the <a href="/docs/next/installation">installation page</a> for
detailed instructions.</p>
<blockquote>
<p>Please <code>cd</code> to <code>singa/examples/cifar10/</code> for the following commands</p>
</blockquote>
<h3><a class="anchor" aria-hidden="true" id="data-preparation"></a><a href="#data-preparation" aria-hidden="true" class="hash-link"><svg class="hash-link-icon" aria-hidden="true" height="16" version="1.1" viewBox="0 0 16 16" width="16"><path fill-rule="evenodd" d="M4 9h1v1H4c-1.5 0-3-1.69-3-3.5S2.55 3 4 3h4c1.45 0 3 1.69 3 3.5 0 1.41-.91 2.72-2 3.25V8.59c.58-.45 1-1.27 1-2.09C10 5.22 8.98 4 8 4H4c-.98 0-2 1.22-2 2.5S3 9 4 9zm9-3h-1v1h1c1 0 2 1.22 2 2.5S13.98 12 13 12H9c-.98 0-2-1.22-2-2.5 0-.83.42-1.64 1-2.09V6.25c-1.09.53-2 1.84-2 3.25C6 11.31 7.55 13 9 13h4c1.45 0 3-1.69 3-3.5S14.5 6 13 6z"></path></svg></a>Data preparation</h3>
<p>The binary Cifar-10 dataset could be downloaded by</p>
<pre><code class="hljs css language-shell">python download_data.py bin
</code></pre>
<p>The Python version could be downloaded by</p>
<pre><code class="hljs css language-shell">python download_data.py py
</code></pre>
<h3><a class="anchor" aria-hidden="true" id="training"></a><a href="#training" aria-hidden="true" class="hash-link"><svg class="hash-link-icon" aria-hidden="true" height="16" version="1.1" viewBox="0 0 16 16" width="16"><path fill-rule="evenodd" d="M4 9h1v1H4c-1.5 0-3-1.69-3-3.5S2.55 3 4 3h4c1.45 0 3 1.69 3 3.5 0 1.41-.91 2.72-2 3.25V8.59c.58-.45 1-1.27 1-2.09C10 5.22 8.98 4 8 4H4c-.98 0-2 1.22-2 2.5S3 9 4 9zm9-3h-1v1h1c1 0 2 1.22 2 2.5S13.98 12 13 12H9c-.98 0-2-1.22-2-2.5 0-.83.42-1.64 1-2.09V6.25c-1.09.53-2 1.84-2 3.25C6 11.31 7.55 13 9 13h4c1.45 0 3-1.69 3-3.5S14.5 6 13 6z"></path></svg></a>Training</h3>
<p>There are four training programs</p>
<ol>
<li><p><code>train.py</code>. The following command would train the VGG model using the python
version of the Cifar-10 dataset in <code>cifar-10-batches-py</code> folder.
<code>shell python train.py vgg cifar-10-batches-py</code></p>
<pre><code class="hljs">To train other models, please replace 'vgg' to 'alexnet', 'resnet' or 'caffe',
where 'caffe' refers to the alexnet model converted from Caffe. By default
the training would run on a CudaGPU device, to run it on CppCPU, add an additional
argument
```shell
python train.py vgg cifar-10-batches-py  --use_cpu
```
</code></pre></li>
<li><p><code>alexnet.cc</code>. It trains the AlexNet model using the CPP APIs on a CudaGPU,</p>
<pre><code class="hljs css language-shell">./run.sh
</code></pre></li>
<li><p><code>alexnet-parallel.cc</code>. It trains the AlexNet model using the CPP APIs on two
CudaGPU devices. The two devices run synchronously to compute the gradients
of the mode parameters, which are averaged on the host CPU device and then
be applied to update the parameters. <code>shell ./run-parallel.sh</code></p></li>
<li><p><code>vgg-parallel.cc</code>. It trains the VGG model using the CPP APIs on two CudaGPU
devices similar to <code>alexnet-parallel.cc</code>.</p></li>
</ol>
<h3><a class="anchor" aria-hidden="true" id="prediction"></a><a href="#prediction" aria-hidden="true" class="hash-link"><svg class="hash-link-icon" aria-hidden="true" height="16" version="1.1" viewBox="0 0 16 16" width="16"><path fill-rule="evenodd" d="M4 9h1v1H4c-1.5 0-3-1.69-3-3.5S2.55 3 4 3h4c1.45 0 3 1.69 3 3.5 0 1.41-.91 2.72-2 3.25V8.59c.58-.45 1-1.27 1-2.09C10 5.22 8.98 4 8 4H4c-.98 0-2 1.22-2 2.5S3 9 4 9zm9-3h-1v1h1c1 0 2 1.22 2 2.5S13.98 12 13 12H9c-.98 0-2-1.22-2-2.5 0-.83.42-1.64 1-2.09V6.25c-1.09.53-2 1.84-2 3.25C6 11.31 7.55 13 9 13h4c1.45 0 3-1.69 3-3.5S14.5 6 13 6z"></path></svg></a>Prediction</h3>
<p><code>predict.py</code> includes the prediction function</p>
<pre><code class="hljs css language-python"><span class="hljs-function"><span class="hljs-keyword">def</span> <span class="hljs-title">predict</span><span class="hljs-params">(net, images, dev, topk=<span class="hljs-number">5</span>)</span>:</span>
</code></pre>
