Skip to content

P
PaddleSlim

PaddlePaddle / PaddleSlim
大约 1 年 前同步成功

通知 51
Star 1434
Fork 344
P
PaddleSlim
提交 9bc13c0f 编写于 作者: W wanghaoshuang
浏览文件
操作

Deployed 42bbcb15 with MkDocs version: 1.0.4

上级 ce4dc254
展开全部 隐藏空白更改
内联 并排
Showing with 1718 addition and 60 deletion
404.html
浏览文件 @ 9bc13c0f
......@@ -78,6 +78,10 @@
</li>
<li class="">
<a class="" href="/tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="/search_space/">搜索空间</a>
</li>
<li class="">
......@@ -105,7 +109,11 @@
</li>
<li class="">
<a class="" href="/api/single_distiller_api/">知识蒸馏</a>
<a class="" href="/api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="/api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -113,6 +121,14 @@
</li>
<li class="">
<a class="" href="/api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="/search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="/table_latency/">硬件延时评估表</a>
</li>
</ul>
......
algo/algo/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,7 +116,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -120,6 +128,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
api/analysis_api/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -124,7 +128,11 @@
</li>
<li class="">
<a class="" href="../single_distiller_api/">知识蒸馏</a>
<a class="" href="../single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -132,6 +140,14 @@
</li>
<li class="">
<a class="" href="../one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -365,7 +381,7 @@
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../single_distiller_api/" class="btn btn-neutral float-right" title="知识蒸馏">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../single_distiller_api/" class="btn btn-neutral float-right" title="简单知识蒸馏">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../prune_api/" class="btn btn-neutral" title="剪枝与敏感度"><span class="icon icon-circle-arrow-left"></span> Previous</a>
......
api/api_guide/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,7 +116,11 @@
</li>
<li class="">
<a class="" href="../single_distiller_api/">知识蒸馏</a>
<a class="" href="../single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -120,6 +128,14 @@
</li>
<li class="">
<a class="" href="../one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -174,6 +190,9 @@
<p><a href="../single_distiller_api/">单进程蒸馏</a></p>
</li>
<li>
<p><a href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a></p>
</li>
<li>
<p><a href="../prune_api/">通道剪裁</a></p>
</li>
</ul>
......
api/nas_api/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,7 +116,11 @@
</li>
<li class="">
<a class="" href="../single_distiller_api/">知识蒸馏</a>
<a class="" href="../single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class=" current">
......@@ -129,6 +137,14 @@
</li>
<li class="">
<a class="" href="../one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -312,10 +328,10 @@
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../../table_latency/" class="btn btn-neutral float-right" title="硬件延时评估表">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../one_shot_api/" class="btn btn-neutral float-right" title="One-shot搜索">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../single_distiller_api/" class="btn btn-neutral" title="知识蒸馏"><span class="icon icon-circle-arrow-left"></span> Previous</a>
<a href="../pantheon_api/" class="btn btn-neutral" title="大规模可扩展知识蒸馏框架 Pantheon"><span class="icon icon-circle-arrow-left"></span> Previous</a>
</div>
......@@ -343,10 +359,10 @@
<a href="https://github.com/PaddlePaddle/PaddleSlim/" class="fa fa-github" style="float: left; color: #fcfcfc"> GitHub</a>
<span><a href="../single_distiller_api/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span><a href="../pantheon_api/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span style="margin-left: 15px"><a href="../../table_latency/" style="color: #fcfcfc">Next &raquo;</a></span>
<span style="margin-left: 15px"><a href="../one_shot_api/" style="color: #fcfcfc">Next &raquo;</a></span>
</span>
</div>
......
