Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 1 年 前同步成功

通知 2299
Star 20931
Fork 5422
P
Paddle
切换分支/标签
查找文件 普通视图历史永久链接Permalink
tensor_array.html 38.4 KB
Newer Older
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 


<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
  <meta charset="utf-8">
  
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  
  <title>Design for TensorArray &mdash; PaddlePaddle  documentation</title>
  

  
  

  

  
  
    

  

  
  
    <link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
  

  
  
        <link rel="index" title="Index"
              href="../genindex.html"/>
        <link rel="search" title="Search" href="../search.html"/>
    <link rel="top" title="PaddlePaddle  documentation" href="../index.html"/> 

  <link rel="stylesheet" href="https://cdn.jsdelivr.net/perfect-scrollbar/0.6.14/css/perfect-scrollbar.min.css" type="text/css" />
  <link rel="stylesheet" href="../_static/css/override.css" type="text/css" />
  <script>
  var _hmt = _hmt || [];
  (function() {
    var hm = document.createElement("script");
    hm.src = "//hm.baidu.com/hm.js?b9a314ab40d04d805655aab1deee08ba";
    var s = document.getElementsByTagName("script")[0]; 
    s.parentNode.insertBefore(hm, s);
  })();
  </script>

  

  
  <script src="../_static/js/modernizr.min.js"></script>

</head>

<body class="wy-body-for-nav" role="document">

  
  <header class="site-header">
    <div class="site-logo">
      <a href="/"><img src="../_static/images/PP_w.png"></a>
    </div>
    <div class="site-nav-links">
      <div class="site-menu">
        <a class="fork-on-github" href="https://github.com/PaddlePaddle/Paddle" target="_blank"><i class="fa fa-github"></i>Fork me on Github</a>
        <div class="language-switcher dropdown">
          <a type="button" data-toggle="dropdown">
            <span>English</span>
            <i class="fa fa-angle-up"></i>
            <i class="fa fa-angle-down"></i>
          </a>
          <ul class="dropdown-menu">
            <li><a href="/doc_cn">中文</a></li>
            <li><a href="/doc">English</a></li>
          </ul>
        </div>
        <ul class="site-page-links">
          <li><a href="/">Home</a></li>
        </ul>
      </div>
      <div class="doc-module">
        
        <ul>
<li class="toctree-l1"><a class="reference internal" href="../getstarted/index_en.html">GET STARTED</a></li>
<li class="toctree-l1"><a class="reference internal" href="../howto/index_en.html">HOW TO</a></li>
<li class="toctree-l1"><a class="reference internal" href="../api/index_en.html">API</a></li>
T
Deploy to GitHub Pages: 664691c89af182de3919fc72a7c875311e03893d  
Travis CI 已提交
87 
<li class="toctree-l1"><a class="reference internal" href="../mobile/index_en.html">MOBILE</a></li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 
</ul>

        
<div role="search">
  <form id="rtd-search-form" class="wy-form" action="../search.html" method="get">
    <input type="text" name="q" placeholder="Search docs" />
    <input type="hidden" name="check_keywords" value="yes" />
    <input type="hidden" name="area" value="default" />
  </form>
</div>        
      </div>
    </div>
  </header>
  
