<!DOCTYPE html><html lang="en"><head><meta charSet="utf-8"/><meta http-equiv="X-UA-Compatible" content="IE=edge"/><title>Software Stack · 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="2.0.0"/><meta name="docsearch:language" content="en"/><meta property="og:title" content="Software Stack · 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>2.0.0</h3></a><div class="navigationWrapper navigationSlider"><nav class="slidingNav"><ul class="nav-site nav-site-internal"><li class="siteNavGroupActive"><a href="/docs/2.0.0/installation" target="_self">Docs</a></li><li class=""><a href="/docs/2.0.0/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="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>Getting Started</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/2.0.0/installation">Installation</a></li><li class="navListItem navListItemActive"><a class="navItem" href="/docs/2.0.0/software-stack">Software Stack</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/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/2.0.0/device">Device</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/tensor">Tensor</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/autograd">Autograd in SINGA</a></li></ul></div><div class="navGroup"><h3 class="navGroupCategoryTitle">Model Zoo</h3><ul class=""><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-cnn-cifar10">Train CNN over Cifar-10</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-char-rnn">Train Char-RNN over plain text</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-rbm-mnist">Train a RBM model against MNIST dataset</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-imagenet-alexnet">Train AlexNet over ImageNet</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-imagenet-densenet">Image Classification using DenseNet</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-imagenet-googlenet">Image Classification using GoogleNet</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-imagenet-inception">Image Classification using Inception V4</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/model-zoo-imagenet-resnet">Image Classification using Residual Networks</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/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/2.0.0/download-singa">Download SINGA</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/build">Build SINGA from Source</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/contribute-code">How to Contribute Code</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/contribute-docs">How to Contribute to Documentation</a></li><li class="navListItem"><a class="navItem" href="/docs/2.0.0/how-to-release">How to Prepare a Release</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/software-stack.md" target="_blank" rel="noreferrer noopener">Edit</a><h1 id="__docusaurus" class="postHeaderTitle">Software Stack</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>SINGA's software stack includes three major components, namely, core, IO and model. Figure 1 illustrates these components together with the hardware. The core component provides memory management and tensor operations; IO has classes for reading (and writing) data from (to) disk and network; The model component provides data structures and algorithms for machine learning models, e.g., layers for neural network models, optimizers/initializer/metric/loss for general machine learning models.</p>
<p><img src="/docs/assets/singav1-sw.png" alt="SINGA V1 software stack"> <br/> <strong>Figure 1 - SINGA V1 software stack.</strong></p>
<h2><a class="anchor" aria-hidden="true" id="core"></a><a href="#core" 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>Core</h2>
<p><a href="/docs/2.0.0/tensor">Tensor</a> and <a href="/docs/2.0.0/device">Device</a> are two core abstractions in SINGA. Tensor class represents a multi-dimensional array, which stores model variables and provides linear algebra operations for machine learning algorithms, including matrix multiplication and random functions. Each tensor instance (i.e. a tensor) is allocated on a Device instance. Each Device instance (i.e. a device) is created against one hardware device, e.g. a GPU card or a CPU core. Devices manage the memory of tensors and execute tensor operations on its execution units, e.g. CPU threads or CUDA streams.</p>
<p>Depending on the hardware and the programming language, SINGA have implemented the following specific device classes:</p>
<ul>
<li><strong>CudaGPU</strong> represents an Nvidia GPU card. The execution units are the CUDA streams.</li>
<li><strong>CppCPU</strong> represents a normal CPU. The execution units are the CPU threads.</li>
<li><strong>OpenclGPU</strong> represents normal GPU card from both Nvidia and AMD. The execution units are the CommandQueues. Given that OpenCL is compatible with many hardware devices, e.g. FPGA and ARM, the OpenclGPU has the potential to be extended for other devices.</li>
</ul>
<p>Different types of devices use different programming languages to write the kernel functions for tensor operations,</p>
<ul>
<li>CppMath (tensor_math_cpp.h) implements the tensor operations using Cpp for CppCPU</li>
<li>CudaMath (tensor_math_cuda.h) implements the tensor operations using CUDA for CudaGPU</li>
<li>OpenclMath (tensor_math_opencl.h) implements the tensor operations using OpenCL for OpenclGPU</li>
</ul>
<p>In addition, different types of data, such as float32 and float16, could be supported by adding the corresponding tensor functions.</p>
<p>Typically, users would create a device instance and pass it to create multiple tensor instances. When users call the Tensor functions, these function would invoke the corresponding implementation (CppMath/CudaMath/OpenclMath) automatically. In other words, the implementation of Tensor operations is transparent to users.</p>
<p>Most machine learning algorithms could be expressed using (dense or sparse) tensors. Therefore, with the Tensor abstraction, SINGA would be able to run a wide range of models, including deep learning models and other traditional machine learning models.</p>
<p>The Tensor and Device abstractions are extensible to support a wide range of hardware device using different programming languages. A new hardware device would be supported by adding a new Device subclass and the corresponding implementation of the Tensor operations (xxxMath).</p>
<p>Optimizations in terms of speed and memory could be implemented by Device, which manages both operation execution and memory malloc/free. More optimization details would be described in the <a href="/docs/2.0.0/device">Device page</a>.</p>
<h2><a class="anchor" aria-hidden="true" id="model"></a><a href="#model" 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>Model</h2>
<p>On top of the Tensor and Device abstractions, SINGA provides some higher level classes for machine learning modules.</p>
<ul>
<li><p><a href="layer.md">Layer</a> and its subclasses are specific for neural networks. Every layer provides functions for forward propagating features and backward propagating gradients w.r.t the training loss functions. They wraps the complex layer operations so that users can easily create neural nets by connecting a set of layers.</p></li>
<li><p><a href="initializer.md">Initializer</a> and its subclasses provide variant methods of initializing model parameters (stored in Tensor instances), following Uniform, Gaussian, etc.</p></li>
<li><p><a href="loss.md">Loss</a> and its subclasses defines the training objective loss functions. Both functions of computing the loss values and computing the gradient of the prediction w.r.t the objective loss are implemented. Example loss functions include squared error and cross entropy.</p></li>
<li><p><a href="metric.md">Metric</a> and its subclasses provide the function to measure the performance of the model, e.g., the accuracy.</p></li>
<li><p><a href="optimizer.md">Optimizer</a> and its subclasses implement the methods for updating model parameter values using parameter gradients, including SGD, AdaGrad, RMSProp etc.</p></li>
</ul>
<h2><a class="anchor" aria-hidden="true" id="io"></a><a href="#io" 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>IO</h2>
<p>The IO module consists of classes for data loading, data preprocessing and message passing.</p>
<ul>
<li>Reader and its subclasses load string records from disk files</li>
<li>Writer and its subclasses write string records to disk files</li>
<li>Encoder and its subclasses encode Tensor instances into string records</li>
<li>Decoder and its subclasses decodes string records into Tensor instances</li>
<li>Endpoint represents a communication endpoint which provides functions for passing messages to each other.</li>
<li>Message represents communication messages between Endpoint instances. It carries both meta data and payload.</li>
</ul>
</span></div></article></div><div class="docs-prevnext"><a class="docs-prev button" href="/docs/2.0.0/installation"><span class="arrow-prev">← </span><span>Installation</span></a><a class="docs-next button" href="/docs/2.0.0/benchmark-train"><span>Benchmark for Distributed Training</span><span class="arrow-next"> →</span></a></div></div></div><nav class="onPageNav"><ul class="toc-headings"><li><a href="#core">Core</a></li><li><a href="#model">Model</a></li><li><a href="#io">IO</a></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>