api/one_shot_api/index.html 0 → 100644
浏览文件 @ 9bc13c0f
<!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">
<title>One-shot搜索 - PaddleSlim Docs</title>
<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">
<link href="../../extra.css" rel="stylesheet">
<script>
// Current page data
var mkdocs_page_name = "One-shot\u641c\u7d22";
var mkdocs_page_input_path = "api/one_shot_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">
<a class="" href="../../model_zoo/">模型库</a>
</li>
<li class="toctree-l1">
<span class="caption-text">教程</span>
<ul class="subnav">
<li class="">
<a class="" href="../../tutorials/pruning_tutorial/">图像分类模型通道剪裁-快速开始</a>
</li>
<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>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../tutorials/distillation_demo/">知识蒸馏</a>
</li>
</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="">
<a class="" href="../prune_api/">剪枝与敏感度</a>
</li>
<li class="">
<a class="" href="../analysis_api/">模型分析</a>
</li>
<li class="">
<a class="" href="../single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
<a class="" href="../nas_api/">SA搜索</a>
</li>
<li class=" current">
<a class="current" href="./">One-shot搜索</a>
<ul class="subnav">
<li class="toctree-l3"><a href="#oneshotsearch">OneShotSearch</a></li>
<li class="toctree-l3"><a href="#oneshotsupernet">OneShotSuperNet</a></li>
<li class="toctree-l3"><a href="#supermnasnet">SuperMnasnet</a></li>
</ul>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
</li>
<li class="toctree-l1">
<a class="" href="../../algo/algo/">算法原理</a>
</li>
</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>
<li>One-shot搜索</li>
<li class="wy-breadcrumbs-aside">
<a href="https://github.com/PaddlePaddle/PaddleSlim/edit/master/docs/api/one_shot_api.md"
class="icon icon-github"> Edit on GitHub</a>
</li>
</ul>
<hr/>
</div>
<div role="main">
<div class="section">
<h2 id="oneshotsearch">OneShotSearch<a class="headerlink" href="#oneshotsearch" title="Permanent link">#</a></h2>
<dl>
<dt>paddleslim.nas.one_shot.OneShotSearch(model, eval_func, strategy='sa', search_steps=100)<a href="">代码</a></dt>
<dd>
<p>从超级网络中搜索出一个最佳的子网络。</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li>
<p><strong>model(fluid.dygraph.layer):</strong> 通过在<code>OneShotSuperNet</code>前后添加若该模块构建的动态图模块。因为<code>OneShotSuperNet</code>是一个超网络，所以<code>model</code>也是一个超网络。换句话说，在<code>model</code>模块的子模块中，至少有一个是<code>OneShotSuperNet</code>的实例。该方法从<code>model</code>超网络中搜索得到一个最佳的子网络。超网络<code>model</code>需要先被训练，具体细节请参考<a href="">OneShotSuperNet</a>。</p>
</li>
<li>
<p><strong>eval_func:</strong> 用于评估子网络性能的回调函数。该回调函数需要接受<code>model</code>为参数，并调用<code>model</code>的<code>forward</code>方法进行性能评估。</p>
</li>
<li>
<p><strong>strategy(str):</strong> 搜索策略的名称。默认为'sa'， 当前仅支持'sa'.</p>
</li>
<li>
<p><strong>search_steps(int):</strong> 搜索轮次数。默认为100。</p>
</li>
</ul>
<p><strong>返回：</strong></p>
<ul>
<li><strong>best_tokens:</strong> 表示最佳子网络的编码信息（tokens）。</li>
</ul>
<p><strong>示例代码：</strong></p>
<p>请参考<a href="">one-shot NAS示例</a></p>
<h2 id="oneshotsupernet">OneShotSuperNet<a class="headerlink" href="#oneshotsupernet" title="Permanent link">#</a></h2>
<p>用于<code>OneShot</code>搜索策略的超级网络的基类，所有超级网络的实现要继承该类。</p>
<dl>
<dt>paddleslim.nas.one_shot.OneShotSuperNet(name_scope)</dt>
<dd>
<p>构造方法。</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li>**name_scope:(str) **超级网络的命名空间。</li>
</ul>
<p><strong>返回：</strong></p>
<ul>
<li><strong>super_net:</strong> 一个<code>OneShotSuperNet</code>实例。</li>
</ul>
<dl>
<dt>init_tokens()</dt>
<dd>
<p>获得当前超级网络的初始化子网络的编码，主要用于搜索。</p>
</dd>
</dl>
<p><strong>返回：</strong></p>
<ul>
<li><strong>tokens(list<int>):</strong> 一个子网络的编码。</li>
</ul>
<dl>
<dt>range_table()</dt>
<dd>
<p>超级网络中各个子网络由一组整型数字编码表示，该方法返回编码每个位置的取值范围。</p>
</dd>
</dl>
<p><strong>返回：</strong></p>
<ul>
<li><strong>range_table(tuple):</strong> 子网络编码每一位的取值范围。<code>range_table</code>格式为<code>(min_values, max_values)</code>，其中，<code>min_values</code>为一个整型数组，表示每个编码位置可选取的最小值；<code>max_values</code>表示每个编码位置可选取的最大值。</li>
</ul>
<dl>
<dt>_forward_impl(input, tokens)</dt>
<dd>
<p>前向计算函数。<code>OneShotSuperNet</code>的子类需要实现该函数。</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li>
<p><strong>input(Variable):</strong> 超级网络的输入。</p>
</li>
<li>
<p><strong>tokens(list<int>):</strong> 执行前向计算所用的子网络的编码。默认为<code>None</code>，即随机选取一个子网络执行前向。</p>
</li>
</ul>
<p><strong>返回：</strong></p>
<ul>
<li><strong>output(Variable):</strong> 前向计算的输出</li>
</ul>
<dl>
<dt>forward(self, input, tokens=None)</dt>
<dd>
<p>执行前向计算。</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li>
<p><strong>input(Variable):</strong> 超级网络的输入。</p>
