Skip to content

P
PaddleSlim

PaddlePaddle / PaddleSlim
1 年多 前同步成功

通知 51
Star 1434
Fork 344
P
PaddleSlim
切换分支/标签
查找文件 普通视图历史永久链接Permalink
index.html 49.1 KB
Newer Older
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
1 2 3 4 5 6 7 8 9 10 
<!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 http-equiv="X-UA-Compatible" content="IE=edge">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  
  
  <link rel="shortcut icon" href="../../img/favicon.ico">
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
11 
  <title>知识蒸馏 - PaddleSlim Docs</title>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
12 13 14 15 16 17 18 19 
  <link href='https://fonts.googleapis.com/css?family=Lato:400,700|Roboto+Slab:400,700|Inconsolata:400,700' rel='stylesheet' type='text/css'>

  <link rel="stylesheet" href="../../css/theme.css" type="text/css" />
  <link rel="stylesheet" href="../../css/theme_extra.css" type="text/css" />
  <link rel="stylesheet" href="//cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css">
  
  <script>
    // Current page data
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
20 
    var mkdocs_page_name = "\u77e5\u8bc6\u84b8\u998f";
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
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 
    var mkdocs_page_input_path = "api/single_distiller_api.md";
    var mkdocs_page_url = null;
  </script>
  
  <script src="../../js/jquery-2.1.1.min.js" defer></script>
  <script src="../../js/modernizr-2.8.3.min.js" defer></script>
  <script src="//cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
  <script>hljs.initHighlightingOnLoad();</script> 
  
</head>

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

  <div class="wy-grid-for-nav">

    
    <nav data-toggle="wy-nav-shift" class="wy-nav-side stickynav">
      <div class="wy-side-nav-search">
        <a href="../.." class="icon icon-home"> PaddleSlim Docs</a>
        <div role="search">
  <form id ="rtd-search-form" class="wy-form" action="../../search.html" method="get">
    <input type="text" name="q" placeholder="Search docs" title="Type search term here" />
  </form>
</div>
      </div>

      <div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
	<ul class="current">
	  
          
            <li class="toctree-l1">
		
    <a class="" href="../..">Home</a>
	    </li>
          
            <li class="toctree-l1">
		
    <span class="caption-text">教程</span>
    <ul class="subnav">
                <li class="">
                    
    <a class="" href="../../tutorials/quant_post_demo/">离线量化</a>
                </li>
                <li class="">
                    
    <a class="" href="../../tutorials/quant_aware_demo/">量化训练</a>
                </li>
                <li class="">
                    
    <a class="" href="../../tutorials/quant_embedding_demo/">Embedding量化</a>
                </li>
                <li class="">
                    
    <a class="" href="../../tutorials/nas_demo/">SA搜索</a>
                </li>
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
76 77 78 79 
                <li class="">
                    
    <a class="" href="../../tutorials/distillation_demo/">知识蒸馏</a>
                </li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
80 81 82 83 84 85 86 87 88 89 90 91 92 
    </ul>
	    </li>
          
            <li class="toctree-l1">
		
    <span class="caption-text">API</span>
    <ul class="subnav">
                <li class="">
                    
    <a class="" href="../quantization_api/">量化</a>
                </li>
                <li class="">
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
93 
    <a class="" href="../prune_api/">剪枝与敏感度</a>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
94 95 96 
                </li>
                <li class="">
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
97 
    <a class="" href="../analysis_api/">模型分析</a>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
98 99 100 
                </li>
                <li class=" current">
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
101 
    <a class="current" href="./">知识蒸馏</a>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
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 
    <ul class="subnav">
            
    <li class="toctree-l3"><a href="#merge">merge</a></li>
    

    <li class="toctree-l3"><a href="#fsp_loss">fsp_loss</a></li>
    

    <li class="toctree-l3"><a href="#l2_loss">l2_loss</a></li>
    

    <li class="toctree-l3"><a href="#soft_label_loss">soft_label_loss</a></li>
    

    <li class="toctree-l3"><a href="#loss">loss</a></li>
    

    </ul>
                </li>
                <li class="">
                    
    <a class="" href="../nas_api/">SA搜索</a>
                </li>
                <li class="">
                    
