gradient.html

<html>
<head>
  <meta charset="utf-8" />
  <meta http-equiv="X-UA-Compatible" content="IE=edge">
  <meta name="viewport" content="width=device-width, initial-scale=1">
  
  <title>TensorFlow for .NET</title>
  <meta name="description" content="Full TensorFlow binding for .NET by Lost Tech">
  <meta name="keywords" content="TensorFlow,.NET,dotnet,C#,F#,VB.NET">

  <meta name="author" content="Victor Milovanov">

  <link rel="icon" href="img/LostTech.ico"/>

  <!-- Bootstrap Core CSS -->
  <link href="css/bootstrap.min.css" rel="stylesheet">
  <!-- Clean CSS -->
  <link href="css/clean-blog.min.css" rel="stylesheet">

  <link href="css/diskache.css" rel="stylesheet" />

  <!-- Custom Fonts -->
  <link href="//maxcdn.bootstrapcdn.com/font-awesome/4.1.0/css/font-awesome.min.css" rel="stylesheet" type="text/css">
  <link href='//fonts.googleapis.com/css?family=Lora:400,700,400italic,700italic' rel='stylesheet' type='text/css'>
  <link href='//fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800' rel='stylesheet' type='text/css'>

  <!-- jQuery -->
  <script src="js/jquery.js"></script>
  <!-- Bootstrap Core JavaScript -->
  <script src="js/bootstrap.min.js"></script>
  <!-- Custom Theme JavaScript -->
  <script src="js/clean-blog.min.js"></script>

  <script src="js/diskache.js"></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','//www.google-analytics.com/analytics.js','ga');

  ga('create', 'UA-107799728-1', 'auto');
  ga('send', 'pageview');

  </script>
</head>
<body>
<!-- Navigation -->
<nav class="navbar navbar-default navbar-custom navbar-fixed-top">
  <div class="container-fluid">
    <!-- Brand and toggle get grouped for better mobile display -->
    <div class="navbar-header page-scroll">
      <button type="button" class="navbar-toggle" data-toggle="collapse" data-target="#bs-example-navbar-collapse-1">
        <span class="sr-only">Toggle navigation</span>
        <span class="icon-bar"></span>
        <span class="icon-bar"></span>
        <span class="icon-bar"></span>
      </button>
      
    </div>
    <!-- Collect the nav links, forms, and other content for toggling -->
    <div class="collapse navbar-collapse" id="bs-example-navbar-collapse-1">
      <ul class="nav navbar-nav navbar-right">
        <li>
          <a href="gradient.html" title="Gradient - full TensorFlow binding for C#">Gradient</a>
        </li>
        <li>
          <a href="https://stackwm.github.io/" title="Stack - customizable tiling window manager for Windows">Stack</a>
        </li>
        <li>
          <a href="https://ml.blogs.losttech.software/" title="Our blog on deep learning: C# examples, Gradient news, etc">ML Blog</a>
        </li>
        <li>
          <a href="about.html">About</a>
        </li>
        <li>
          <a href="contact.html">Contact</a>
        </li>
      </ul>
    </div>
    <!-- /.navbar-collapse -->
  </div>
  <!-- /.container -->
</nav>


<!-- Page Header -->
<!-- Set your background image for this header on the line below. -->
<header class="intro-header" style="background-image: linear-gradient(rgba(0, 0, 0, 0.25), rgba(0, 0, 0, 0.7), rgba(0, 0, 0, 0.25)), url('img/gradient-post-bg.jpg');">
  <div class="container">
    <div class="row">
      <div class="col-lg-8 col-lg-offset-2 col-md-10 col-md-offset-1">
        <div class="post-heading">
          <h1>TensorFlow for .NET</h1>
          <div class="meta">train neural networks with TensorFlow in C#</div>
          <div class="meta padding">&nbsp;</div>
          <div class="meta padding">&nbsp;</div>
          <div class="meta">
            <a href="https://losttech.software/buy_gradient.html" class="btn btn-primary">Get Now</a>
            <p>
              <small><a href="https://ml.blogs.losttech.software/TF-1.15-Release-Candidate/">See What's New!</a></small>
            </p>
          </div>
        </div>
      </div>
    </div>
  </div>
</header>