  <div class="main-content-wrap">

    
    <nav class="doc-menu-vertical" role="navigation">
        
          
          <ul>
<li class="toctree-l1"><a class="reference internal" href="../getstarted/index_en.html">GET STARTED</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../getstarted/build_and_install/index_en.html">Install and Build</a><ul>
T
Deploy to GitHub Pages: 0c0ff2828ccedb51db23290d6df9e4c83839d6af  
Travis CI 已提交
111 112 
<li class="toctree-l3"><a class="reference internal" href="../getstarted/build_and_install/pip_install_en.html">Install Using pip</a></li>
<li class="toctree-l3"><a class="reference internal" href="../getstarted/build_and_install/docker_install_en.html">Run in Docker Containers</a></li>
T
Deploy to GitHub Pages: 458ffbf4fe7fe6da5e2ea133c027d560815cc3cc  
Travis CI 已提交
113 
<li class="toctree-l3"><a class="reference internal" href="../howto/dev/build_en.html">Build using Docker</a></li>
T
Deploy to GitHub Pages: 0c0ff2828ccedb51db23290d6df9e4c83839d6af  
Travis CI 已提交
114 
<li class="toctree-l3"><a class="reference internal" href="../getstarted/build_and_install/build_from_source_en.html">Build from Sources</a></li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
115 116 117 118 119 120 121 122 123 124 125 
</ul>
</li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="../howto/index_en.html">HOW TO</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../howto/usage/cmd_parameter/index_en.html">Set Command-line Parameters</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cmd_parameter/use_case_en.html">Use Case</a></li>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cmd_parameter/arguments_en.html">Argument Outline</a></li>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cmd_parameter/detail_introduction_en.html">Detail Description</a></li>
</ul>
</li>
T
Deploy to GitHub Pages: 63ffe5250a120ff430469b8d000deb2b031c4881  
Travis CI 已提交
126 
<li class="toctree-l2"><a class="reference internal" href="../howto/usage/cluster/cluster_train_en.html">PaddlePaddle Distributed Training</a></li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
127 
<li class="toctree-l2"><a class="reference internal" href="../howto/dev/new_layer_en.html">Write New Layers</a></li>
T
Deploy to GitHub Pages: f122a5da2f27038b48f6ed607e296d762050e920  
Travis CI 已提交
128 
<li class="toctree-l2"><a class="reference internal" href="../howto/dev/contribute_to_paddle_en.html">Contribute Code</a></li>
T
Deploy to GitHub Pages: 70ee79c2567d5128a9d527b444fcee30f83c02a5  
Travis CI 已提交
129 
<li class="toctree-l2"><a class="reference internal" href="../howto/dev/write_docs_en.html">Contribute Documentation</a></li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 
<li class="toctree-l2"><a class="reference internal" href="../howto/deep_model/rnn/index_en.html">RNN Models</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../howto/deep_model/rnn/rnn_config_en.html">RNN Configuration</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../howto/optimization/gpu_profiling_en.html">Tune GPU Performance</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="../api/index_en.html">API</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../api/v2/model_configs.html">Model Configuration</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/activation.html">Activation</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/layer.html">Layers</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/evaluators.html">Evaluators</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/optimizer.html">Optimizer</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/pooling.html">Pooling</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/networks.html">Networks</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/config/attr.html">Parameter Attribute</a></li>
</ul>
</li>
T
Deploy to GitHub Pages: d4d8f742a3bc4366bf410bc9ff29c0c32666bf60  
Travis CI 已提交
148 149 150 151 152 153 
<li class="toctree-l2"><a class="reference internal" href="../api/v2/data.html">Data Reader Interface and DataSets</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/data/data_reader.html">Data Reader Interface</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/data/image.html">Image Interface</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/data/dataset.html">Dataset</a></li>
</ul>
</li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
154 
<li class="toctree-l2"><a class="reference internal" href="../api/v2/run_logic.html">Training and Inference</a></li>
T
Deploy to GitHub Pages: 1fe05c458fa1d7ff1949759c2a06ed6d19ab8048  
Travis CI 已提交
155 156 157 158 159 160 161 162 163 164 165 166 167 
<li class="toctree-l2"><a class="reference internal" href="../api/v2/fluid.html">Fluid</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/layers.html">Layers</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/data_feeder.html">DataFeeder</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/executor.html">Executor</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/initializer.html">Initializer</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/evaluator.html">Evaluator</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/nets.html">Nets</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/optimizer.html">Optimizer</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/param_attr.html">ParamAttr</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/profiler.html">Profiler</a></li>
<li class="toctree-l3"><a class="reference internal" href="../api/v2/fluid/regularizer.html">Regularizer</a></li>
</ul>
</li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
168 169 
</ul>
</li>
T
Deploy to GitHub Pages: 664691c89af182de3919fc72a7c875311e03893d  
Travis CI 已提交
170 171 
<li class="toctree-l1"><a class="reference internal" href="../mobile/index_en.html">MOBILE</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../mobile/cross_compiling_for_android_en.html">Build PaddlePaddle for Android</a></li>
T
Deploy to GitHub Pages: 78c20e3ed3ca9c03239292ec1c3769697dc976c4  
Travis CI 已提交
172 
<li class="toctree-l2"><a class="reference internal" href="../mobile/cross_compiling_for_ios_en.html">PaddlePaddle Compiling Guide for iOS</a></li>
T
Deploy to GitHub Pages: 664691c89af182de3919fc72a7c875311e03893d  
Travis CI 已提交
173 174 175 
<li class="toctree-l2"><a class="reference internal" href="../mobile/cross_compiling_for_raspberry_en.html">Build PaddlePaddle for Raspberry Pi</a></li>
</ul>
</li>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 
</ul>