    <a class="" href="../search_space/">搜索空间</a>
                </li>
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
129 130 131 132 
                <li class="">
                    
    <a class="" href="../../table_latency/">硬件延时评估表</a>
                </li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
133 134 135 
    </ul>
	    </li>
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
136 137 138 139 140 
            <li class="toctree-l1">
		
    <a class="" href="../../algo/algo/">算法原理</a>
	    </li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
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 
        </ul>
      </div>
      &nbsp;
    </nav>

    <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap">

      
      <nav class="wy-nav-top" role="navigation" aria-label="top navigation">
        <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
        <a href="../..">PaddleSlim Docs</a>
      </nav>

      
      <div class="wy-nav-content">
        <div class="rst-content">
          <div role="navigation" aria-label="breadcrumbs navigation">
  <ul class="wy-breadcrumbs">
    <li><a href="../..">Docs</a> &raquo;</li>
    
      
        
          <li>API &raquo;</li>
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
167 
    <li>知识蒸馏</li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
168 169 170 171 172 173 174 175 176 177 178 179 180 181 
    <li class="wy-breadcrumbs-aside">
      
        <a href="https://github.com/PaddlePaddle/PaddleSlim/edit/master/docs/api/single_distiller_api.md"
          class="icon icon-github"> Edit on GitHub</a>
      
    </li>
  </ul>
  <hr/>
</div>
          <div role="main">
            <div class="section">
              
                <h2 id="merge">merge<a class="headerlink" href="#merge" title="Permanent link">#</a></h2>
<dl>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
182 
<dt>paddleslim.dist.merge(teacher_program, student_program, data_name_map, place, scope=fluid.global_scope(), name_prefix='teacher_') <a href="https://github.com/PaddlePaddle/PaddleSlim/blob/develop/paddleslim/dist/single_distiller.py#L19">[源代码]</a> </dt>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
183 
<dd>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
184 
<p>merge将两个paddle program（teacher_program, student_program）融合为一个program，并将融合得到的program返回。在融合的program中，可以为其中合适的teacher特征图和student特征图添加蒸馏损失函数，从而达到用teacher模型的暗知识（Dark Knowledge）指导student模型学习的目的。</p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
185 186 187 188 
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
B
Deployed c953571 with MkDocs version: 1.0.4  
baiyfbupt 已提交
189 190 
<li><strong>teacher_program</strong>(Program)-定义了teacher模型的 <a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/api_cn/fluid_cn/Program_cn.html#program"><em>paddle program</em></a></li>
<li><strong>student_program</strong>(Program)-定义了student模型的 <a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/api_cn/fluid_cn/Program_cn.html#program"><em>paddle program</em></a></li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
191 
<li><strong>data_name_map</strong>(dict)-teacher输入接口名与student输入接口名的映射，其中dict的 <em>key</em> 为teacher的输入名，<em>value</em> 为student的输入名</li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
192 
<li><strong>place</strong>(fluid.CPUPlace()|fluid.CUDAPlace(N))-该参数表示程序运行在何种设备上，这里的N为GPU对应的ID</li>
B
Deployed c953571 with MkDocs version: 1.0.4  
baiyfbupt 已提交
193 194 
<li><strong>scope</strong>(Scope)-该参数表示程序使用的变量作用域，如果不指定将使用默认的全局作用域。默认值：<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/api_cn/fluid_cn/global_scope_cn.html#global-scope"><em>fluid.global_scope()</em></a></li>
<li><strong>name_prefix</strong>(str)-merge操作将统一为teacher的<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/1.3/api_guides/low_level/program.html#variable"><em>Variables</em></a>添加的名称前缀name_prefix。默认值：'teacher_'</li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
195 196 
</ul>
<p><strong>返回：</strong> 由student_program和teacher_program merge得到的program</p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
197 198 199 200 
<div class="admonition note">
<p class="admonition-title">Note</p>
<p><em>data_name_map</em> 是 <strong>teacher_var name到student_var name的映射</strong>，如果写反可能无法正确进行merge</p>
</div>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
201 
<p><strong>使用示例：</strong></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span> 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="kn">import</span> <span class="nn">paddle.fluid</span> <span class="kn">as</span> <span class="nn">fluid</span>
<span class="kn">import</span> <span class="nn">paddleslim.dist</span> <span class="kn">as</span> <span class="nn">dist</span>
<span class="n">student_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">student_program</span><span class="p">):</span>
    <span class="n">x</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;x&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
224 225 226 227 228 229 230 231 232 233 234 
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
<span class="n">teacher_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">):</span>
    <span class="n">y</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;y&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">y</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
<span class="n">data_name_map</span> <span class="o">=</span> <span class="p">{</span><span class="s1">&#39;y&#39;</span><span class="p">:</span><span class="s1">&#39;x&#39;</span><span class="p">}</span>
<span class="n">USE_GPU</span> <span class="o">=</span> <span class="bp">False</span>
<span class="n">place</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CUDAPlace</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="k">if</span> <span class="n">USE_GPU</span> <span class="k">else</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CPUPlace</span><span class="p">()</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
235 236 237 238 
<span class="hll"><span class="n">main_program</span> <span class="o">=</span> <span class="n">dist</span><span class="o">.</span><span class="n">merge</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">,</span> <span class="n">student_program</span><span class="p">,</span>
</span><span class="hll">                          <span class="n">data_name_map</span><span class="p">,</span> <span class="n">place</span><span class="p">)</span>
</span></pre></div>
</td></tr></table>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
239 240 241 

