documentation/data_import.xhtml - platform/external/ComputeLibrary - Gitiles

 <!-- HTML header for doxygen 1.8.15-->
 <!-- Remember to use version doxygen 1.8.15 +-->
 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
 <html xmlns="http://www.w3.org/1999/xhtml">
 <head>
 <meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
 <meta http-equiv="X-UA-Compatible" content="IE=9"/>
 <meta name="generator" content="Doxygen 1.8.15"/>
 <meta name="robots" content="NOINDEX, NOFOLLOW" /> <!-- Prevent indexing by search engines -->
 <title>Compute Library: Importing data from existing models</title>
 <link href="tabs.css" rel="stylesheet" type="text/css"/>
 <script type="text/javascript" src="jquery.js"></script>
 <script type="text/javascript" src="dynsections.js"></script>
 <link href="navtree.css" rel="stylesheet" type="text/css"/>
 <script type="text/javascript" src="resize.js"></script>
 <script type="text/javascript" src="navtreedata.js"></script>
 <script type="text/javascript" src="navtree.js"></script>
 <script type="text/javascript">
 /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
   $(document).ready(initResizable);
 /* @license-end */</script>
 <link href="search/search.css" rel="stylesheet" type="text/css"/>
 <script type="text/javascript" src="search/searchdata.js"></script>
 <script type="text/javascript" src="search/search.js"></script>
 <script type="text/x-mathjax-config">
   MathJax.Hub.Config({
     extensions: ["tex2jax.js"],
     jax: ["input/TeX","output/HTML-CSS"],
 });
 </script><script type="text/javascript" async="async" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></script>
 <link href="doxygen.css" rel="stylesheet" type="text/css" />
 <link href="stylesheet.css" rel="stylesheet" type="text/css"/>
 </head>
 <body>
 <div id="top"><!-- do not remove this div, it is closed by doxygen! -->
 <div id="titlearea">
 <table cellspacing="0" cellpadding="0">
  <tbody>
  <tr style="height: 56px;">
   <img alt="Compute Library" src="https://raw.githubusercontent.com/ARM-software/ComputeLibrary/gh-pages/ACL_logo.png" style="max-width: 100%;margin-top: 15px;margin-left: 10px"/>
   <td style="padding-left: 0.5em;">
    <div id="projectname">
    &#160;<span id="projectnumber">19.11</span>
    </div>
   </td>
  </tr>
  </tbody>
 </table>
 </div>
 <!-- end header part -->
 <!-- Generated by Doxygen 1.8.15 -->
 <script type="text/javascript">
 /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
 var searchBox = new SearchBox("searchBox", "search",false,'Search');
 /* @license-end */
 </script>
 <script type="text/javascript" src="menudata.js"></script>
 <script type="text/javascript" src="menu.js"></script>
 <script type="text/javascript">
 /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
 $(function() {
   initMenu('',true,false,'search.php','Search');
   $(document).ready(function() { init_search(); });
 });
 /* @license-end */</script>
 <div id="main-nav"></div>
 </div><!-- top -->
 <div id="side-nav" class="ui-resizable side-nav-resizable">
   <div id="nav-tree">
     <div id="nav-tree-contents">
       <div id="nav-sync" class="sync"></div>
     </div>
   </div>
   <div id="splitbar" style="-moz-user-select:none;"
        class="ui-resizable-handle">
   </div>
 </div>
 <script type="text/javascript">
 /* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
 $(document).ready(function(){initNavTree('data_import.xhtml','');});
 /* @license-end */
 </script>
 <div id="doc-content">
 <!-- window showing the filter options -->
 <div id="MSearchSelectWindow"
      onmouseover="return searchBox.OnSearchSelectShow()"
      onmouseout="return searchBox.OnSearchSelectHide()"
      onkeydown="return searchBox.OnSearchSelectKey(event)">
 </div>