<!-- Main Content -->
<article>
  <div class="container">
    <div class="row">
      <div class="col-lg-8 col-lg-offset-2 col-md-10 col-md-offset-1">
        <p>
          TensorFlow for .NET by Lost Tech allows you to create, train, and use machine learning
          models with the full power of TensorFlow API on C#, F# or any other .NET language.
        </p>

        <pre style="font-family:Consolas;font-size:13px;color:black;background:white;"><span style="color:blue;">var</span>&nbsp;input&nbsp;=&nbsp;<span style="color:#2b91af;">tf</span>.placeholder(<span style="color:#2b91af;">tf</span>.float32,&nbsp;<span style="color:blue;">new</span>&nbsp;<span style="color:#2b91af;">TensorShape</span>(<span style="color:blue;">null</span>,&nbsp;1),&nbsp;name:&nbsp;<span style="color:#a31515;">&quot;x&quot;</span>);
<span style="color:blue;">var</span>&nbsp;output&nbsp;=&nbsp;<span style="color:#2b91af;">tf</span>.placeholder(<span style="color:#2b91af;">tf</span>.float32,&nbsp;<span style="color:blue;">new</span>&nbsp;<span style="color:#2b91af;">TensorShape</span>(<span style="color:blue;">null</span>,&nbsp;1),&nbsp;name:&nbsp;<span style="color:#a31515;">&quot;y&quot;</span>);
 
<span style="color:blue;">var</span>&nbsp;hiddenLayer&nbsp;=&nbsp;<span style="color:#2b91af;">tf</span>.<span style="color:#2b91af;">layers</span>.dense(input,&nbsp;hiddenSize,
&nbsp;&nbsp;&nbsp;&nbsp;activation:&nbsp;<span style="color:#2b91af;">tf</span>.sigmoid_fn,
&nbsp;&nbsp;&nbsp;&nbsp;kernel_initializer:&nbsp;<span style="color:blue;">new</span>&nbsp;<span style="color:#2b91af;">ones_initializer</span>(),
&nbsp;&nbsp;&nbsp;&nbsp;bias_initializer:&nbsp;<span style="color:blue;">new</span>&nbsp;<span style="color:#2b91af;">random_uniform_initializer</span>(minval:&nbsp;-x1,&nbsp;maxval:&nbsp;-x0),
&nbsp;&nbsp;&nbsp;&nbsp;name:&nbsp;<span style="color:#a31515;">&quot;hidden&quot;</span>);
 
<span style="color:blue;">var</span>&nbsp;model&nbsp;=&nbsp;<span style="color:#2b91af;">tf</span>.<span style="color:#2b91af;">layers</span>.dense(hiddenLayer,&nbsp;units:&nbsp;1,&nbsp;name:&nbsp;<span style="color:#a31515;">&quot;output&quot;</span>);
 
<span style="color:blue;">var</span>&nbsp;cost&nbsp;=&nbsp;<span style="color:#2b91af;">tf</span>.<span style="color:#2b91af;">losses</span>.mean_squared_error(output,&nbsp;model);
 