<h2 id="fsp_loss">fsp_loss<a class="headerlink" href="#fsp_loss" title="Permanent link">#</a></h2>
<dl>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
242 
<dt>paddleslim.dist.fsp_loss(teacher_var1_name, teacher_var2_name, student_var1_name, student_var2_name, program=fluid.default_main_program()) <a href="https://github.com/PaddlePaddle/PaddleSlim/blob/develop/paddleslim/dist/single_distiller.py#L90">[源代码]</a></dt>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
243 
<dd>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
244 
<p>fsp_loss为program内的teacher var和student var添加fsp loss，出自论文<a href="http://openaccess.thecvf.com/content_cvpr_2017/papers/Yim_A_Gift_From_CVPR_2017_paper.pdf">&lt;&lt;A Gift from Knowledge Distillation: Fast Optimization, Network Minimization and Transfer Learning>></a></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
245 246 247 248 249 250 251 252 
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li><strong>teacher_var1_name</strong>(str): teacher_var1的名称. 对应的variable是一个形为<code>[batch_size, x_channel, height, width]</code>的4-D特征图Tensor，数据类型为float32或float64</li>
<li><strong>teacher_var2_name</strong>(str): teacher_var2的名称. 对应的variable是一个形为<code>[batch_size, y_channel, height, width]</code>的4-D特征图Tensor，数据类型为float32或float64。只有y_channel可以与teacher_var1的x_channel不同，其他维度必须与teacher_var1相同</li>
<li><strong>student_var1_name</strong>(str): student_var1的名称. 对应的variable需与teacher_var1尺寸保持一致，是一个形为<code>[batch_size, x_channel, height, width]</code>的4-D特征图Tensor，数据类型为float32或float64</li>
<li><strong>student_var2_name</strong>(str): student_var2的名称. 对应的variable需与teacher_var2尺寸保持一致，是一个形为<code>[batch_size, y_channel, height, width]</code>的4-D特征图Tensor，数据类型为float32或float64。只有y_channel可以与student_var1的x_channel不同，其他维度必须与student_var1相同</li>
B
Deployed c953571 with MkDocs version: 1.0.4  
baiyfbupt 已提交
253 
<li><strong>program</strong>(Program): 用于蒸馏训练的fluid program。默认值：<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/1.3/api_cn/fluid_cn.html#default-main-program"><em>fluid.default_main_program()</em></a></li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
254 
</ul>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
255 
<p><strong>返回：</strong> 由teacher_var1, teacher_var2, student_var1, student_var2组合得到的fsp_loss</p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
256 
<p><strong>使用示例：</strong></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span> 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="kn">import</span> <span class="nn">paddle.fluid</span> <span class="kn">as</span> <span class="nn">fluid</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 
<span class="kn">import</span> <span class="nn">paddleslim.dist</span> <span class="kn">as</span> <span class="nn">dist</span>
<span class="n">student_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">student_program</span><span class="p">):</span>
    <span class="n">x</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;x&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s1&#39;</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s2&#39;</span><span class="p">)</span>
<span class="n">teacher_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">):</span>
    <span class="n">y</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;y&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">y</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t1&#39;</span><span class="p">)</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t2&#39;</span><span class="p">)</span>
<span class="n">data_name_map</span> <span class="o">=</span> <span class="p">{</span><span class="s1">&#39;y&#39;</span><span class="p">:</span><span class="s1">&#39;x&#39;</span><span class="p">}</span>
<span class="n">USE_GPU</span> <span class="o">=</span> <span class="bp">False</span>
<span class="n">place</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CUDAPlace</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="k">if</span> <span class="n">USE_GPU</span> <span class="k">else</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CPUPlace</span><span class="p">()</span>
<span class="n">main_program</span> <span class="o">=</span> <span class="n">merge</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">,</span> <span class="n">student_program</span><span class="p">,</span> <span class="n">data_name_map</span><span class="p">,</span> <span class="n">place</span><span class="p">)</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">main_program</span><span class="p">):</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
294 295 296 297 
<span class="hll">    <span class="n">distillation_loss</span> <span class="o">=</span> <span class="n">dist</span><span class="o">.</span><span class="n">fsp_loss</span><span class="p">(</span><span class="s1">&#39;teacher_t1.tmp_1&#39;</span><span class="p">,</span> <span class="s1">&#39;teacher_t2.tmp_1&#39;</span><span class="p">,</span>
</span><span class="hll">                                      <span class="s1">&#39;s1.tmp_1&#39;</span><span class="p">,</span> <span class="s1">&#39;s2.tmp_1&#39;</span><span class="p">,</span> <span class="n">main_program</span><span class="p">)</span>
</span></pre></div>
</td></tr></table>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
298 299 300 