</li>
<li>
<p><strong>tokens(list<int>):</strong> 执行前向计算所用的子网络的编码。默认为<code>None</code>，即随机选取一个子网络执行前向。</p>
</li>
</ul>
<p><strong>返回：</strong></p>
<ul>
<li><strong>output(Variable):</strong> 前向计算的输出</li>
</ul>
<dl>
<dt>_random_tokens()</dt>
<dd>
<p>随机选取一个子网络，并返回其编码。</p>
</dd>
</dl>
<p><strong>返回：</strong></p>
<ul>
<li><strong>tokens(list<int>):</strong> 一个子网络的编码。</li>
</ul>
<h2 id="supermnasnet">SuperMnasnet<a class="headerlink" href="#supermnasnet" title="Permanent link">#</a></h2>
<p>在<a href="https://arxiv.org/abs/1807.11626">Mnasnet</a>基础上修改得到的超级网络, 该类继承自<code>OneShotSuperNet</code>.</p>
<dl>
<dt>paddleslim.nas.one_shot.SuperMnasnet(name_scope, input_channels=3, out_channels=1280, repeat_times=[6, 6, 6, 6, 6, 6], stride=[1, 1, 1, 1, 2, 1], channels=[16, 24, 40, 80, 96, 192, 320], use_auxhead=False)</dt>
<dd>
<p>构造函数。</p>
</dd>
</dl>
<p><strong>参数：</strong></p>
<ul>
<li>
<p><strong>name_scope(str):</strong> 命名空间。</p>
</li>
<li>
<p><strong>input_channels(str):</strong> 当前超级网络的输入的特征图的通道数量。</p>
</li>
<li>
<p><strong>out_channels(str):</strong> 当前超级网络的输出的特征图的通道数量。</p>
</li>
<li>
<p><strong>repeat_times(list):</strong> 每种<code>block</code>重复的次数。</p>
</li>
<li>
<p><strong>stride(list):</strong> 一种<code>block</code>重复堆叠成<code>repeat_block</code>，<code>stride</code>表示每个<code>repeat_block</code>的下采样比例。</p>
</li>
<li>
<p><strong>channels(list):</strong> channels[i]和channels[i+1]分别表示第i个<code>repeat_block</code>的输入特征图的通道数和输出特征图的通道数。</p>
</li>
<li>
<p><strong>use_auxhead(bool):</strong> 是否使用辅助特征图。如果设置为<code>True</code>，则<code>SuperMnasnet</code>除了返回输出特征图，还还返回辅助特征图。默认为False.</p>
</li>
</ul>
<p><strong>返回：</strong></p>
<ul>
<li><strong>instance(SuperMnasnet):</strong> 一个<code>SuperMnasnet</code>实例</li>
</ul>
<p><strong>示例：</strong>
<div class="codehilite"><pre><span></span><span class="kn">import</span> <span class="nn">paddle</span>
<span class="kn">import</span> <span class="nn">paddle.fluid</span> <span class="kn">as</span> <span class="nn">fluid</span>
<span class="k">class</span> <span class="nc">MNIST</span><span class="p">(</span><span class="n">fluid</span><span class="o">.</span><span class="n">dygraph</span><span class="o">.</span><span class="n">Layer</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="nb">super</span><span class="p">(</span><span class="n">MNIST</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">arch</span> <span class="o">=</span> <span class="n">SuperMnasnet</span><span class="p">(</span>
<span class="n">name_scope</span><span class="o">=</span><span class="s2">&quot;super_net&quot;</span><span class="p">,</span> <span class="n">input_channels</span><span class="o">=</span><span class="mi">20</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">20</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">pool_2_shape</span> <span class="o">=</span> <span class="mi">50</span> <span class="o">*</span> <span class="mi">13</span> <span class="o">*</span> <span class="mi">13</span>
<span class="n">SIZE</span> <span class="o">=</span> <span class="mi">10</span>
<span class="n">scale</span> <span class="o">=</span> <span class="p">(</span><span class="mf">2.0</span> <span class="o">/</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">pool_2_shape</span><span class="o">**</span><span class="mi">2</span> <span class="o">*</span> <span class="n">SIZE</span><span class="p">))</span><span class="o">**</span><span class="mf">0.5</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_fc</span> <span class="o">=</span> <span class="n">Linear</span><span class="p">(</span>
<span class="bp">self</span><span class="o">.</span><span class="n">pool_2_shape</span><span class="p">,</span>
<span class="mi">10</span><span class="p">,</span>
<span class="n">param_attr</span><span class="o">=</span><span class="n">fluid</span><span class="o">.</span><span class="n">param_attr</span><span class="o">.</span><span class="n">ParamAttr</span><span class="p">(</span>
<span class="n">initializer</span><span class="o">=</span><span class="n">fluid</span><span class="o">.</span><span class="n">initializer</span><span class="o">.</span><span class="n">NormalInitializer</span><span class="p">(</span>
<span class="n">loc</span><span class="o">=</span><span class="mf">0.0</span><span class="p">,</span> <span class="n">scale</span><span class="o">=</span><span class="n">scale</span><span class="p">)),</span>