        
    </nav>
    
    <section class="doc-content-wrap">

      

 







<div role="navigation" aria-label="breadcrumbs navigation">
  <ul class="wy-breadcrumbs">
      
    <li>Design for TensorArray</li>
  </ul>
</div>
      
      <div class="wy-nav-content" id="doc-content">
        <div class="rst-content">
          <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
           <div itemprop="articleBody">
            
  <div class="section" id="design-for-tensorarray">
<span id="design-for-tensorarray"></span><h1>Design for TensorArray<a class="headerlink" href="#design-for-tensorarray" title="Permalink to this headline"></a></h1>
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 
<p>This design doc presents the necessity of a new C++ class <code class="docutils literal"><span class="pre">TensorArray</span></code>.
In addition to the very simple C++ implementation</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">TensorArray</span> <span class="p">{</span>
 <span class="k">public</span><span class="o">:</span>
  <span class="k">explicit</span> <span class="n">TensorArray</span><span class="p">(</span><span class="k">const</span> <span class="n">LoDTensor</span><span class="o">&amp;</span><span class="p">);</span>
  <span class="k">explicit</span> <span class="nf">TensorArray</span><span class="p">(</span><span class="kt">size_t</span> <span class="n">size</span><span class="p">);</span>

 <span class="k">private</span><span class="o">:</span>
  <span class="n">vector</span><span class="o">&lt;</span><span class="n">LoDTensor</span><span class="o">&gt;</span> <span class="n">values_</span><span class="p">;</span>
<span class="p">};</span>
</pre></div>
</div>
<p>We also need to expose it to PaddlePaddle&#8217;s Python API,
because users would want to use it with our very flexible operators <code class="docutils literal"><span class="pre">WhileLoop</span></code>.
An example for a RNN based on dynamic operators is</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="nb">input</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="n">num_steps</span> <span class="o">=</span> <span class="n">Var</span><span class="p">(</span><span class="mi">12</span><span class="p">)</span>

<span class="n">TensorArray</span> <span class="n">states</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">num_steps</span><span class="p">)</span>
<span class="n">TensorArray</span> <span class="n">step_inputs</span><span class="p">(</span><span class="n">unstack_from</span><span class="o">=</span><span class="nb">input</span><span class="p">)</span>
<span class="n">TensorArray</span> <span class="n">step_outputs</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">num_steps</span><span class="p">)</span>

<span class="n">W</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="n">U</span> <span class="o">=</span> <span class="n">Tensor</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="n">default_state</span> <span class="o">=</span> <span class="n">some_op</span><span class="p">()</span>

<span class="n">step</span> <span class="o">=</span> <span class="n">Var</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>

<span class="n">wloop</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">create_whileloop</span><span class="p">(</span><span class="n">loop_vars</span><span class="o">=</span><span class="p">[</span><span class="n">step</span><span class="p">])</span>
<span class="k">with</span> <span class="n">wloop</span><span class="o">.</span><span class="n">frame</span><span class="p">():</span>
    <span class="n">wloop</span><span class="o">.</span><span class="n">break_if</span><span class="p">(</span><span class="n">pd</span><span class="o">.</span><span class="n">equal</span><span class="p">(</span><span class="n">step</span><span class="p">,</span> <span class="n">num_steps</span><span class="p">)</span>
    <span class="n">pre_state</span> <span class="o">=</span> <span class="n">states</span><span class="o">.</span><span class="n">read</span><span class="p">(</span><span class="n">step</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">default_state</span><span class="p">)</span>
    <span class="n">step_input</span> <span class="o">=</span> <span class="n">step_inputs</span><span class="o">.</span><span class="n">read</span><span class="p">(</span><span class="n">step</span><span class="p">)</span>
    <span class="n">state</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">sigmoid</span><span class="p">(</span><span class="n">pd</span><span class="o">.</span><span class="n">matmul</span><span class="p">(</span><span class="n">U</span><span class="p">,</span> <span class="n">pre_state</span><span class="p">)</span> <span class="o">+</span> <span class="n">pd</span><span class="o">.</span><span class="n">matmul</span><span class="p">(</span><span class="n">W</span><span class="p">,</span> <span class="n">step_input</span><span class="p">))</span>
    <span class="n">states</span><span class="o">.</span><span class="n">write</span><span class="p">(</span><span class="n">step</span><span class="p">,</span> <span class="n">state</span><span class="p">)</span>
    <span class="n">step_outputs</span><span class="o">.</span><span class="n">write</span><span class="p">(</span><span class="n">step</span><span class="p">,</span> <span class="n">state</span><span class="p">)</span> <span class="c1"># output state</span>
    <span class="n">step</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">state</span><span class="o">+</span><span class="mi">1</span><span class="p">)</span>