<h2 id="l2_loss">l2_loss<a class="headerlink" href="#l2_loss" title="Permanent link">#</a></h2>
<dl>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
301 
<dt>paddleslim.dist.l2_loss(teacher_var_name, student_var_name, program=fluid.default_main_program())<a href="https://github.com/PaddlePaddle/PaddleSlim/blob/develop/paddleslim/dist/single_distiller.py#L118">[源代码]</a></dt>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
302 303 304 305 306 307 308 309 
<dd>
<p>l2_loss为program内的teacher var和student var添加l2 loss</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li><strong>teacher_var_name</strong>(str): teacher_var的名称. </li>
<li><strong>student_var_name</strong>(str): student_var的名称.</li>
B
Deployed c953571 with MkDocs version: 1.0.4  
baiyfbupt 已提交
310 
<li><strong>program</strong>(Program): 用于蒸馏训练的fluid program。默认值：<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/1.3/api_cn/fluid_cn.html#default-main-program"><em>fluid.default_main_program()</em></a></li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
311 
</ul>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
312 
<p><strong>返回：</strong> 由teacher_var, student_var组合得到的l2_loss</p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
313 
<p><strong>使用示例：</strong></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span> 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="kn">import</span> <span class="nn">paddle.fluid</span> <span class="kn">as</span> <span class="nn">fluid</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 
<span class="kn">import</span> <span class="nn">paddleslim.dist</span> <span class="kn">as</span> <span class="nn">dist</span>
<span class="n">student_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">student_program</span><span class="p">):</span>
    <span class="n">x</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;x&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s1&#39;</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s2&#39;</span><span class="p">)</span>
<span class="n">teacher_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">):</span>
    <span class="n">y</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;y&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">y</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t1&#39;</span><span class="p">)</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t2&#39;</span><span class="p">)</span>
<span class="n">data_name_map</span> <span class="o">=</span> <span class="p">{</span><span class="s1">&#39;y&#39;</span><span class="p">:</span><span class="s1">&#39;x&#39;</span><span class="p">}</span>
<span class="n">USE_GPU</span> <span class="o">=</span> <span class="bp">False</span>
<span class="n">place</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CUDAPlace</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="k">if</span> <span class="n">USE_GPU</span> <span class="k">else</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CPUPlace</span><span class="p">()</span>
<span class="n">main_program</span> <span class="o">=</span> <span class="n">merge</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">,</span> <span class="n">student_program</span><span class="p">,</span> <span class="n">data_name_map</span><span class="p">,</span> <span class="n">place</span><span class="p">)</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">main_program</span><span class="p">):</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
351 352 353 354 
<span class="hll">    <span class="n">distillation_loss</span> <span class="o">=</span> <span class="n">dist</span><span class="o">.</span><span class="n">l2_loss</span><span class="p">(</span><span class="s1">&#39;teacher_t2.tmp_1&#39;</span><span class="p">,</span> <span class="s1">&#39;s2.tmp_1&#39;</span><span class="p">,</span>
</span><span class="hll">                                     <span class="n">main_program</span><span class="p">)</span>
</span></pre></div>
</td></tr></table>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
355 356 357 