<span class="n">act</span><span class="o">=</span><span class="s2">&quot;softmax&quot;</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">inputs</span><span class="p">,</span> <span class="n">label</span><span class="o">=</span><span class="bp">None</span><span class="p">,</span> <span class="n">tokens</span><span class="o">=</span><span class="bp">None</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">arch</span><span class="p">(</span><span class="n">inputs</span><span class="p">,</span> <span class="n">tokens</span><span class="o">=</span><span class="n">tokens</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">reshape</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">shape</span><span class="o">=</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">pool_2_shape</span><span class="p">])</span>
<span class="n">x</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_fc</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">if</span> <span class="n">label</span> <span class="ow">is</span> <span class="ow">not</span> <span class="bp">None</span><span class="p">:</span>
<span class="n">acc</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">accuracy</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">x</span><span class="p">,</span> <span class="n">label</span><span class="o">=</span><span class="n">label</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span><span class="p">,</span> <span class="n">acc</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="n">x</span>
</pre></div></p>
</div>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../../search_space/" class="btn btn-neutral float-right" title="搜索空间">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../nas_api/" class="btn btn-neutral" title="SA搜索"><span class="icon icon-circle-arrow-left"></span> Previous</a>
</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="../nas_api/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span style="margin-left: 15px"><a href="../../search_space/" 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>
<script src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS-MML_HTMLorMML" defer></script>
<script src="../../search/main.js" defer></script>
</body>
</html>
api/pantheon_api/index.html 0 → 100644
浏览文件 @ 9bc13c0f
此差异已折叠。
api/prune_api/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -130,7 +134,11 @@
</li>
<li class="">
<a class="" href="../single_distiller_api/">知识蒸馏</a>
<a class="" href="../single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -138,6 +146,14 @@
</li>
<li class="">
<a class="" href="../one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
api/quantization_api/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -130,7 +134,11 @@
</li>
<li class="">
<a class="" href="../single_distiller_api/">知识蒸馏</a>
<a class="" href="../single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -138,6 +146,14 @@
</li>
<li class="">
<a class="" href="../one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
api/single_distiller_api/index.html
浏览文件 @ 9bc13c0f
......@@ -8,7 +8,7 @@
<link rel="shortcut icon" href="../../img/favicon.ico">
<title>知识蒸馏 - PaddleSlim Docs</title>
<title>简单知识蒸馏 - PaddleSlim Docs</title>
<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" />
......@@ -18,7 +18,7 @@
<script>
// Current page data
var mkdocs_page_name = "\u77e5\u8bc6\u84b8\u998f";
var mkdocs_page_name = "\u7b80\u5355\u77e5\u8bc6\u84b8\u998f";
var mkdocs_page_input_path = "api/single_distiller_api.md";
var mkdocs_page_url = null;
</script>
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,7 +116,7 @@
</li>
<li class=" current">
<a class="current" href="./">知识蒸馏</a>
<a class="current" href="./">简单知识蒸馏</a>
<ul class="subnav">
<li class="toctree-l3"><a href="#merge">merge</a></li>
......@@ -134,10 +138,22 @@
</li>
<li class="">
<a class="" href="../pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
<a class="" href="../nas_api/">SA搜索</a>
</li>
<li class="">
<a class="" href="../one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -174,7 +190,7 @@
<li>知识蒸馏</li>
<li>简单知识蒸馏</li>
<li class="wy-breadcrumbs-aside">
<a href="https://github.com/PaddlePaddle/PaddleSlim/edit/master/docs/api/single_distiller_api.md"
......@@ -401,7 +417,7 @@
<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>
<a href="../pantheon_api/" class="btn btn-neutral float-right" title="大规模可扩展知识蒸馏框架 Pantheon">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../analysis_api/" class="btn btn-neutral" title="模型分析"><span class="icon icon-circle-arrow-left"></span> Previous</a>
......@@ -435,7 +451,7 @@
<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 style="margin-left: 15px"><a href="../pantheon_api/" style="color: #fcfcfc">Next &raquo;</a></span>
</span>
</div>
......