 <!-- iframe showing the search results (closed by default) -->
 <div id="MSearchResultsWindow">
 <iframe src="javascript:void(0)" frameborder="0"
         name="MSearchResults" id="MSearchResults">
 </iframe>
 </div>

 <div class="PageDoc"><div class="header">
   <div class="headertitle">
 <div class="title">Importing data from existing models </div>  </div>
 </div><!--header-->
 <div class="contents">
 <div class="toc"><h3>Table of Contents</h3>
 <ul><li class="level1"><a href="#caffe_data_extractor">Extract data from pre-trained caffe model</a><ul><li class="level2"><a href="#caffe_how_to">How to use the script</a></li>
 <li class="level2"><a href="#caffe_result">What is the expected output from the script</a></li>
 </ul>
 </li>
 <li class="level1"><a href="#tensorflow_data_extractor">Extract data from pre-trained tensorflow model</a><ul><li class="level2"><a href="#tensorflow_how_to">How to use the script</a></li>
 <li class="level2"><a href="#tensorflow_result">What is the expected output from the script</a></li>
 </ul>
 </li>
 <li class="level1"><a href="#tf_frozen_model_extractor">Extract data from pre-trained frozen tensorflow model</a><ul><li class="level2"><a href="#tensorflow_frozen_how_to">How to use the script</a></li>
 <li class="level2"><a href="#tensorflow_frozen_result">What is the expected output from the script</a></li>
 </ul>
 </li>
 <li class="level1"><a href="#validate_examples">Validating examples</a></li>
 </ul>
 </div>
 <div class="textblock"><h1><a class="anchor" id="caffe_data_extractor"></a>
 Extract data from pre-trained caffe model</h1>
 <p>One can find caffe <a href="https://github.com/BVLC/caffe/wiki/Model-Zoo">pre-trained models</a> on caffe's official github repository.</p>
 <p>The caffe_data_extractor.py provided in the scripts folder is an example script that shows how to extract the parameter values from a trained model.</p>
 <dl class="section note"><dt>Note</dt><dd>complex networks might require altering the script to properly work.</dd></dl>
 <h2><a class="anchor" id="caffe_how_to"></a>
 How to use the script</h2>
 <p>Install caffe following <a href="http://caffe.berkeleyvision.org/installation.html">caffe's document</a>. Make sure the pycaffe has been added into the PYTHONPATH.</p>
 <p>Download the pre-trained caffe model.</p>
 <p>Run the caffe_data_extractor.py script by </p><pre class="fragment">    python caffe_data_extractor.py -m &lt;caffe model&gt; -n &lt;caffe netlist&gt;
 </pre><p>For example, to extract the data from pre-trained caffe Alex model to binary file: </p><pre class="fragment">    python caffe_data_extractor.py -m /path/to/bvlc_alexnet.caffemodel -n /path/to/caffe/models/bvlc_alexnet/deploy.prototxt
 </pre><p>The script has been tested under Python2.7.</p>
 <h2><a class="anchor" id="caffe_result"></a>
 What is the expected output from the script</h2>
 <p>If the script runs successfully, it prints the names and shapes of each layer onto the standard output and generates *.npy files containing the weights and biases of each layer.</p>
 <p>The <a class="el" href="namespacearm__compute_1_1utils.xhtml#af214346f90d640ac468dd90fa2a275cc" title="Load the tensor with pre-trained data from a binary file.">arm_compute::utils::load_trained_data</a> shows how one could load the weights and biases into tensor from the .npy file by the help of Accessor.</p>
 <h1><a class="anchor" id="tensorflow_data_extractor"></a>
 Extract data from pre-trained tensorflow model</h1>
 <p>The script tensorflow_data_extractor.py extracts trainable parameters (e.g. values of weights and biases) from a trained tensorflow model. A tensorflow model consists of the following two files:</p>
 <p>{model_name}.data-{step}-{global_step}: A binary file containing values of each variable.</p>