<p>The net is created by loading the previously trained model; Images consist of a
numpy array of images (one row per image); dev is the training device, e.g., a
CudaGPU device or the host CppCPU device; It returns the topk labels for each
instance.</p>
<p>The predict.py file's main function provides an example of using the pre-trained
alexnet model to do prediction for new images. The 'model.bin' file generated by
the training program should be placed at the cifar10 folder to run</p>
<pre><code class="hljs css language-shell">python predict.py
</code></pre>
</span></div></article></div><div class="docLastUpdate"><em>Last updated on 4/2/2020</em></div><div class="docs-prevnext"><a class="docs-prev button" href="/docs/next/dist-train"><span class="arrow-prev">← </span><span>Distributed Training</span></a><a class="docs-next button" href="/docs/next/model-zoo-char-rnn"><span>Train Char-RNN over plain text</span><span class="arrow-next"> →</span></a></div></div></div><nav class="onPageNav"><ul class="toc-headings"><li><a href="#instructions">Instructions</a><ul class="toc-headings"><li><a href="#singa-installation">SINGA installation</a></li><li><a href="#data-preparation">Data preparation</a></li><li><a href="#training">Training</a></li><li><a href="#prediction">Prediction</a></li></ul></li></ul></nav></div><footer class="nav-footer" id="footer"><section class="sitemap"><a href="/" class="nav-home"><img src="/img/singa-logo-square.png" alt="Apache SINGA" width="66" height="58"/></a><div><h5>Docs</h5><a href="/docs/installation">Getting Started</a><a href="/docs/device">Guides</a><a href="/en/#">API Reference (coming soon)</a><a href="/docs/model-zoo-cnn-cifar10">Model Zoo</a><a href="/docs/download-singa">Development</a></div><div><h5>Community</h5><a href="/en/users.html">User Showcase</a><a href="/docs/history-singa">SINGA History</a><a href="/docs/team-list">SINGA Team</a><a href="/news">SINGA News</a><a href="https://github.com/apache/singa-doc">GitHub</a><div class="social"><a class="github-button" href="https://github.com/apache/singa-doc" data-count-href="/apache/singa/stargazers" data-show-count="true" data-count-aria-label="# stargazers on GitHub" aria-label="Star this project on GitHub">apache/singa-doc</a></div><div class="social"><a href="https://twitter.com/ApacheSINGA" class="twitter-follow-button">Follow @ApacheSINGA</a></div></div><div><h5>Apache Software Foundation</h5><a href="https://apache.org/" target="_blank" rel="noreferrer noopener">Foundation</a><a href="http://www.apache.org/licenses/" target="_blank" rel="noreferrer noopener">License</a><a href="http://www.apache.org/foundation/sponsorship.html" target="_blank" rel="noreferrer noopener">Sponsorship</a><a href="http://www.apache.org/foundation/thanks.html" target="_blank" rel="noreferrer noopener">Thanks</a><a href="http://www.apache.org/events/current-event" target="_blank" rel="noreferrer noopener">Events</a><a href="http://www.apache.org/security/" target="_blank" rel="noreferrer noopener">Security</a></div></section><div style="width:100%;text-align:center"><a href="https://apache.org/" target="_blank" rel="noreferrer noopener" class="ApacheOpenSource"><img src="/img/asf_logo_wide.svg" alt="Apache Open Source"/></a><section class="copyright" style="max-width:60%;margin:0 auto">Copyright © 2020
   The Apache Software Foundation. All rights reserved.
   Apache SINGA, Apache, the Apache feather logo, and
   the Apache SINGA project logos are trademarks of The
   Apache Software Foundation. All other marks mentioned
   may be trademarks or registered trademarks of their
   respective owners.</section></div></footer></div><script>window.twttr=(function(d,s, id){var js,fjs=d.getElementsByTagName(s)[0],t=window.twttr||{};if(d.getElementById(id))return t;js=d.createElement(s);js.id=id;js.src='https://platform.twitter.com/widgets.js';fjs.parentNode.insertBefore(js, fjs);t._e = [];t.ready = function(f) {t._e.push(f);};return t;}(document, 'script', 'twitter-wjs'));</script></body></html>