index.html
浏览文件 @ 9bc13c0f
......@@ -103,6 +103,10 @@
</li>
<li class="">
<a class="" href="tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="search_space/">搜索空间</a>
</li>
<li class="">
......@@ -130,7 +134,11 @@
</li>
<li class="">
<a class="" href="api/single_distiller_api/">知识蒸馏</a>
<a class="" href="api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -138,6 +146,14 @@
</li>
<li class="">
<a class="" href="api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -294,5 +310,5 @@ python setup.py install
<!--
MkDocs version : 1.0.4
Build Date UTC : 2020-02-03 06:29:19
Build Date UTC : 2020-02-04 06:35:51
-->
model_zoo/index.html
浏览文件 @ 9bc13c0f
......@@ -125,6 +125,10 @@
</li>
<li class="">
<a class="" href="../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -152,7 +156,11 @@
</li>
<li class="">
<a class="" href="../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -160,6 +168,14 @@
</li>
<li class="">
<a class="" href="../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
search.html
浏览文件 @ 9bc13c0f
......@@ -78,6 +78,10 @@
</li>
<li class="">
<a class="" href="./tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="./search_space/">搜索空间</a>
</li>
<li class="">
......@@ -105,7 +109,11 @@
</li>
<li class="">
<a class="" href="./api/single_distiller_api/">知识蒸馏</a>
<a class="" href="./api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="./api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -113,6 +121,14 @@
</li>
<li class="">
<a class="" href="./api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="./search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="./table_latency/">硬件延时评估表</a>
</li>
</ul>
......
search/search_index.json
浏览文件 @ 9bc13c0f
此差异已折叠。
search_space/index.html
浏览文件 @ 9bc13c0f
......@@ -83,9 +83,13 @@
<a class="" href="../tutorials/nas_demo/">SA搜索</a>
</li>
<li class=" current">
<li class="">
<a class="current" href="./">搜索空间</a>
<a class="" href="../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="./">搜索空间</a>
<ul class="subnav">
<li class="toctree-l3"><a href="#_1">搜索空间简介</a></li>
......@@ -135,7 +139,11 @@
</li>
<li class="">
<a class="" href="../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -143,6 +151,37 @@
</li>
<li class="">
<a class="" href="../api/one_shot_api/">One-shot搜索</a>
</li>
<li class=" current">
<a class="current" href="./">搜索空间</a>
<ul class="subnav">
<li class="toctree-l3"><a href="#_1">搜索空间简介</a></li>
<li class="toctree-l3"><a href="#paddleslimnas">paddleslim.nas 提供的搜索空间</a></li>
<ul>
<li><a class="toctree-l4" href="#_2">根据初始模型结构构造搜索空间</a></li>
<li><a class="toctree-l4" href="#block">根据相应模型的block构造搜索空间</a></li>
</ul>
<li class="toctree-l3"><a href="#_3">搜索空间示例</a></li>
<li class="toctree-l3"><a href="#search-space">自定义搜索空间(search space)</a></li>
</ul>
</li>
<li class="">
<a class="" href="../table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -175,7 +214,7 @@
<li>教程 &raquo;</li>
<li>API &raquo;</li>
......@@ -318,10 +357,10 @@
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../tutorials/distillation_demo/" class="btn btn-neutral float-right" title="知识蒸馏">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../table_latency/" class="btn btn-neutral float-right" title="硬件延时评估表">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../tutorials/nas_demo/" class="btn btn-neutral" title="SA搜索"><span class="icon icon-circle-arrow-left"></span> Previous</a>
<a href="../api/one_shot_api/" class="btn btn-neutral" title="One-shot搜索"><span class="icon icon-circle-arrow-left"></span> Previous</a>
</div>
......@@ -349,10 +388,10 @@
<a href="https://github.com/PaddlePaddle/PaddleSlim/" class="fa fa-github" style="float: left; color: #fcfcfc"> GitHub</a>
<span><a href="../tutorials/nas_demo/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span><a href="../api/one_shot_api/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span style="margin-left: 15px"><a href="../tutorials/distillation_demo/" style="color: #fcfcfc">Next &raquo;</a></span>
<span style="margin-left: 15px"><a href="../table_latency/" style="color: #fcfcfc">Next &raquo;</a></span>
</span>
</div>
......
sitemap.xml
浏览文件 @ 9bc13c0f
......@@ -2,82 +2,102 @@
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-03</lastmod>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
<url>
<loc>None</loc>
<lastmod>2020-02-04</lastmod>
<changefreq>daily</changefreq>
</url>
</urlset>
\ No newline at end of file
sitemap.xml.gz
浏览文件 @ 9bc13c0f
无法预览此类型文件
table_latency/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../tutorials/one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,11 +116,23 @@
</li>
<li class="">
<a class="" href="../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
<a class="" href="../api/nas_api/">SA搜索</a>
</li>
<li class="">
<a class="" href="../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../search_space/">搜索空间</a>
</li>
<li class=" current">
......@@ -321,7 +337,7 @@
<a href="../algo/algo/" class="btn btn-neutral float-right" title="算法原理">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../api/nas_api/" class="btn btn-neutral" title="SA搜索"><span class="icon icon-circle-arrow-left"></span> Previous</a>
<a href="../search_space/" class="btn btn-neutral" title="搜索空间"><span class="icon icon-circle-arrow-left"></span> Previous</a>
</div>
......@@ -349,7 +365,7 @@
<a href="https://github.com/PaddlePaddle/PaddleSlim/" class="fa fa-github" style="float: left; color: #fcfcfc"> GitHub</a>
<span><a href="../api/nas_api/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span><a href="../search_space/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span style="margin-left: 15px"><a href="../algo/algo/" style="color: #fcfcfc">Next &raquo;</a></span>
......