<span class="n">output</span> <span class="o">=</span> <span class="n">step_outputs</span><span class="o">.</span><span class="n">stack</span><span class="p">()</span>
</pre></div>
</div>
<div class="section" id="background">
<span id="background"></span><h2>Background<a class="headerlink" href="#background" title="Permalink to this headline"></a></h2>
<p>Steps are one of the core concepts of RNN. In each time step of RNN, there should be several input segments, states, and output segments; all these components act like arrays, for example, call <code class="docutils literal"><span class="pre">states[step_id]</span></code> will get the state in <code class="docutils literal"><span class="pre">step_id</span></code>th time step.</p>
<p>An RNN can be implemented with the following pseudocode</p>
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="n">Array</span> <span class="n">states</span><span class="p">;</span>
<span class="n">Array</span> <span class="n">input_segments</span><span class="p">;</span>
<span class="n">Array</span> <span class="n">output_segments</span><span class="p">;</span>
<span class="n">Parameter</span> <span class="n">W</span><span class="p">,</span> <span class="n">U</span><span class="p">;</span>

<span class="n">step</span> <span class="o">=</span> <span class="mi">1</span>
<span class="n">seq_len</span> <span class="o">=</span> <span class="mi">12</span>
<span class="n">while_loop</span> <span class="p">{</span>
   <span class="k">if</span> <span class="p">(</span><span class="n">step</span> <span class="o">==</span> <span class="n">seq_len</span><span class="p">)</span> <span class="k">break</span><span class="p">;</span>
    <span class="n">states</span><span class="p">[</span><span class="n">step</span><span class="p">]</span> <span class="o">=</span> <span class="n">sigmoid</span><span class="p">(</span><span class="n">W</span> <span class="o">*</span> <span class="n">states</span><span class="p">[</span><span class="n">step</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span> <span class="o">+</span> <span class="n">U</span> <span class="o">*</span> <span class="n">input_segments</span><span class="p">[</span><span class="n">step</span><span class="p">]);</span>
    <span class="n">output_segments</span><span class="p">[</span><span class="n">step</span><span class="p">]</span> <span class="o">=</span> <span class="n">states</span><span class="p">[</span><span class="n">step</span><span class="p">]</span> <span class="c1">// take state as output</span>
   <span class="n">step</span><span class="o">++</span><span class="p">;</span>
<span class="p">}</span>
</pre></div>
</div>
<p>According to the <a class="reference external" href="https://github.com/PaddlePaddle/Paddle/issues/4561">RNN roadmap</a>, there are several different RNNs that PaddlePaddle will eventually support.</p>
<p>Currently, the basic RNN implementation supported by PaddlePaddle is the <code class="docutils literal"><span class="pre">recurrent_op</span></code> which takes tensors as input and splits them into <code class="docutils literal"><span class="pre">input_segments</span></code>.</p>
<p>Since a tensor cannot store variable-length sequences directly, PaddlePaddle implements the tensor with level of details (<code class="docutils literal"><span class="pre">LoDTensor</span></code> for short).
Segmenting the <code class="docutils literal"><span class="pre">LoDTensor</span></code> is much more complicated than splitting a tensor, that makes it necessary to refactor the <code class="docutils literal"><span class="pre">recurrent_op</span></code> with <code class="docutils literal"><span class="pre">LoDTensor</span></code> segmenting support.</p>
<p>As the next step in RNN support, <code class="docutils literal"><span class="pre">dynamic_recurrent_op</span></code> should be introduced to handle inputs with variable-length sequences.</p>
<p>The implementation is similar to <code class="docutils literal"><span class="pre">recurrent_op</span></code>.
The key difference is the way <strong>the original input <code class="docutils literal"><span class="pre">LoDTensors</span></code> and outupts are split to get the <code class="docutils literal"><span class="pre">input_segments</span></code> and the <code class="docutils literal"><span class="pre">output_segments</span></code>.</strong></p>