 <p>{model_name}.meta: A binary file containing a MetaGraph struct which defines the graph structure of the neural network.</p>
 <dl class="section note"><dt>Note</dt><dd>Since Tensorflow version 0.11 the binary checkpoint file which contains the values for each parameter has the format of: {model_name}.data-{step}-of-{max_step} instead of: {model_name}.ckpt When dealing with binary files with version &gt;= 0.11, only pass {model_name} to -m option; when dealing with binary files with version &lt; 0.11, pass the whole file name {model_name}.ckpt to -m option.</dd>
 <dd>
 This script relies on the parameters to be extracted being in the 'trainable_variables' tensor collection. By default all variables are automatically added to this collection unless specified otherwise by the user. Thus should a user alter this default behavior and/or want to extract parameters from other collections, tf.GraphKeys.TRAINABLE_VARIABLES should be replaced accordingly.</dd></dl>
 <h2><a class="anchor" id="tensorflow_how_to"></a>
 How to use the script</h2>
 <p>Install tensorflow and numpy.</p>
 <p>Download the pre-trained tensorflow model.</p>
 <p>Run tensorflow_data_extractor.py with </p><pre class="fragment">    python tensorflow_data_extractor -m &lt;path_to_binary_checkpoint_file&gt; -n &lt;path_to_metagraph_file&gt;
 </pre><p>For example, to extract the data from pre-trained tensorflow Alex model to binary files: </p><pre class="fragment">    python tensorflow_data_extractor -m /path/to/bvlc_alexnet -n /path/to/bvlc_alexnet.meta
 </pre><p>Or for binary checkpoint files before Tensorflow 0.11: </p><pre class="fragment">    python tensorflow_data_extractor -m /path/to/bvlc_alexnet.ckpt -n /path/to/bvlc_alexnet.meta
 </pre><dl class="section note"><dt>Note</dt><dd>with versions &gt;= Tensorflow 0.11 only model name is passed to the -m option</dd></dl>
 <p>The script has been tested with Tensorflow 1.2, 1.3 on Python 2.7.6 and Python 3.4.3.</p>
 <h2><a class="anchor" id="tensorflow_result"></a>
 What is the expected output from the script</h2>
 <p>If the script runs successfully, it prints the names and shapes of each parameter onto the standard output and generates .npy files containing the weights and biases of each layer.</p>
 <p>The <a class="el" href="namespacearm__compute_1_1utils.xhtml#af214346f90d640ac468dd90fa2a275cc" title="Load the tensor with pre-trained data from a binary file.">arm_compute::utils::load_trained_data</a> shows how one could load the weights and biases into tensor from the .npy file by the help of Accessor.</p>
 <h1><a class="anchor" id="tf_frozen_model_extractor"></a>
 Extract data from pre-trained frozen tensorflow model</h1>
 <p>The script tf_frozen_model_extractor.py extracts trainable parameters (e.g. values of weights and biases) from a frozen trained Tensorflow model.</p>
 <h2><a class="anchor" id="tensorflow_frozen_how_to"></a>
 How to use the script</h2>
 <p>Install Tensorflow and NumPy.</p>
 <p>Download the pre-trained Tensorflow model and freeze the model using the architecture and the checkpoint file.</p>
 <p>Run tf_frozen_model_extractor.py with </p><pre class="fragment">    python tf_frozen_model_extractor -m &lt;path_to_frozen_pb_model_file&gt; -d &lt;path_to_store_parameters&gt;
 </pre><p>For example, to extract the data from pre-trained Tensorflow model to binary files: </p><pre class="fragment">    python tf_frozen_model_extractor -m /path/to/inceptionv3.pb -d ./data
 </pre><h2><a class="anchor" id="tensorflow_frozen_result"></a>
 What is the expected output from the script</h2>
 <p>If the script runs successfully, it prints the names and shapes of each parameter onto the standard output and generates .npy files containing the weights and biases of each layer.</p>
