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
backward.html 29.5 KB
Newer Older
T
Deploy to GitHub Pages: 6a5cf28a7f53dc44691f52849b8759a251cfebd7  
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 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 


<!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>Backward Building &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>
<li class="toctree-l1"><a class="reference internal" href="../mobile/index_en.html">MOBILE</a></li>
</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>
<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>
<li class="toctree-l3"><a class="reference internal" href="../howto/dev/build_en.html">Build using Docker</a></li>
<li class="toctree-l3"><a class="reference internal" href="../getstarted/build_and_install/build_from_source_en.html">Build from Sources</a></li>
</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>
<li class="toctree-l2"><a class="reference internal" href="../howto/usage/cluster/cluster_train_en.html">Distributed Training</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cluster/fabric_en.html">fabric</a></li>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cluster/openmpi_en.html">openmpi</a></li>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cluster/k8s_en.html">kubernetes</a></li>
<li class="toctree-l3"><a class="reference internal" href="../howto/usage/cluster/k8s_aws_en.html">kubernetes on AWS</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../howto/dev/new_layer_en.html">Write New Layers</a></li>
<li class="toctree-l2"><a class="reference internal" href="../howto/dev/contribute_to_paddle_en.html">Contribute Code</a></li>
<li class="toctree-l2"><a class="reference internal" href="../howto/dev/write_docs_en.html">Contribute Documentation</a></li>
<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>
<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>
<li class="toctree-l2"><a class="reference internal" href="../api/v2/run_logic.html">Training and Inference</a></li>
<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>
</ul>
</li>
<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>
<li class="toctree-l2"><a class="reference internal" href="../mobile/cross_compiling_for_ios_en.html">PaddlePaddle Compiling Guide for iOS</a></li>
<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>
</ul>

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

      

 







<div role="navigation" aria-label="breadcrumbs navigation">
  <ul class="wy-breadcrumbs">
      
    <li>Backward Building</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="backward-building">
<span id="backward-building"></span><h1>Backward Building<a class="headerlink" href="#backward-building" title="Permalink to this headline"></a></h1>
<div class="section" id="motivation">
<span id="motivation"></span><h2>Motivation<a class="headerlink" href="#motivation" title="Permalink to this headline"></a></h2>
<p>In Neural Network, most models are solved by the backpropagation algorithm(known as <strong>BP</strong>) at present. Technically, BP calculates the gradient of the loss function, then propagates it back through the networks following the chain rule. However, when configuring the model structure, users do not need to define the backward part. So a mechanism is required by the framework which can complete the model&#8217;s backward part automatically according to the given forward part.</p>
<p>When implementing a specific <code class="docutils literal"><span class="pre">op</span></code>, the developer is also asked to implement its backward version, called <code class="docutils literal"><span class="pre">grad_op</span></code>. A <code class="docutils literal"><span class="pre">grad_op</span></code> takes gradients of its corresponding <code class="docutils literal"><span class="pre">op</span></code>&#8216;s outputs, and calculate gradients of the <code class="docutils literal"><span class="pre">op</span></code>&#8216;s inputs. During the building of a model&#8217;s backward part, the framework creates each forward <code class="docutils literal"><span class="pre">op</span></code>&#8216;s <code class="docutils literal"><span class="pre">grad_op</span></code>, and then string them together in reverse order of forwarding part. In this way, gradients spread from the end to the beginning of the model, in another word, from the loss to parameters.</p>
</div>
<div class="section" id="challenges">
<span id="challenges"></span><h2>Challenges<a class="headerlink" href="#challenges" title="Permalink to this headline"></a></h2>
<p>The motivation of backward building is apparent. However, implementation it correctly is not so easy. In the <strong>Fluid</strong> design, a deep learning model is described by <code class="docutils literal"><span class="pre">Program</span></code>, <code class="docutils literal"><span class="pre">Block</span></code>, <code class="docutils literal"><span class="pre">Op</span></code> and <code class="docutils literal"><span class="pre">Variable</span></code>. The <code class="docutils literal"><span class="pre">Block</span></code> itself can be nested. It means that the <code class="docutils literal"><span class="pre">op</span></code>s and <code class="docutils literal"><span class="pre">variable</span></code>s are scattered across different blocks rather than all be gathered in a single graph. Our backward building algorithm shall visit blocks in recursive order and be able to insert <code class="docutils literal"><span class="pre">grad_op</span></code>s and new created <code class="docutils literal"><span class="pre">variable</span></code>s into the right place.</p>
</div>
<div class="section" id="usage">
<span id="usage"></span><h2>Usage<a class="headerlink" href="#usage" title="Permalink to this headline"></a></h2>
<p>Although the whole algorithm is comprised of many functions, only one is exposed as API:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">append_backward</span><span class="p">(</span><span class="n">loss</span><span class="p">,</span> <span class="n">parameter_list</span><span class="o">=</span><span class="bp">None</span><span class="p">,</span> <span class="n">no_grad_set</span><span class="o">=</span><span class="bp">None</span><span class="p">):</span>
    <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">    Append backward part to main_program</span>