<h2 id="soft_label_loss">soft_label_loss<a class="headerlink" href="#soft_label_loss" title="Permanent link">#</a></h2>
<dl>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
358 
<dt>paddleslim.dist.soft_label_loss(teacher_var_name, student_var_name, program=fluid.default_main_program(), teacher_temperature=1., student_temperature=1.)<a href="https://github.com/PaddlePaddle/PaddleSlim/blob/develop/paddleslim/dist/single_distiller.py#L136">[源代码]</a></dt>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
359 
<dd>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
360 
<p>soft_label_loss为program内的teacher var和student var添加soft label loss，出自论文<a href="https://arxiv.org/pdf/1503.02531.pdf">&lt;&lt;Distilling the Knowledge in a Neural Network>></a></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
361 362 363 364 365 366 
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li><strong>teacher_var_name</strong>(str): teacher_var的名称. </li>
<li><strong>student_var_name</strong>(str): student_var的名称. </li>
B
Deployed c953571 with MkDocs version: 1.0.4  
baiyfbupt 已提交
367 
<li><strong>program</strong>(Program): 用于蒸馏训练的fluid program。默认值：<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/1.3/api_cn/fluid_cn.html#default-main-program"><em>fluid.default_main_program()</em></a></li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
368 369 370 
<li><strong>teacher_temperature</strong>(float): 对teacher_var进行soft操作的温度值，温度值越大得到的特征图越平滑 </li>
<li><strong>student_temperature</strong>(float): 对student_var进行soft操作的温度值，温度值越大得到的特征图越平滑 </li>
</ul>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
371 
<p><strong>返回：</strong> 由teacher_var, student_var组合得到的soft_label_loss</p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
372 
<p><strong>使用示例：</strong></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span> 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="kn">import</span> <span class="nn">paddle.fluid</span> <span class="kn">as</span> <span class="nn">fluid</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 
<span class="kn">import</span> <span class="nn">paddleslim.dist</span> <span class="kn">as</span> <span class="nn">dist</span>
<span class="n">student_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">student_program</span><span class="p">):</span>
    <span class="n">x</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;x&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s1&#39;</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s2&#39;</span><span class="p">)</span>
<span class="n">teacher_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">):</span>
    <span class="n">y</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;y&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">y</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t1&#39;</span><span class="p">)</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t2&#39;</span><span class="p">)</span>
<span class="n">data_name_map</span> <span class="o">=</span> <span class="p">{</span><span class="s1">&#39;y&#39;</span><span class="p">:</span><span class="s1">&#39;x&#39;</span><span class="p">}</span>
<span class="n">USE_GPU</span> <span class="o">=</span> <span class="bp">False</span>
<span class="n">place</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CUDAPlace</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="k">if</span> <span class="n">USE_GPU</span> <span class="k">else</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CPUPlace</span><span class="p">()</span>
<span class="n">main_program</span> <span class="o">=</span> <span class="n">merge</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">,</span> <span class="n">student_program</span><span class="p">,</span> <span class="n">data_name_map</span><span class="p">,</span> <span class="n">place</span><span class="p">)</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">main_program</span><span class="p">):</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
410 411 412 413 
<span class="hll">    <span class="n">distillation_loss</span> <span class="o">=</span> <span class="n">dist</span><span class="o">.</span><span class="n">soft_label_loss</span><span class="p">(</span><span class="s1">&#39;teacher_t2.tmp_1&#39;</span><span class="p">,</span>
</span><span class="hll">                                             <span class="s1">&#39;s2.tmp_1&#39;</span><span class="p">,</span> <span class="n">main_program</span><span class="p">,</span> <span class="mf">1.</span><span class="p">,</span> <span class="mf">1.</span><span class="p">)</span>
</span></pre></div>
</td></tr></table>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
414 415 416 