 <p>The <a class="el" href="namespacearm__compute_1_1utils.xhtml#af214346f90d640ac468dd90fa2a275cc" title="Load the tensor with pre-trained data from a binary file.">arm_compute::utils::load_trained_data</a> shows how one could load the weights and biases into tensor from the .npy file by the help of Accessor.</p>
 <h1><a class="anchor" id="validate_examples"></a>
 Validating examples</h1>
 <p>Using one of the provided scripts will generate files containing the trainable parameters.</p>
 <p>You can validate a given graph example on a list of inputs by running: </p><pre class="fragment">LD_LIBRARY_PATH=lib ./&lt;graph_example&gt; --validation-range='&lt;validation_range&gt;' --validation-file='&lt;validation_file&gt;' --validation-path='/path/to/test/images/' --data='/path/to/weights/'
 </pre><p>e.g:</p>
 <p>LD_LIBRARY_PATH=lib ./bin/graph_alexnet &ndash;target=CL &ndash;layout=NHWC &ndash;type=F32 &ndash;threads=4 &ndash;validation-range='16666,24998' &ndash;validation-file='val.txt' &ndash;validation-path='images/' &ndash;data='data/'</p>
 <p>where: validation file is a plain document containing a list of images along with their expected label value. e.g: </p><pre class="fragment">val_00000001.JPEG 65
 val_00000002.JPEG 970
 val_00000003.JPEG 230
 val_00000004.JPEG 809
 val_00000005.JPEG 516
 </pre><p>&ndash;validation-range is the index range of the images within the validation file you want to check: e.g:</p>
 <p>&ndash;validation-range='100,200' will validate 100 images starting from 100th one in the validation file.</p>
 <p>This can be useful when parallelizing the validation process is needed. </p>
 </div></div><!-- PageDoc -->
 </div><!-- contents -->
 </div><!-- doc-content -->
 <!-- start footer part -->
 <div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
   <ul>
     <li class="footer">Generated on Thu Nov 28 2019 16:53:08 for Compute Library by
     <a href="http://www.doxygen.org/index.html">
     <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.15 </li>
   </ul>
 </div>
 </body>
 </html>
	<!-- HTML header for doxygen 1.8.15-->
	<!-- Remember to use version doxygen 1.8.15 +-->
	<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
	<html xmlns="http://www.w3.org/1999/xhtml">
	<head>
	<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
	<meta http-equiv="X-UA-Compatible" content="IE=9"/>
	<meta name="generator" content="Doxygen 1.8.15"/>
	<meta name="robots" content="NOINDEX, NOFOLLOW" /> <!-- Prevent indexing by search engines -->
	<title>Compute Library: Importing data from existing models</title>
	<link href="tabs.css" rel="stylesheet" type="text/css"/>
	<script type="text/javascript" src="jquery.js"></script>
	<script type="text/javascript" src="dynsections.js"></script>
	<link href="navtree.css" rel="stylesheet" type="text/css"/>
	<script type="text/javascript" src="resize.js"></script>
	<script type="text/javascript" src="navtreedata.js"></script>
	<script type="text/javascript" src="navtree.js"></script>
	<script type="text/javascript">
	/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */
	$(document).ready(initResizable);
	/* @license-end */</script>
	<link href="search/search.css" rel="stylesheet" type="text/css"/>
	<script type="text/javascript" src="search/searchdata.js"></script>
	<script type="text/javascript" src="search/search.js"></script>
	<script type="text/x-mathjax-config">
	MathJax.Hub.Config({
	extensions: ["tex2jax.js"],
	jax: ["input/TeX","output/HTML-CSS"],
	});
	</script><script type="text/javascript" async="async" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></script>
	<link href="doxygen.css" rel="stylesheet" type="text/css" />
	<link href="stylesheet.css" rel="stylesheet" type="text/css"/>
	</head>
	<body>