<span class="sd">    Args:</span>
<span class="sd">        loss(Variable): The variable generated by the cost function.</span>
<span class="sd">        parameter_list(list): Parameters that need to be updated by optimizers.</span>
<span class="sd">            If None, it means all parameters need to be updated.</span>

<span class="sd">        no_grad_set(set): Variables that have no gradients in Block 0. </span>
<span class="sd">            If None, the set will be generated inside the function and </span>
<span class="sd">            contains all variables with `step_gradient=True` from all blocks.</span>
<span class="sd">        </span>
<span class="sd">    Return:</span>
<span class="sd">        (list[Variable]): list of (parameters, gradients) pair.</span>
<span class="sd">    &quot;&quot;&quot;</span>
</pre></div>
</div>
<p>By invoking this API, the framework appends backward part of the program where the <code class="docutils literal"><span class="pre">loss</span></code> is. It takes three arguments. <code class="docutils literal"><span class="pre">loss</span></code> means the final loss value. It must be a scalar and is usually the output of the loss layer. It is also where the gradient generated and backpropagation starts. <code class="docutils literal"><span class="pre">parameter_list</span></code> marks all parameters needs updating. If it&#8217;s <code class="docutils literal"><span class="pre">None</span></code>, all parameter will be updated by optimizers. <code class="docutils literal"><span class="pre">no_grad_set</span></code> marks variables without gradient. if all outputs of some <code class="docutils literal"><span class="pre">grad_op</span></code> are in <code class="docutils literal"><span class="pre">no_grad_set</span></code>, the <code class="docutils literal"><span class="pre">grad_op</span></code> will not be run.</p>
<p>This API will be invoked automatically before optimizer building.
As a result, in most cases, users do not need to invoke the API by themselves to append backward part.</p>
</div>
<div class="section" id="implementation">
<span id="implementation"></span><h2>Implementation<a class="headerlink" href="#implementation" title="Permalink to this headline"></a></h2>
<p>The implementation of backward building algorithm is in <code class="docutils literal"><span class="pre">backward.py</span></code> file. The whole algorithm can be divided into two independent parts: creating <code class="docutils literal"><span class="pre">grad_op</span></code>s and creating new variables.</p>
<div class="section" id="creating-grad-ops">
<span id="creating-grad-ops"></span><h3>Creating <code class="docutils literal"><span class="pre">grad_op</span></code>s<a class="headerlink" href="#creating-grad-ops" title="Permalink to this headline"></a></h3>
<p>The creating of <code class="docutils literal"><span class="pre">grad_op</span></code>s is implemented by:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">_append_backward_ops_</span><span class="p">(</span><span class="n">target</span><span class="p">,</span>
                          <span class="n">block</span><span class="p">,</span>
                          <span class="n">target_block</span><span class="p">,</span>
                          <span class="n">no_grad_dict</span><span class="p">,</span>
                          <span class="n">grad_to_var</span><span class="p">):</span>
    <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">    Create all grad ops, and insert them into given block</span>