<h2 id="loss">loss<a class="headerlink" href="#loss" title="Permanent link">#</a></h2>
<dl>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
417 
<dt>paddleslim.dist.loss(loss_func, program=fluid.default_main_program(), **kwargs) <a href="https://github.com/PaddlePaddle/PaddleSlim/blob/develop/paddleslim/dist/single_distiller.py#L165">[源代码]</a></dt>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
418 419 420 421 422 423 424 
<dd>
<p>loss函数支持对任意多对teacher_var和student_var使用自定义损失函数</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li><strong>loss_func</strong>(python function): 自定义的损失函数，输入为teacher var和student var，输出为自定义的loss </li>
B
Deployed c953571 with MkDocs version: 1.0.4  
baiyfbupt 已提交
425 
<li><strong>program</strong>(Program): 用于蒸馏训练的fluid program。默认值：<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/1.3/api_cn/fluid_cn.html#default-main-program"><em>fluid.default_main_program()</em></a></li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
426 427 428 429 
<li><strong>**kwargs</strong>: loss_func输入名与对应variable名称</li>
</ul>
<p><strong>返回</strong>：自定义的损失函数loss</p>
<p><strong>使用示例：</strong></p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre><span></span> 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</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="kn">import</span> <span class="nn">paddle.fluid</span> <span class="kn">as</span> <span class="nn">fluid</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 
<span class="kn">import</span> <span class="nn">paddleslim.dist</span> <span class="kn">as</span> <span class="nn">dist</span>
<span class="n">student_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">student_program</span><span class="p">):</span>
    <span class="n">x</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;x&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s1&#39;</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;s2&#39;</span><span class="p">)</span>
<span class="n">teacher_program</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">Program</span><span class="p">()</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">):</span>
    <span class="n">y</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;y&#39;</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">])</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">y</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t1&#39;</span><span class="p">)</span>
    <span class="n">conv</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
    <span class="n">out</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">conv</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="n">padding</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">&#39;t2&#39;</span><span class="p">)</span>
<span class="n">data_name_map</span> <span class="o">=</span> <span class="p">{</span><span class="s1">&#39;y&#39;</span><span class="p">:</span><span class="s1">&#39;x&#39;</span><span class="p">}</span>
<span class="n">USE_GPU</span> <span class="o">=</span> <span class="bp">False</span>
<span class="n">place</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CUDAPlace</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="k">if</span> <span class="n">USE_GPU</span> <span class="k">else</span> <span class="n">fluid</span><span class="o">.</span><span class="n">CPUPlace</span><span class="p">()</span>
<span class="n">main_program</span> <span class="o">=</span> <span class="n">merge</span><span class="p">(</span><span class="n">teacher_program</span><span class="p">,</span> <span class="n">student_program</span><span class="p">,</span> <span class="n">data_name_map</span><span class="p">,</span> <span class="n">place</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">adaptation_loss</span><span class="p">(</span><span class="n">t_var</span><span class="p">,</span> <span class="n">s_var</span><span class="p">):</span>
    <span class="n">teacher_channel</span> <span class="o">=</span> <span class="n">t_var</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span>
    <span class="n">s_hint</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">conv2d</span><span class="p">(</span><span class="n">s_var</span><span class="p">,</span> <span class="n">teacher_channel</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
    <span class="n">hint_loss</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">reduce_mean</span><span class="p">(</span><span class="n">fluid</span><span class="o">.</span><span class="n">layers</span><span class="o">.</span><span class="n">square</span><span class="p">(</span><span class="n">s_hint</span> <span class="o">-</span> <span class="n">t_var</span><span class="p">))</span>
    <span class="k">return</span> <span class="n">hint_loss</span>
<span class="k">with</span> <span class="n">fluid</span><span class="o">.</span><span class="n">program_guard</span><span class="p">(</span><span class="n">main_program</span><span class="p">):</span>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
477 478 479 480 
<span class="hll">    <span class="n">distillation_loss</span> <span class="o">=</span> <span class="n">dist</span><span class="o">.</span><span class="n">loss</span><span class="p">(</span><span class="n">main_program</span><span class="p">,</span> <span class="n">adaptation_loss</span><span class="p">,</span>
</span><span class="hll">            <span class="n">t_var</span><span class="o">=</span><span class="s1">&#39;teacher_t2.tmp_1&#39;</span><span class="p">,</span> <span class="n">s_var</span><span class="o">=</span><span class="s1">&#39;s2.tmp_1&#39;</span><span class="p">)</span>
</span></pre></div>
</td></tr></table>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
481
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
482 483 
<div class="admonition note">
<p class="admonition-title">注意事项</p>
B
Deployed 76338d9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
484 
<p>在添加蒸馏loss时会引入新的variable，需要注意新引入的variable不要与student variables命名冲突。这里建议两种用法（两种方法任选其一即可）：</p>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
485 486 
<ol>
<li>建议与student_program使用同一个命名空间，以避免一些未指定名称的variables(例如tmp_0, tmp_1...)多次定义为同一名称出现命名冲突</li>
B
Deployed 76338d9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
487 
<li>建议在添加蒸馏loss时指定一个命名空间前缀，具体用法请参考Paddle官方文档<a href="https://www.paddlepaddle.org.cn/documentation/docs/zh/api_cn/fluid_cn/name_scope_cn.html#name-scope"><em>fluid.name_scope</em></a></li>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
488 
</ol>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
489 
</div>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
490 491 492 493 494 495 496 497 498 499 
              