<p>Though it can&#8217;t be built over <code class="docutils literal"><span class="pre">recurrent_op</span></code> or <code class="docutils literal"><span class="pre">dynamic_recurrent_op</span></code> directly,
the logic behind splitting a tensor or a LoD tensor into <code class="docutils literal"><span class="pre">input_segments</span></code> remains the same.</p>
</div>
<div class="section" id="why-tensorarray">
<span id="why-tensorarray"></span><h2>Why <code class="docutils literal"><span class="pre">TensorArray</span></code><a class="headerlink" href="#why-tensorarray" title="Permalink to this headline"></a></h2>
<p>The logic behind splitting the inputs to segments, states and outputs is similar and can be shared in a seperate module.</p>
<p>The array of <code class="docutils literal"><span class="pre">states</span></code>, <code class="docutils literal"><span class="pre">input_segments</span></code> and <code class="docutils literal"><span class="pre">output_segments</span></code> would be exposed to users when writing a dynamic RNN model similar to the above pseudo codes.</p>
<p>So there should be an array-like container, which can store the segments of a tensor or LoD tensor.</p>
<p><strong>This container can store an array of tensors and provides several methods to split a tensor or a LoD tensor</strong> .
This is where the notion of <code class="docutils literal"><span class="pre">TensorArray</span></code> comes from.</p>
</div>
<div class="section" id="introduce-tensorarray-to-uniform-all-the-three-rnns">
<span id="introduce-tensorarray-to-uniform-all-the-three-rnns"></span><h2>Introduce TensorArray to uniform all the three RNNs<a class="headerlink" href="#introduce-tensorarray-to-uniform-all-the-three-rnns" title="Permalink to this headline"></a></h2>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
287 288 289 
<p>TensorArray as a new concept is borrowed from TensorFlow,
it is meant to be used with dynamic iteration primitives such as <code class="docutils literal"><span class="pre">while_loop</span></code> and <code class="docutils literal"><span class="pre">map_fn</span></code>.</p>
<p>This concept can be used to support our new design of dynamic operations, and help to refactor some existing variant-sentence-related layers,
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
290 
such as <code class="docutils literal"><span class="pre">recurrent_op</span></code>, <code class="docutils literal"><span class="pre">RecurrentGradientMachine</span></code>.</p>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
291 292 293 
<p>In <a class="reference external" href="https://github.com/PaddlePaddle/Paddle/pull/4401">our design for dynamic RNN</a>,
<code class="docutils literal"><span class="pre">TensorArray</span></code> is used to segment inputs and store states in all time steps.
By providing some methods similar to a C++ array,
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
294 
the definition of some state-based dynamic models such as RNN can be more natural and highly flexible.</p>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
295 
</div>
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 
<div class="section" id="dynamic-operations-on-tensorarray">
<span id="dynamic-operations-on-tensorarray"></span><h2>Dynamic-operations on TensorArray<a class="headerlink" href="#dynamic-operations-on-tensorarray" title="Permalink to this headline"></a></h2>
<p><code class="docutils literal"><span class="pre">TensorArray</span></code> will be used directly when defining dynamic models, so some operators listed below should be implemented</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># several helper operators for TensorArray</span>
<span class="k">def</span> <span class="nf">tensor_array_stack</span><span class="p">(</span><span class="n">ta</span><span class="p">,</span> <span class="n">tensor</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    get a tensor array `ta`, return a packed `tensor`.</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>

<span class="k">def</span> <span class="nf">tensor_array_unstack</span><span class="p">(</span><span class="n">tensor</span><span class="p">,</span> <span class="n">ta</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    get a `tensor`, unstack it and get a tensor array `ta`.</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>