	<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
	<div id="titlearea">
	<table cellspacing="0" cellpadding="0">
	<tbody>
	<tr style="height: 56px;">
	<img alt="Compute Library" src="https://raw.githubusercontent.com/ARM-software/ComputeLibrary/gh-pages/ACL_logo.png" style="max-width: 100%;margin-top: 15px;margin-left: 10px"/>
	<td style="padding-left: 0.5em;">
	<div id="projectname">
	<span id="projectnumber">19.11</span>
	</div>
	</td>
	</tr>
	</tbody>
	</table>
	</div>
	<!-- end header part -->
	<!-- Generated by Doxygen 1.8.15 -->
	<script type="text/javascript">
	/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */
	var searchBox = new SearchBox("searchBox", "search",false,'Search');
	/* @license-end */
	</script>
	<script type="text/javascript" src="menudata.js"></script>
	<script type="text/javascript" src="menu.js"></script>
	<script type="text/javascript">
	/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */
	$(function() {
	initMenu('',true,false,'search.php','Search');
	$(document).ready(function() { init_search(); });
	});
	/* @license-end */</script>
	<div id="main-nav"></div>
	</div><!-- top -->
	<div id="side-nav" class="ui-resizable side-nav-resizable">
	<div id="nav-tree">
	<div id="nav-tree-contents">
	<div id="nav-sync" class="sync"></div>
	</div>
	</div>
	<div id="splitbar" style="-moz-user-select:none;"
	class="ui-resizable-handle">
	</div>
	</div>
	<script type="text/javascript">
	/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&dn=gpl-2.0.txt GPL-v2 */
	$(document).ready(function(){initNavTree('data_import.xhtml','');});
	/* @license-end */
	</script>
	<div id="doc-content">
	<!-- window showing the filter options -->
	<div id="MSearchSelectWindow"
	onmouseover="return searchBox.OnSearchSelectShow()"
	onmouseout="return searchBox.OnSearchSelectHide()"
	onkeydown="return searchBox.OnSearchSelectKey(event)">
	</div>

	<!-- iframe showing the search results (closed by default) -->
	<div id="MSearchResultsWindow">
	<iframe src="javascript:void(0)" frameborder="0"
	name="MSearchResults" id="MSearchResults">
	</iframe>
	</div>

	<div class="PageDoc"><div class="header">
	<div class="headertitle">
	<div class="title">Importing data from existing models </div> </div>
	</div><!--header-->
	<div class="contents">
	<div class="toc"><h3>Table of Contents</h3>
	<ul><li class="level1"><a href="#caffe_data_extractor">Extract data from pre-trained caffe model</a><ul><li class="level2"><a href="#caffe_how_to">How to use the script</a></li>
	<li class="level2"><a href="#caffe_result">What is the expected output from the script</a></li>
	</ul>
	</li>
	<li class="level1"><a href="#tensorflow_data_extractor">Extract data from pre-trained tensorflow model</a><ul><li class="level2"><a href="#tensorflow_how_to">How to use the script</a></li>
	<li class="level2"><a href="#tensorflow_result">What is the expected output from the script</a></li>
	</ul>
	</li>
	<li class="level1"><a href="#tf_frozen_model_extractor">Extract data from pre-trained frozen tensorflow model</a><ul><li class="level2"><a href="#tensorflow_frozen_how_to">How to use the script</a></li>
	<li class="level2"><a href="#tensorflow_frozen_result">What is the expected output from the script</a></li>
	</ul>
	</li>
	<li class="level1"><a href="#validate_examples">Validating examples</a></li>
	</ul>
	</div>
	<div class="textblock"><h1><a class="anchor" id="caffe_data_extractor"></a>
	Extract data from pre-trained caffe model</h1>
	<p>One can find caffe <a href="https://github.com/BVLC/caffe/wiki/Model-Zoo">pre-trained models</a> on caffe's official github repository.</p>
	<p>The caffe_data_extractor.py provided in the scripts folder is an example script that shows how to extract the parameter values from a trained model.</p>