<span style="color:blue;">var</span>&nbsp;training&nbsp;=&nbsp;<span style="color:blue;">new</span>&nbsp;<span style="color:#2b91af;">GradientDescentOptimizer</span>(learning_rate:&nbsp;learningRate).minimize(cost);</pre>
        <span class="caption text-muted">Code sample.</span>
        <hr />
        <a href="https://ml.blogs.losttech.software/" class="btn btn-primary btn-block">Read Our ML Blog</a>
        <hr />
        <h3>Features</h3>
        <hr />
        <ul>
          <li>
            <h4>Access the full set of <a href="https://www.tensorflow.org/versions/r1.15/api_docs/python/tf">TensorFlow APIs</a></h4>
            <ul>
              <li>Build computation graphs, and run them in sessions</li>
              <li>Use Keras-style <a href="https://medium.com/tensorflow/standardizing-on-keras-guidance-on-high-level-apis-in-tensorflow-2-0-bad2b04c819a">high-level APIs</a></li>
              <li>Build fast data pipelines, keep logs and model checkpoints</li>
              <li>Use estimators and the full power of tf.contrib</li>
              <li>Use eager mode to transform data interactively</li>
              <li>Many more</li>
            </ul>
          </li>
          <li><h4>Train and run models on any hardware platform: CPUs, GPUs, TPUs</h4></li>
          <li><h4>Use distributed training features</h4></li>
          <li>
            <h4>Track your training progress with <a href="https://github.com/tensorflow/tensorboard/blob/master/docs/r1/summaries.md">TensorBoard</a></h4>
            <p>
              <img class="img-responsive" src="https://www.tensorflow.org/images/mnist_tensorboard.png" title="TensorBoard screenshot" />
            </p>
          </li>
          <li>
            <h4>Easily port numerous existing TensorFlow examples</h4>
            <ul>
              <li>
                From simple numerical computation samples to state-of-art models like
                <a href="https://en.wikipedia.org/wiki/AlphaZero">AlphaZero</a> - the new world's Go champion by DeepMind.
              </li>
              <li>
                Our <a href="https://habr.com/post/453232/">GPT-2 demo</a> was built in one week just a month
                after OpenAI released their first model, and included both training and infrerence,
                demonstrating <strong>incredible time-to-market with LostTech.TensorFlow</strong>
              </li>
            </ul>
          </li>
          <li>
            <h4>Get started quickly with a collection of <a href="https://github.com/losttech/Gradient-Samples">samples</a></h4>
          </li>
          <li>
            <h4>Seek help with the <a href="https://stackoverflow.com/questions/tagged/tensorflow">growing community</a></h4>
          </li>
          <li>
            <h4>Use C# for machine learning</h4>
            <ul>
              <li>Static typing when possible, fallback to dynamic in corner cases</li>
              <li>IDE support: code completion, documentation hints for classes, functions, and parameters</li>
              <li>Support for C# 8.0 features, such as ranges</li>
              <li>Can be used from C# interactive, and C# <a href="https://github.com/zabirauf/icsharp">kernel for Jupyter</a></li>
            </ul>
          </li>
          <li>
            <h4>
              LostTech.TensorFlow is about 18% faster than TensorFlow.NET
              on a <a href="https://github.com/losttech/Gradient-Perf">CNN training benchmark</a>
            </h4>
            <p>
              <img class="img-responsive" src="https://ml.blogs.losttech.software/images/perf-vs-tf.net.png" title="performance comparison" />
            </p>
          </li>
        </ul>
        <p></p>
        <hr />
        <h3>Comparison with <a href="https://github.com/migueldeicaza/TensorFlowSharp">TensorFlowSharp</a> </h3>
        <hr />
        <table class="feature-comparison" style="width: 100%">
          <thead>
            <tr>
              <td></td>
              <td style="width: 30%">TensorFlow.NET</td>
              <td style="width: 30%">LostTech.TensorFlow</td>
            </tr>
          </thead>
          <tbody>
            <tr><td>Load TensorFlow models</td><td class="win">✓</td><td class="win">✓</td></tr>
            <tr><td>Train existing models</td><td class="win">✓</td><td class="win">✓</td></tr>
            <tr><td>Create new models with low-level API</td><td class="win">✓</td><td class="win">✓</td></tr>
            <tr><td>Create new models with high-level API</td><td class="loose">✗</td><td class="win">✓</td></tr>
            <tr><td>Latest algorithms</td><td class="loose">Some</td><td class="win">Most</td></tr>
            <tr><td>Dependencies</td><td class="win">TF</td><td class="loose">TF&nbsp;+ Python</td></tr>
            <tr><td>TensorBoard integration</td><td class="loose">✗</td><td class="win">✓</td></tr>
            <tr><td>Estimators</td><td class="loose">✗</td><td class="win">✓</td></tr>
            <tr><td>Dataset manipulation via tf.data</td><td class="loose">✗</td><td class="win">✓</td></tr>
            <tr><td>tf.contrib</td><td class="loose">✗</td><td class="win" title="Available, but not supported">✓*</td></tr>
            <tr><td>Commercial support</td><td class="loose">✗</td><td class="win">✓</td></tr>
            <tr><td>Performance</td><td class="loose">1x</td><td class="win">1.18x</td></tr>
            <tr><td>TPU support</td><td class="loose">✗</td><td class="win">✓</td></tr>
            <tr><td>AMD GPU support</td><td class="loose">✗</td><td class="win">✓</td></tr>
          </tbody>
        </table>
        <p></p>
        <hr />
        <h3>Documentation &amp; Tutorials</h3>
        <hr />
        <ul>
          <li><a href="https://ml.blogs.losttech.software/">Our Machine Learning Blog</a>: cool samples, LostTech.TensorFlow news, etc</li>
          <li>See <a href="https://ml.blogs.losttech.software/TF-1.15-Release-Candidate/">What's New</a> in the latest version</li>
          <li><a href="https://github.com/losttech/Gradient/#contents">Getting started</a></li>
          <li><a href="https://ml.blogs.losttech.software/Reinforcement-Learning-With-Unity-ML-Agents/">Reinforcement learning</a> with Unity ML Agents</li>
          <li><a href="https://habr.com/post/453232/">Writing billion songs with C# and Deep Learning</a> + <a href="https://billionsongs.azurewebsites.net/">Demo</a></li>
          <li><a href="https://lostmsu.github.io/Not-CSharp/">C# or NOT</a>: train deep convolutional network to classify programming language from a code fragment</li>
          <li><a href="https://habr.com/post/437174/">.NET, TensorFlow, and the windmills of Kaggle</a></li>
        </ul>
        <hr />
        <h4>
          <a href="https://losttech.software/buy_gradient.html">Get it now!</a>
        </h4>
        <hr />
      </div>
    </div>
  </div>
</article>