<span class="k">def</span> <span class="nf">tensor_array_write</span><span class="p">(</span><span class="n">ta</span><span class="p">,</span> <span class="n">index</span><span class="p">,</span> <span class="n">tensor</span><span class="p">,</span> <span class="n">data_shared</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    get a `tensor` and a scalar tensor `index`, write `tensor` into index-th</span>
<span class="sd">    value of the tensor array `ta`.</span>
<span class="sd">    `data_shared` is an attribute that specifies whether to copy or reference the tensors.</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>

<span class="k">def</span> <span class="nf">tensor_array_read</span><span class="p">(</span><span class="n">ta</span><span class="p">,</span> <span class="n">index</span><span class="p">,</span> <span class="n">tensor</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    get a tensor array `ta`, a scalar tensor `index`, read the index-th value of</span>
<span class="sd">    `ta` and return as the `tensor`.</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>

<span class="k">def</span> <span class="nf">tensor_array_size</span><span class="p">(</span><span class="n">ta</span><span class="p">,</span> <span class="n">tensor</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    get a tensor array `ta`, return the size of `ta` and return as the scalar `tensor`.</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>
</pre></div>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
333 
</div>
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 
<p>It is trivial for users to use so many low-level operators, so some helper methods should be proposed in python wrapper to make <code class="docutils literal"><span class="pre">TensorArray</span></code> easier to use,
for example</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">TensorArray</span><span class="p">:</span>
    <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">name</span><span class="p">):</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">name</span> <span class="o">=</span> <span class="n">name</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">desc</span> <span class="o">=</span> <span class="n">TensorArrayDesc</span><span class="p">()</span>

    <span class="k">def</span> <span class="nf">stack</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="bp">None</span><span class="p">):</span>
        <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">        Pack the values in a `TensorArray` into a tensor with rank one higher</span>
<span class="sd">        than each tensor in `values`.</span>
<span class="sd">        `stack` can be used to split tensor into time steps for RNN or whileloop.</span>

<span class="sd">        @name: str</span>
<span class="sd">            the name of the variable to output.</span>
<span class="sd">        &#39;&#39;&#39;</span>
T
Deploy to GitHub Pages: e593113a84a6227e8fc98875f66b9b865c6bc17b  
Travis CI 已提交
350 
        <span class="n">tensor</span> <span class="o">=</span> <span class="n">Var</span><span class="p">(</span><span class="n">name</span><span class="p">)</span>
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 
        <span class="n">tensor_array_stack</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">name</span><span class="p">,</span> <span class="n">tensor</span><span class="p">)</span>
        <span class="k">return</span> <span class="n">tensor</span>

    <span class="k">def</span> <span class="nf">unstack</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="nb">input</span><span class="p">):</span>
        <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">        Unpacks the given dimension of a rank-`R` tensor into rank-`(R-1)` tensors.</span>
<span class="sd">        `unstack` can be used to concatenate all the time steps for RNN or whileloop.</span>

<span class="sd">        @input: str</span>
<span class="sd">            the name of input tensor</span>
<span class="sd">        &#39;&#39;&#39;</span>
        <span class="n">tensor_array_unstack</span><span class="p">(</span><span class="n">tensor</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">name</span><span class="p">)</span>

    <span class="k">def</span> <span class="nf">write</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">index</span><span class="p">,</span> <span class="n">value</span><span class="p">,</span> <span class="n">data_shared</span><span class="o">=</span><span class="bp">True</span><span class="p">):</span>
        <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">        Write value into index of the TensorArray.</span>
<span class="sd">        If `data_shared` is set to True, than the index-th value in TensorArray will</span>
<span class="sd">        be shared with the tensor passed in.</span>

<span class="sd">        @index: str</span>
<span class="sd">            name of a scalar tensor</span>
<span class="sd">        @value: str</span>
<span class="sd">            name of a tensor</span>
<span class="sd">        @data_shared: bool</span>
<span class="sd">        &#39;&#39;&#39;</span>
        <span class="n">tensor_array_write</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">name</span><span class="p">,</span> <span class="n">index</span><span class="p">,</span> <span class="n">value</span><span class="p">,</span> <span class="n">data_shared</span><span class="p">)</span>

    <span class="k">def</span> <span class="nf">read</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">index</span><span class="p">,</span> <span class="n">output</span><span class="p">):</span>
        <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">        Read the value at location `index` in the `TensorArray`.</span>