	<dl class="section note"><dt>Note</dt><dd>complex networks might require altering the script to properly work.</dd></dl>
	<h2><a class="anchor" id="caffe_how_to"></a>
	How to use the script</h2>
	<p>Install caffe following <a href="http://caffe.berkeleyvision.org/installation.html">caffe's document</a>. Make sure the pycaffe has been added into the PYTHONPATH.</p>
	<p>Download the pre-trained caffe model.</p>
	<p>Run the caffe_data_extractor.py script by </p><pre class="fragment"> python caffe_data_extractor.py -m <caffe model> -n <caffe netlist>
	</pre><p>For example, to extract the data from pre-trained caffe Alex model to binary file: </p><pre class="fragment"> python caffe_data_extractor.py -m /path/to/bvlc_alexnet.caffemodel -n /path/to/caffe/models/bvlc_alexnet/deploy.prototxt
	</pre><p>The script has been tested under Python2.7.</p>
	<h2><a class="anchor" id="caffe_result"></a>
	What is the expected output from the script</h2>
	<p>If the script runs successfully, it prints the names and shapes of each layer onto the standard output and generates *.npy files containing the weights and biases of each layer.</p>
	<p>The <a class="el" href="namespacearm__compute_1_1utils.xhtml#af214346f90d640ac468dd90fa2a275cc" title="Load the tensor with pre-trained data from a binary file.">arm_compute::utils::load_trained_data</a> shows how one could load the weights and biases into tensor from the .npy file by the help of Accessor.</p>
	<h1><a class="anchor" id="tensorflow_data_extractor"></a>
	Extract data from pre-trained tensorflow model</h1>
	<p>The script tensorflow_data_extractor.py extracts trainable parameters (e.g. values of weights and biases) from a trained tensorflow model. A tensorflow model consists of the following two files:</p>
	<p>{model_name}.data-{step}-{global_step}: A binary file containing values of each variable.</p>
	<p>{model_name}.meta: A binary file containing a MetaGraph struct which defines the graph structure of the neural network.</p>
	<dl class="section note"><dt>Note</dt><dd>Since Tensorflow version 0.11 the binary checkpoint file which contains the values for each parameter has the format of: {model_name}.data-{step}-of-{max_step} instead of: {model_name}.ckpt When dealing with binary files with version >= 0.11, only pass {model_name} to -m option; when dealing with binary files with version < 0.11, pass the whole file name {model_name}.ckpt to -m option.</dd>
	<dd>
	This script relies on the parameters to be extracted being in the 'trainable_variables' tensor collection. By default all variables are automatically added to this collection unless specified otherwise by the user. Thus should a user alter this default behavior and/or want to extract parameters from other collections, tf.GraphKeys.TRAINABLE_VARIABLES should be replaced accordingly.</dd></dl>
	<h2><a class="anchor" id="tensorflow_how_to"></a>
	How to use the script</h2>
	<p>Install tensorflow and numpy.</p>
	<p>Download the pre-trained tensorflow model.</p>
	<p>Run tensorflow_data_extractor.py with </p><pre class="fragment"> python tensorflow_data_extractor -m <path_to_binary_checkpoint_file> -n <path_to_metagraph_file>
	</pre><p>For example, to extract the data from pre-trained tensorflow Alex model to binary files: </p><pre class="fragment"> python tensorflow_data_extractor -m /path/to/bvlc_alexnet -n /path/to/bvlc_alexnet.meta
	</pre><p>Or for binary checkpoint files before Tensorflow 0.11: </p><pre class="fragment"> python tensorflow_data_extractor -m /path/to/bvlc_alexnet.ckpt -n /path/to/bvlc_alexnet.meta
	</pre><dl class="section note"><dt>Note</dt><dd>with versions >= Tensorflow 0.11 only model name is passed to the -m option</dd></dl>
	<p>The script has been tested with Tensorflow 1.2, 1.3 on Python 2.7.6 and Python 3.4.3.</p>