tutorials/demo_guide/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,7 +116,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -120,6 +128,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
tutorials/distillation_demo/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class=" current">
......@@ -135,7 +139,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -143,6 +151,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
tutorials/nas_demo/index.html
浏览文件 @ 9bc13c0f
......@@ -97,6 +97,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -124,7 +128,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -132,6 +140,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......@@ -246,7 +262,7 @@
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../../search_space/" class="btn btn-neutral float-right" title="搜索空间">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../one_shot_nas_demo/" class="btn btn-neutral float-right" title="One-shot搜索">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../quant_embedding_demo/" class="btn btn-neutral" title="Embedding量化"><span class="icon icon-circle-arrow-left"></span> Previous</a>
......@@ -280,7 +296,7 @@
<span><a href="../quant_embedding_demo/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span style="margin-left: 15px"><a href="../../search_space/" style="color: #fcfcfc">Next &raquo;</a></span>
<span style="margin-left: 15px"><a href="../one_shot_nas_demo/" style="color: #fcfcfc">Next &raquo;</a></span>
</span>
</div>
......
tutorials/one_shot_nas_demo/index.html 0 → 100644
浏览文件 @ 9bc13c0f
<!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">
<title>One-shot搜索 - PaddleSlim Docs</title>
<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">
<link href="../../extra.css" rel="stylesheet">
<script>
// Current page data
var mkdocs_page_name = "One-shot\u641c\u7d22";
var mkdocs_page_input_path = "tutorials/one_shot_nas_demo.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">
<a class="" href="../../model_zoo/">模型库</a>
</li>
<li class="toctree-l1">
<span class="caption-text">教程</span>
<ul class="subnav">
<li class="">
<a class="" href="../pruning_tutorial/">图像分类模型通道剪裁-快速开始</a>
</li>
<li class="">
<a class="" href="../quant_post_demo/">离线量化</a>
</li>
<li class="">
<a class="" href="../quant_aware_demo/">量化训练</a>
</li>
<li class="">
<a class="" href="../quant_embedding_demo/">Embedding量化</a>
</li>
<li class="">
<a class="" href="../nas_demo/">SA搜索</a>
</li>
<li class=" current">
<a class="current" href="./">One-shot搜索</a>
<ul class="subnav">
<li class="toctree-l3"><a href="#one-shot-nas">One Shot NAS 示例</a></li>
<ul>
<li><a class="toctree-l4" href="#_1">关键代码介绍</a></li>
<li><a class="toctree-l4" href="#_6">启动示例</a></li>
<li><a class="toctree-l4" href="#faq">FAQ</a></li>
</ul>
</ul>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../distillation_demo/">知识蒸馏</a>
</li>
</ul>
</li>
<li class="toctree-l1">
<span class="caption-text">API</span>
<ul class="subnav">
<li class="">
<a class="" href="../../api/quantization_api/">量化</a>
</li>
<li class="">
<a class="" href="../../api/prune_api/">剪枝与敏感度</a>
</li>
<li class="">
<a class="" href="../../api/analysis_api/">模型分析</a>
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
<a class="" href="../../api/nas_api/">SA搜索</a>
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
</li>
<li class="toctree-l1">
<a class="" href="../../algo/algo/">算法原理</a>
</li>
</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>教程 &raquo;</li>
<li>One-shot搜索</li>
<li class="wy-breadcrumbs-aside">
<a href="https://github.com/PaddlePaddle/PaddleSlim/edit/master/docs/tutorials/one_shot_nas_demo.md"
class="icon icon-github"> Edit on GitHub</a>
</li>
</ul>
<hr/>
</div>
<div role="main">
<div class="section">
<h1 id="one-shot-nas">One Shot NAS 示例<a class="headerlink" href="#one-shot-nas" title="Permanent link">#</a></h1>
<blockquote>
<p>该示例依赖Paddle1.7.0或Paddle develop版本。</p>
</blockquote>
<p>该示例使用MNIST数据，介绍了如何使用PaddleSlim的OneShotNAS接口搜索出一个分类网络。OneShotNAS仅支持动态图，所以该示例完全使用Paddle动态图模式。</p>
<h2 id="_1">关键代码介绍<a class="headerlink" href="#_1" title="Permanent link">#</a></h2>
<p>One-shot网络结构搜索策略包含以下步骤：</p>
<ol>
<li>定义超网络</li>
<li>训练超网络</li>
<li>基于超网络搜索子网络</li>
<li>训练最佳子网络</li>