<span class="sd">        @index: str</span>
<span class="sd">            name of a scalar tensor</span>
<span class="sd">        @output:</span>
<span class="sd">            name of a output variable</span>
<span class="sd">        &#39;&#39;&#39;</span>
        <span class="n">tensor_array_read</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">name</span><span class="p">,</span> <span class="n">index</span><span class="p">,</span> <span class="n">output</span><span class="p">)</span>


    <span class="k">def</span> <span class="nf">size</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">output</span><span class="p">):</span>
        <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">        Return the number of values.</span>

<span class="sd">        @output: str</span>
<span class="sd">            name of a scalar tensor</span>
<span class="sd">        &#39;&#39;&#39;</span>
        <span class="n">tensor_array_size</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">name</span><span class="p">,</span> <span class="n">output</span><span class="p">)</span>
</pre></div>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
399 400 401 402 
</div>
</div>
<div class="section" id="lodtensor-related-supports">
<span id="lodtensor-related-supports"></span><h2>LoDTensor-related Supports<a class="headerlink" href="#lodtensor-related-supports" title="Permalink to this headline"></a></h2>
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
403 404 
<p>The <code class="docutils literal"><span class="pre">RecurrentGradientMachine</span></code> in Paddle serves as a flexible RNN layer; it takes varience-length sequences as input, and output sequences too.</p>
<p>Since each step of RNN can only take a tensor-represented batch of data as input,
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
405 
some preprocess should be taken on the inputs such as sorting the sentences by their length in descending order and cut each word and pack to new batches.</p>
T
Deploy to GitHub Pages: 34193844adc9d67e4a9c8bb00ff9e67ffbc4ba3c  
Travis CI 已提交
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 
<p>Such cut-like operations can be embedded into <code class="docutils literal"><span class="pre">TensorArray</span></code> as general methods called <code class="docutils literal"><span class="pre">unpack</span></code> and <code class="docutils literal"><span class="pre">pack</span></code>,
these two operations are similar to <code class="docutils literal"><span class="pre">stack</span></code> and <code class="docutils literal"><span class="pre">unstack</span></code> except that they operate on variable-length sequences formated as a LoD tensor rather than a tensor.</p>
<p>Some definitions are like</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">unpack</span><span class="p">(</span><span class="n">level</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    Split LodTensor in some `level` and generate batches, if set `sort_by_length`,</span>
<span class="sd">    will sort by length.</span>

<span class="sd">    Returns:</span>
<span class="sd">        - a new `TensorArray`, whose values are LodTensors and represents batches</span>
<span class="sd">          of data.</span>
<span class="sd">        - an int32 Tensor, which stores the map from the new batch&#39;s indices to</span>
<span class="sd">          original LoDTensor</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>

<span class="k">def</span> <span class="nf">pack</span><span class="p">(</span><span class="n">level</span><span class="p">,</span> <span class="n">indices_map</span><span class="p">):</span>
    <span class="sd">&#39;&#39;&#39;</span>
<span class="sd">    Recover the original LoD-arranged LoDTensor with the values in a `TensorArray`</span>
<span class="sd">    and `level` and `indices_map`.</span>
<span class="sd">    &#39;&#39;&#39;</span>
    <span class="k">pass</span>
</pre></div>
</div>
<p>With these two methods, a varience-length sentence supported RNN can be implemented like</p>
T
Deploy to GitHub Pages: 88f3f009e82810ce8e68d897de587bf7598ef7ad  
Travis CI 已提交
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 
<div class="highlight-c++"><div class="highlight"><pre><span></span><span class="c1">// input is the varient-length data</span>
<span class="n">LodTensor</span> <span class="nf">sentence_input</span><span class="p">(</span><span class="n">xxx</span><span class="p">);</span>
<span class="n">TensorArray</span> <span class="n">ta</span><span class="p">;</span>
<span class="n">Tensor</span> <span class="n">indice_map</span><span class="p">;</span>
<span class="n">Tensor</span> <span class="n">boot_state</span> <span class="o">=</span> <span class="n">xxx</span><span class="p">;</span> <span class="c1">// to initialize rnn&#39;s first state</span>
<span class="n">TensorArray</span><span class="o">::</span><span class="n">unpack</span><span class="p">(</span><span class="n">input</span><span class="p">,</span> <span class="mi">1</span><span class="cm">/*level*/</span><span class="p">,</span> <span class="nb">true</span><span class="cm">/*sort_by_length*/</span><span class="p">,</span> <span class="o">&amp;</span><span class="n">ta</span><span class="p">,</span> <span class="o">&amp;</span><span class="n">indice_map</span><span class="p">);</span>
<span class="n">TessorArray</span> <span class="n">step_outputs</span><span class="p">;</span>
<span class="n">TensorArray</span> <span class="n">states</span><span class="p">;</span>