	<h2><a class="anchor" id="tensorflow_result"></a>
	What is the expected output from the script</h2>
	<p>If the script runs successfully, it prints the names and shapes of each parameter onto the standard output and generates .npy files containing the weights and biases of each layer.</p>
	<p>The <a class="el" href="namespacearm__compute_1_1utils.xhtml#af214346f90d640ac468dd90fa2a275cc" title="Load the tensor with pre-trained data from a binary file.">arm_compute::utils::load_trained_data</a> shows how one could load the weights and biases into tensor from the .npy file by the help of Accessor.</p>
	<h1><a class="anchor" id="tf_frozen_model_extractor"></a>
	Extract data from pre-trained frozen tensorflow model</h1>
	<p>The script tf_frozen_model_extractor.py extracts trainable parameters (e.g. values of weights and biases) from a frozen trained Tensorflow model.</p>
	<h2><a class="anchor" id="tensorflow_frozen_how_to"></a>
	How to use the script</h2>
	<p>Install Tensorflow and NumPy.</p>
	<p>Download the pre-trained Tensorflow model and freeze the model using the architecture and the checkpoint file.</p>
	<p>Run tf_frozen_model_extractor.py with </p><pre class="fragment"> python tf_frozen_model_extractor -m <path_to_frozen_pb_model_file> -d <path_to_store_parameters>
	</pre><p>For example, to extract the data from pre-trained Tensorflow model to binary files: </p><pre class="fragment"> python tf_frozen_model_extractor -m /path/to/inceptionv3.pb -d ./data
	</pre><h2><a class="anchor" id="tensorflow_frozen_result"></a>
	What is the expected output from the script</h2>
	<p>If the script runs successfully, it prints the names and shapes of each parameter onto the standard output and generates .npy files containing the weights and biases of each layer.</p>
	<p>The <a class="el" href="namespacearm__compute_1_1utils.xhtml#af214346f90d640ac468dd90fa2a275cc" title="Load the tensor with pre-trained data from a binary file.">arm_compute::utils::load_trained_data</a> shows how one could load the weights and biases into tensor from the .npy file by the help of Accessor.</p>
	<h1><a class="anchor" id="validate_examples"></a>
	Validating examples</h1>
	<p>Using one of the provided scripts will generate files containing the trainable parameters.</p>
	<p>You can validate a given graph example on a list of inputs by running: </p><pre class="fragment">LD_LIBRARY_PATH=lib ./<graph_example> --validation-range='<validation_range>' --validation-file='<validation_file>' --validation-path='/path/to/test/images/' --data='/path/to/weights/'
	</pre><p>e.g:</p>
	<p>LD_LIBRARY_PATH=lib ./bin/graph_alexnet –target=CL –layout=NHWC –type=F32 –threads=4 –validation-range='16666,24998' –validation-file='val.txt' –validation-path='images/' –data='data/'</p>
	<p>where: validation file is a plain document containing a list of images along with their expected label value. e.g: </p><pre class="fragment">val_00000001.JPEG 65
	val_00000002.JPEG 970
	val_00000003.JPEG 230
	val_00000004.JPEG 809
	val_00000005.JPEG 516
	</pre><p>–validation-range is the index range of the images within the validation file you want to check: e.g:</p>
	<p>–validation-range='100,200' will validate 100 images starting from 100th one in the validation file.</p>
	<p>This can be useful when parallelizing the validation process is needed. </p>
	</div></div><!-- PageDoc -->
	</div><!-- contents -->
	</div><!-- doc-content -->
	<!-- start footer part -->
	<div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
	<ul>
	<li class="footer">Generated on Thu Nov 28 2019 16:53:08 for Compute Library by
	<a href="http://www.doxygen.org/index.html">
	<img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.15 </li>
	</ul>
	</div>
	</body>
	</html>