</ol>
<p>以下按序介绍各个步骤的关键代码。</p>
<h3 id="_2">定义超级网络<a class="headerlink" href="#_2" title="Permanent link">#</a></h3>
<p>按照动态图教程，定义一个分类网络模块，该模块包含4个子模块：<code>_simple_img_conv_pool_1</code>,<code>_simple_img_conv_pool_2</code>,<code>super_net</code>和<code>fc</code>，其中<code>super_net</code>为<code>SuperMnasnet</code>的一个实例。</p>
<p>在前向计算过程中，输入图像先后经过子模块<code>_simple_img_conv_pool_1</code>、<code>super_net</code>、<code>_simple_img_conv_pool_2</code>和<code>fc</code>的前向计算。</p>
<p>代码如下所示：
<div class="codehilite"><pre><span></span><span class="k">class</span> <span class="n">MNIST</span>(<span class="n">fluid</span>.<span class="n">dygraph</span>.<span class="n">Layer</span>):
<span class="n">def</span> <span class="n">__init__</span>(<span class="k">self</span>):
<span class="n">super</span>(<span class="n">MNIST</span>, <span class="k">self</span>).<span class="n">__init__</span>()
<span class="k">self</span>.<span class="n">_simple_img_conv_pool_1</span> = <span class="n">SimpleImgConv</span>(<span class="mi">1</span>, <span class="mi">20</span>, <span class="mi">2</span>, <span class="n">act</span>=<span class="s">&quot;relu&quot;</span>)
<span class="k">self</span>.<span class="n">arch</span> = <span class="n">SuperMnasnet</span>(
<span class="n">name_scope</span>=<span class="s">&quot;super_net&quot;</span>, <span class="n">input_channels</span>=<span class="mi">20</span>, <span class="n">out_channels</span>=<span class="mi">20</span>)
<span class="k">self</span>.<span class="n">_simple_img_conv_pool_2</span> = <span class="n">SimpleImgConv</span>(<span class="mi">20</span>, <span class="mi">50</span>, <span class="mi">2</span>, <span class="n">act</span>=<span class="s">&quot;relu&quot;</span>)
<span class="k">self</span>.<span class="n">pool_2_shape</span> = <span class="mi">50</span> * <span class="mi">13</span> * <span class="mi">13</span>
<span class="n">SIZE</span> = <span class="mi">10</span>
<span class="n">scale</span> = (<span class="mf">2.0</span> / (<span class="k">self</span>.<span class="n">pool_2_shape</span><span class="o">**</span><span class="mi">2</span> * <span class="n">SIZE</span>))**<span class="mf">0.5</span>
<span class="k">self</span>.<span class="n">_fc</span> = <span class="n">Linear</span>(
<span class="k">self</span>.<span class="n">pool_2_shape</span>,
<span class="mi">10</span>,
<span class="n">param_attr</span>=<span class="n">fluid</span>.<span class="n">param_attr</span>.<span class="n">ParamAttr</span>(
<span class="n">initializer</span>=<span class="n">fluid</span>.<span class="n">initializer</span>.<span class="n">NormalInitializer</span>(
<span class="n">loc</span>=<span class="mf">0.0</span>, <span class="n">scale</span>=<span class="n">scale</span>)),
<span class="n">act</span>=<span class="s">&quot;softmax&quot;</span>)
<span class="n">def</span> <span class="n">forward</span>(<span class="k">self</span>, <span class="n">inputs</span>, <span class="n">label</span>=<span class="n">None</span>, <span class="n">tokens</span>=<span class="n">None</span>):
<span class="o">x</span> = <span class="k">self</span>.<span class="n">_simple_img_conv_pool_1</span>(<span class="n">inputs</span>)
<span class="o">x</span> = <span class="k">self</span>.<span class="n">arch</span>(<span class="o">x</span>, <span class="n">tokens</span>=<span class="n">tokens</span>) <span class="c1"># addddddd</span>
<span class="o">x</span> = <span class="k">self</span>.<span class="n">_simple_img_conv_pool_2</span>(<span class="o">x</span>)
<span class="o">x</span> = <span class="n">fluid</span>.<span class="n">layers</span>.<span class="n">reshape</span>(<span class="o">x</span>, <span class="nb">shape</span>=[-<span class="mi">1</span>, <span class="k">self</span>.<span class="n">pool_2_shape</span>])
<span class="o">x</span> = <span class="k">self</span>.<span class="n">_fc</span>(<span class="o">x</span>)
<span class="k">if</span> <span class="n">label</span> <span class="k">is</span> <span class="nb">not</span> <span class="n">None:</span>
<span class="n">acc</span> = <span class="n">fluid</span>.<span class="n">layers</span>.<span class="n">accuracy</span>(<span class="n">input</span>=<span class="o">x</span>, <span class="n">label</span>=<span class="n">label</span>)