            </div>
          </div>
          <footer>
  
    <div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
      
        <a href="../nas_api/" class="btn btn-neutral float-right" title="SA搜索">Next <span class="icon icon-circle-arrow-right"></span></a>
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
500 
        <a href="../analysis_api/" class="btn btn-neutral" title="模型分析"><span class="icon icon-circle-arrow-left"></span> Previous</a>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 
      
    </div>
  

  <hr/>

  <div role="contentinfo">
    <!-- Copyright etc -->
    
  </div>

  Built with <a href="http://www.mkdocs.org">MkDocs</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>

  <div class="rst-versions" role="note" style="cursor: pointer">
    <span class="rst-current-version" data-toggle="rst-current-version">
      
          <a href="https://github.com/PaddlePaddle/PaddleSlim/" class="fa fa-github" style="float: left; color: #fcfcfc"> GitHub</a>
      
      
        <span><a href="../analysis_api/" style="color: #fcfcfc;">&laquo; Previous</a></span>
      
      
        <span style="margin-left: 15px"><a href="../nas_api/" style="color: #fcfcfc">Next &raquo;</a></span>
      
    </span>
</div>
    <script>var base_url = '../..';</script>
    <script src="../../js/theme.js" defer></script>
      <script src="../../mathjax-config.js" defer></script>
B
Deployed 0f9c3ce with MkDocs version: 1.0.4  
baiyfbupt 已提交
538 
      <script src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS-MML_HTMLorMML" defer></script>
B
Deployed 09f99c9 with MkDocs version: 1.0.4  
baiyfbupt 已提交
539 540 541 542 
      <script src="../../search/main.js" defer></script>

</body>
</html>