content/docs/model-zoo-cnn-cifar10/index.html - singa-site - Git at Google

 <!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="3.0.0.rc1"/><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>3.0.0.rc1</h3></a><div class="navigationWrapper navigationSlider"><nav class="slidingNav"><ul class="nav-site nav-site-internal"><li class=""><a href="/docs/installation" target="_self">Docs</a></li><li class=""><a href="/docs/source-repository" target="_self">Community</a></li><li class=""><a href="/blog/" target="_self">News</a></li><li class=""><a href="https://apache-singa.readthedocs.io/en/latest/" target="_self">API</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="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/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="docs-prevnext"></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>
	<!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="<!--- 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. -->"/><meta name="docsearch:version" content="3.0.0.rc1"/><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="<!--- 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. -->"/><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>3.0.0.rc1</h3></a><div class="navigationWrapper navigationSlider"><nav class="slidingNav"><ul class="nav-site nav-site-internal"><li class=""><a href="/docs/installation" target="_self">Docs</a></li><li class=""><a href="/docs/source-repository" target="_self">Community</a></li><li class=""><a href="/blog/" target="_self">News</a></li><li class=""><a href="https://apache-singa.readthedocs.io/en/latest/" target="_self">API</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="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/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="docs-prevnext"></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>