<span class="sd">    Args:</span>
<span class="sd">        target(Variable): the target variable of forward pass</span>
<span class="sd">        block(Block): the block where forward ops are</span>
<span class="sd">        target_block(Block): the block which is going to hold new generated grad ops</span>
<span class="sd">        no_grad_dict(dict): </span>
<span class="sd">            key(int)  block index</span>
<span class="sd">            val(set) a set of varibale names. These varibales have no gradient</span>
<span class="sd">        grad_to_var(dict)(output argument):</span>
<span class="sd">            key(str): grad variable name</span>
<span class="sd">            val(str): corresponding forward variable name</span>
<span class="sd">    &quot;&quot;&quot;</span>
</pre></div>
</div>
<p>Given a <code class="docutils literal"><span class="pre">block</span></code>, the function will traverses all <code class="docutils literal"><span class="pre">op</span></code>s in this block in reverse order, gets corresponding <code class="docutils literal"><span class="pre">grad_op</span></code> from the C++ core via <code class="docutils literal"><span class="pre">core.get_grad_op_desc()</span></code>, then append it to <code class="docutils literal"><span class="pre">target_block</span></code>.</p>
<p>However, some specific <code class="docutils literal"><span class="pre">op</span></code>(e.g. <code class="docutils literal"><span class="pre">while_op</span></code>, <code class="docutils literal"><span class="pre">if_else_op</span></code>) can hold its own sub-block. For these sub-blocks contains <code class="docutils literal"><span class="pre">op</span></code>s as well, the <code class="docutils literal"><span class="pre">grad_op</span></code> creating should be recursive.</p>
<p>During the reverse traversal, we check each <code class="docutils literal"><span class="pre">op</span></code> whether it has an attribute named <code class="docutils literal"><span class="pre">sub_block</span></code>. If so, it means there is a sub-block and we need to deal with it first. After creating a new block whose father is the one in <code class="docutils literal"><span class="pre">op</span></code>&#8216;s attribute, we invoke <code class="docutils literal"><span class="pre">_append_backward_ops_()</span></code> recursively, assigning the new block to parameter <code class="docutils literal"><span class="pre">target_block</span></code> and the one in <code class="docutils literal"><span class="pre">op</span></code>&#8216;s attribute to <code class="docutils literal"><span class="pre">block</span></code>. The <em>pseudo-code</em> shows this process:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span>******* pseudo-code ********
for op in reversed(block.ops):
    if op has an attribute named &#39;sub_block&#39;:
        Get the sub-block(`s_block`) from op&#39;s attribute.
        Create a new block(`grad_s_block`), whose father is `s_block`.
        Invoke _append_backward_ops_(), with `block=s_block` and `target_block=grad_s_block`
    
    Invoke `core.get_grad_op_desc()` to get op&#39;s grad_op.
    Insert name correspondings between variables and their gradients of the grad_op to grad_to_var
    Assign grad_s_block to grad_op as it&#39;s &#39;sub_block&#39; attribute.
    Append grad_op to current target_block.
</pre></div>
</div>
<p>The first invoking of <code class="docutils literal"><span class="pre">_append_backward_ops_()</span></code> is initiated by <code class="docutils literal"><span class="pre">append_backward()</span></code>, in which parameters <code class="docutils literal"><span class="pre">block</span></code> and <code class="docutils literal"><span class="pre">target_block</span></code> are all assigned with root block(the block with index 0).</p>
</div>
<div class="section" id="corner-cases-of-grad-op-creating">
<span id="corner-cases-of-grad-op-creating"></span><h3>Corner Cases of <code class="docutils literal"><span class="pre">grad_op</span></code> Creating<a class="headerlink" href="#corner-cases-of-grad-op-creating" title="Permalink to this headline"></a></h3>
<p>In the previous section, we show the regular process of <code class="docutils literal"><span class="pre">grad_op</span></code> creating. However, in some corner cases, the conventional algorithm is not enough to get the correct result and appending handling is required. These additional processes run after the algorithm mentioned above and do some special adjusts on its output <code class="docutils literal"><span class="pre">grad_op</span></code>s.</p>
<div class="section" id="shared-variables">
<span id="shared-variables"></span><h4>Shared Variables<a class="headerlink" href="#shared-variables" title="Permalink to this headline"></a></h4>
<p>If a variable is read by more than one <code class="docutils literal"><span class="pre">op</span></code> in the forward pass, its gradient is likely to be written by more than one <code class="docutils literal"><span class="pre">grad_op</span></code>s in the next backward pass. To make the gradient result being the sum of all <code class="docutils literal"><span class="pre">grad_op</span></code>s&#8217; outputs instead of the last running one, we assign each output with a temporary variable and then add a <code class="docutils literal"><span class="pre">sum_op</span></code> to add them up.</p>
<p>For the debug convenience, if the final gradient name is <code class="docutils literal"><span class="pre">w&#64;GRAD</span></code>, it&#8217;s corresponding temporary variables will be named as <code class="docutils literal"><span class="pre">w&#64;GRAD&#64;RENAME&#64;0</span></code>, <code class="docutils literal"><span class="pre">w&#64;GRAD&#64;RENAME&#64;1</span></code>...</p>
<p>See function <code class="docutils literal"><span class="pre">_addup_repetitive_outputs_</span></code> in <code class="docutils literal"><span class="pre">backward.py</span></code> for implementation details.</p>
</div>
<div class="section" id="no-gradient-variables">
<span id="no-gradient-variables"></span><h4>No Gradient Variables<a class="headerlink" href="#no-gradient-variables" title="Permalink to this headline"></a></h4>
<p>In our framework, variables can be marked as <em>no_gradient</em>, it means that the gradient of this variable is unnecessary and can be considered as zero in model training. Apparently, when all the outputs of some <code class="docutils literal"><span class="pre">grad_op</span></code> are marked as <em>no_gradient</em>, the <code class="docutils literal"><span class="pre">grad_op</span></code> itself can be skipped in backward pass.</p>
<p>But these unnecessary gradients still need to be creating and initialized by something, otherwise following <code class="docutils literal"><span class="pre">grad_op</span></code>s who take these gradients as inputs take the risk of using uninitialized memory. In our code, we employ <code class="docutils literal"><span class="pre">fill_zeros_like_op</span></code> to initialize them as all zeros.</p>
<p>This features are implemented in function <code class="docutils literal"><span class="pre">_remove_no_grad_branch_</span></code>. It checks new created <code class="docutils literal"><span class="pre">grad_op</span></code>s one-by-one, removes whose outputs are all in <code class="docutils literal"><span class="pre">no_grad_set</span></code> or inserts <code class="docutils literal"><span class="pre">fill_zeros_like_op</span></code> when its necessary. We can get the <code class="docutils literal"><span class="pre">no_grad_set</span></code> from the <code class="docutils literal"><span class="pre">_append_backward_ops_</span></code> argument <code class="docutils literal"><span class="pre">no_grad_dict</span></code> or generate it on the fly by scanning all variables&#8217; <code class="docutils literal"><span class="pre">no_gradient</span></code> attribute(True or False).</p>
</div>
</div>
<div class="section" id="creating-backward-variables">
<span id="creating-backward-variables"></span><h3>Creating Backward Variables<a class="headerlink" href="#creating-backward-variables" title="Permalink to this headline"></a></h3>
<p>Up to now, we have completed all creating and adjusting jobs of <code class="docutils literal"><span class="pre">grad_op</span></code>s. However, backward variables have not been created. Now they are only represented by <code class="docutils literal"><span class="pre">grad_op</span></code>&#8216;s input and output arguments. The backward variable creating job will be done by:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">_append_backward_vars_</span><span class="p">(</span><span class="n">block</span><span class="p">,</span> 
                           <span class="n">start_op_idx</span><span class="p">,</span> 
                           <span class="n">grad_to_var</span><span class="p">,</span> 
                           <span class="n">grad_info_map</span><span class="p">):</span>
    <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">    Create new variables required by backward pass.</span>