<p class="copyright text-muted photo-credit">
  Photo credit: <a href="https://www.pexels.com/@florian-weihmann-459786">Florian Weihmann from Pexels</a>
  <br />
  TensorFlow, the TensorFlow logo and any related marks are trademarks of Google Inc.
</p>


<hr />

<!-- Footer -->
<footer>
  <div class="container">
    <div class="row">
      <div class="col-lg-8 col-lg-offset-2 col-md-10 col-md-offset-1">
        <ul class="list-inline text-center">
          <li>
            <a href="https://www.facebook.com/losttechllc">
              <span class="fa-stack fa-lg">
                <i class="fa fa-circle fa-stack-2x"></i>
                <i class="fa fa-facebook fa-stack-1x fa-inverse"></i>
              </span>
            </a>
          </li>
          <li>
            <span class="fa-stack fa-lg">
              <i class="fa fa-circle fa-stack-2x"></i>
              <i class="fa fa-windows fa-stack-1x fa-inverse"></i>
            </span>
          </li>
        </ul>
        <p class="copyright text-muted">Copyright &copy; <a href="about.html">Lost Tech LLC</a> 2015-2021</p>
        <p class="copyright text-muted photo-credit">Site design: <a href="http://www.ironsummitmedia.com/">Iron Summit Media Strategies, LLC</a></p>
      </div>
    </div>
  </div>
</footer>
</body>

<script src="js/instant.page.js"></script>
</html>