<span class="k">return</span> <span class="o">x</span>, <span class="n">acc</span>
<span class="n">else:</span>
<span class="k">return</span> <span class="o">x</span>
</pre></div></p>
<p>动态图模块MNIST的forward函数接受一个参数<code>tokens</code>，用于指定在前向计算中使用的子网络，如果<code>tokens</code>为None，则随机选取一个子网络进行前向计算。</p>
<h3 id="_3">训练超级网络<a class="headerlink" href="#_3" title="Permanent link">#</a></h3>
<p>网络训练的逻辑定义在<code>train_mnist</code>函数中，将<code>tokens</code>参数设置为None，进行超网络训练，即在每个batch选取一个超网络进行训练。</p>
<p>代码如下所示：</p>
<div class="codehilite"><pre><span></span><span class="err">with fluid.dygraph.guard(place):</span>
<span class="err"> model = MNIST()</span>
<span class="err"> train_mnist(args, model)</span>
</pre></div>
<h3 id="_4">搜索最佳子网络<a class="headerlink" href="#_4" title="Permanent link">#</a></h3>
<p>使用PaddleSlim提供的<code>OneShotSearch</code>接口搜索最佳子网络。传入已定义且训练好的超网络实例<code>model</code>和一个用于评估子网络的回调函数<code>test_mnist</code>.</p>
<p>代码如下：</p>
<div class="codehilite"><pre><span></span><span class="err">best_tokens = OneShotSearch(model, test_mnist)</span>
</pre></div>
<h3 id="_5">训练最佳子网络<a class="headerlink" href="#_5" title="Permanent link">#</a></h3>
<p>获得最佳的子网络的编码<code>best_tokens</code>后，调用之前定义的<code>train_mnist</code>方法进行子网络的训练。代码如下：</p>
<div class="codehilite"><pre><span></span><span class="err">train_mnist(args, model, best_tokens)</span>
</pre></div>
<h2 id="_6">启动示例<a class="headerlink" href="#_6" title="Permanent link">#</a></h2>
<p>执行以下代码运行示例：</p>
<div class="codehilite"><pre><span></span><span class="err">python train.py</span>
</pre></div>
<p>执行<code>python train.py --help</code>查看更多可配置选项。</p>
<h2 id="faq">FAQ<a class="headerlink" href="#faq" title="Permanent link">#</a></h2>
</div>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../../search_space/" class="btn btn-neutral float-right" title="搜索空间">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../nas_demo/" class="btn btn-neutral" title="SA搜索"><span class="icon icon-circle-arrow-left"></span> Previous</a>
</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="../nas_demo/" style="color: #fcfcfc;">&laquo; Previous</a></span>
<span style="margin-left: 15px"><a href="../../search_space/" 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>
<script src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS-MML_HTMLorMML" defer></script>
<script src="../../search/main.js" defer></script>
</body>
</html>
tutorials/pruning_tutorial/index.html
浏览文件 @ 9bc13c0f
......@@ -103,6 +103,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -130,7 +134,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -138,6 +146,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
tutorials/quant_aware_demo/index.html
浏览文件 @ 9bc13c0f
......@@ -99,6 +99,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -126,7 +130,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -134,6 +142,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
tutorials/quant_embedding_demo/index.html
浏览文件 @ 9bc13c0f
......@@ -99,6 +99,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -126,7 +130,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -134,6 +142,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
tutorials/quant_post_demo/index.html
浏览文件 @ 9bc13c0f
......@@ -99,6 +99,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -126,7 +130,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -134,6 +142,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
tutorials/sensitivity_demo/index.html
浏览文件 @ 9bc13c0f
......@@ -85,6 +85,10 @@
</li>
<li class="">
<a class="" href="../one_shot_nas_demo/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
......@@ -112,7 +116,11 @@
</li>
<li class="">
<a class="" href="../../api/single_distiller_api/">知识蒸馏</a>
<a class="" href="../../api/single_distiller_api/">简单知识蒸馏</a>
</li>
<li class="">
<a class="" href="../../api/pantheon_api/">大规模可扩展知识蒸馏框架 Pantheon</a>
</li>
<li class="">
......@@ -120,6 +128,14 @@
</li>
<li class="">
<a class="" href="../../api/one_shot_api/">One-shot搜索</a>
</li>
<li class="">
<a class="" href="../../search_space/">搜索空间</a>
</li>
<li class="">
<a class="" href="../../table_latency/">硬件延时评估表</a>
</li>
</ul>
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或