<span class="sd">    Args:</span>
<span class="sd">        block(Block): the block where new variables will be created</span>
<span class="sd">        start_op_idx(int): Only variables required by ops in block.ops[start_op_idx : ] will be created</span>
<span class="sd">        grad_to_var(dict):</span>
<span class="sd">            key(str): grad variable name</span>
<span class="sd">            val(str): corresponding forward variable name</span>
<span class="sd">            In most cases, this dict is generated by _append_backward_ops_()</span>
<span class="sd">        grad_info_map(dict)(output argument):</span>
<span class="sd">            key(str): forward variable name</span>
<span class="sd">            val(tuple): a tuple of (str, int), str is the corresponding grad name, int is the block index</span>
<span class="sd">    &quot;&quot;&quot;</span>
</pre></div>
</div>
<p>Given a <code class="docutils literal"><span class="pre">block</span></code>, this function traverses all the <code class="docutils literal"><span class="pre">grad_op</span></code>s in it(The argument <code class="docutils literal"><span class="pre">start_op_idx</span></code> indicates where the grad_op sequence starts.) and creates all the uncreated outputs. The <em>pseudo-code</em> shows this process:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span>for op in block.ops[start_op_idx : ]:

    if op has an attribute named &#39;sub_block&#39;:
        Get the sub-block(`s_block`) from op&#39;s attribute.
        Invoke _append_backward_vars_(), with `block=s_block`
        
    for var_name in op.all_output_names():
        if block.has_var_recursive(var_name) or var_name is the name of empty variable:
            continue
        create a new variable named &#39;var_name&#39; in block
        if grad_to_var.has_key(var_name):
            set grad_info_map[grad_to_var[var_name]] as a tuple of (var_name. block)
            
    do op&#39;s var type inference
    do op&#39;s shape inference
</pre></div>
</div>
</div>
</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>