<span class="k">for</span> <span class="p">(</span><span class="kt">int</span> <span class="n">step</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span> <span class="n">step</span> <span class="o">=</span> <span class="n">ta</span><span class="p">.</span><span class="n">size</span><span class="p">();</span> <span class="n">step</span><span class="o">++</span><span class="p">)</span> <span class="p">{</span>
  <span class="k">auto</span> <span class="n">state</span> <span class="o">=</span> <span class="n">states</span><span class="p">.</span><span class="n">read</span><span class="p">(</span><span class="n">step</span><span class="p">);</span>
  <span class="c1">// rnnstep is a function which acts like a step of RNN</span>
  <span class="k">auto</span> <span class="n">step_input</span> <span class="o">=</span> <span class="n">ta</span><span class="p">.</span><span class="n">read</span><span class="p">(</span><span class="n">step</span><span class="p">);</span>
  <span class="k">auto</span> <span class="n">step_output</span> <span class="o">=</span> <span class="n">rnnstep</span><span class="p">(</span><span class="n">step_input</span><span class="p">,</span> <span class="n">state</span><span class="p">);</span>
  <span class="n">step_outputs</span><span class="p">.</span><span class="n">write</span><span class="p">(</span><span class="n">step_output</span><span class="p">,</span> <span class="nb">true</span><span class="cm">/*data_shared*/</span><span class="p">);</span>
<span class="p">}</span>

<span class="c1">// rnn_output is the final output of an rnn</span>
<span class="n">LoDTensor</span> <span class="n">rnn_output</span> <span class="o">=</span> <span class="n">ta</span><span class="p">.</span><span class="n">pack</span><span class="p">(</span><span class="n">ta</span><span class="p">,</span> <span class="n">indice_map</span><span class="p">);</span>
</pre></div>
</div>
<p>the code above shows that by embedding the LoDTensor-related preprocess operations into <code class="docutils literal"><span class="pre">TensorArray</span></code>,
the implementation of a RNN that supports varient-length sentences is far more concise than <code class="docutils literal"><span class="pre">RecurrentGradientMachine</span></code> because the latter mixes all the codes together, hard to read and extend.</p>
</div>
</div>


           </div>
          </div>
          <footer>
  

  <hr/>

  <div role="contentinfo">
    <p>
        &copy; Copyright 2016, PaddlePaddle developers.

    </p>
  </div>
  Built with <a href="http://sphinx-doc.org/">Sphinx</a> using a <a href="https://github.com/snide/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>. 

</footer>

        </div>
      </div>

    </section>

  </div>
  


  

    <script type="text/javascript">
        var DOCUMENTATION_OPTIONS = {
            URL_ROOT:'../',
            VERSION:'',
            COLLAPSE_INDEX:false,
            FILE_SUFFIX:'.html',
            HAS_SOURCE:  true,
            SOURCELINK_SUFFIX: ".txt",
        };
    </script>
      <script type="text/javascript" src="../_static/jquery.js"></script>
      <script type="text/javascript" src="../_static/underscore.js"></script>
      <script type="text/javascript" src="../_static/doctools.js"></script>
      <script type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
       
  

  
  
    <script type="text/javascript" src="../_static/js/theme.js"></script>
  
  
  <script src="https://maxcdn.bootstrapcdn.com/bootstrap/3.3.7/js/bootstrap.min.js" integrity="sha384-Tc5IQib027qvyjSMfHjOMaLkfuWVxZxUPnCJA7l2mCWNIpG9mGCD8wGNIcPD7Txa" crossorigin="anonymous"></script>
  <script src="https://cdn.jsdelivr.net/perfect-scrollbar/0.6.14/js/perfect-scrollbar.jquery.min.js"></script>
  <script src="../_static/js/paddle_doc_init.js"></script> 

</body>
</html>