Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 1 年 前同步成功

通知 2298
Star 20931
Fork 5422
P
Paddle
切换分支/标签
查找文件 普通视图历史永久链接Permalink
index_cn.html 36.5 KB
Newer Older
T
Deploy to GitHub Pages: 1ff47b284c565d030b198705d5f18b4bd4ce53e5  
Travis CI 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 


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

  
  

  

  
  
    

  

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

  
  
        <link rel="index" title="索引"
              href="../../genindex.html"/>
        <link rel="search" title="搜索" href="../../search.html"/>
    <link rel="top" title="PaddlePaddle  文档" href="../../index.html"/>
        <link rel="up" title="FAQ" href="../index_cn.html"/>
        <link rel="next" title="本地训练与预测" href="../local/index_cn.html"/>
        <link rel="prev" title="模型配置" href="../model/index_cn.html"/> 

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

  

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

</head>

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

  
  <header class="site-header">
    <div class="site-logo">
      <a href="/"><img src="../../_static/images/PP_w.png"></a>
    </div>
    <div class="site-nav-links">
      <div class="site-menu">
        <a class="fork-on-github" href="https://github.com/PaddlePaddle/Paddle" target="_blank"><i class="fa fa-github"></i>Fork me on Github</a>
        <div class="language-switcher dropdown">
          <a type="button" data-toggle="dropdown">
            <span>English</span>
            <i class="fa fa-angle-up"></i>
            <i class="fa fa-angle-down"></i>
          </a>
          <ul class="dropdown-menu">
            <li><a href="/doc_cn">中文</a></li>
            <li><a href="/doc">English</a></li>
          </ul>
        </div>
        <ul class="site-page-links">
          <li><a href="/">Home</a></li>
        </ul>
      </div>
      <div class="doc-module">
        
        <ul class="current">
<li class="toctree-l1"><a class="reference internal" href="../../getstarted/index_cn.html">新手入门</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../howto/index_cn.html">进阶指南</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../api/index_cn.html">API</a></li>
<li class="toctree-l1 current"><a class="reference internal" href="../index_cn.html">FAQ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../mobile/index_cn.html">MOBILE</a></li>
</ul>

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

    
    <nav class="doc-menu-vertical" role="navigation">
        
          
          <ul class="current">
<li class="toctree-l1"><a class="reference internal" href="../../getstarted/index_cn.html">新手入门</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../../getstarted/build_and_install/index_cn.html">安装与编译</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../../getstarted/build_and_install/pip_install_cn.html">使用pip安装</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../getstarted/build_and_install/docker_install_cn.html">使用Docker安装运行</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../howto/dev/build_cn.html">用Docker编译和测试PaddlePaddle</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../getstarted/build_and_install/build_from_source_cn.html">从源码编译</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../../getstarted/concepts/use_concepts_cn.html">基本使用概念</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="../../howto/index_cn.html">进阶指南</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../../howto/usage/cmd_parameter/index_cn.html">设置命令行参数</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../../howto/usage/cmd_parameter/use_case_cn.html">使用案例</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../howto/usage/cmd_parameter/arguments_cn.html">参数概述</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../howto/usage/cmd_parameter/detail_introduction_cn.html">细节描述</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/usage/cluster/cluster_train_cn.html">PaddlePaddle分布式训练</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/usage/k8s/k8s_basis_cn.html">Kubernetes 简介</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/usage/k8s/k8s_cn.html">Kubernetes单机训练</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/usage/k8s/k8s_distributed_cn.html">Kubernetes分布式训练</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/dev/contribute_to_paddle_cn.html">如何贡献代码</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/dev/write_docs_cn.html">如何贡献/修改文档</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/deep_model/rnn/index_cn.html">RNN相关模型</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../../howto/deep_model/rnn/rnn_config_cn.html">RNN配置</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../howto/deep_model/rnn/recurrent_group_cn.html">Recurrent Group教程</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../howto/deep_model/rnn/hierarchical_layer_cn.html">支持双层序列作为输入的Layer</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../howto/deep_model/rnn/hrnn_rnn_api_compare_cn.html">单双层RNN API对比介绍</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../../howto/optimization/gpu_profiling_cn.html">GPU性能分析与调优</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="../../api/index_cn.html">API</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../../api/v2/model_configs.html">模型配置</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/activation.html">Activation</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/layer.html">Layers</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/evaluators.html">Evaluators</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/optimizer.html">Optimizer</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/pooling.html">Pooling</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/networks.html">Networks</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/config/attr.html">Parameter Attribute</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../../api/v2/data.html">数据访问</a><ul>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/data/data_reader.html">Data Reader Interface</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/data/image.html">Image Interface</a></li>
<li class="toctree-l3"><a class="reference internal" href="../../api/v2/data/dataset.html">Dataset</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="../../api/v2/run_logic.html">训练与应用</a></li>
</ul>
</li>
<li class="toctree-l1 current"><a class="reference internal" href="../index_cn.html">FAQ</a><ul class="current">
<li class="toctree-l2"><a class="reference internal" href="../build_and_install/index_cn.html">编译安装与单元测试</a></li>
<li class="toctree-l2"><a class="reference internal" href="../model/index_cn.html">模型配置</a></li>
<li class="toctree-l2 current"><a class="current reference internal" href="#">参数设置</a></li>
<li class="toctree-l2"><a class="reference internal" href="../local/index_cn.html">本地训练与预测</a></li>
<li class="toctree-l2"><a class="reference internal" href="../cluster/index_cn.html">集群训练与预测</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="../../mobile/index_cn.html">MOBILE</a><ul>
<li class="toctree-l2"><a class="reference internal" href="../../mobile/cross_compiling_for_android_cn.html">Android平台编译指南</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../mobile/cross_compiling_for_ios_cn.html">iOS平台编译指南</a></li>
<li class="toctree-l2"><a class="reference internal" href="../../mobile/cross_compiling_for_raspberry_cn.html">Raspberry Pi平台编译指南</a></li>
</ul>
</li>
</ul>

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

      

 







<div role="navigation" aria-label="breadcrumbs navigation">
  <ul class="wy-breadcrumbs">
      
        <li><a href="../index_cn.html">FAQ</a> > </li>
      
    <li>参数设置</li>
  </ul>
</div>
      
      <div class="wy-nav-content" id="doc-content">
        <div class="rst-content">
          <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
           <div itemprop="articleBody">
            
  <div class="section" id="id1">
<h1><a class="toc-backref" href="#id6">参数设置</a><a class="headerlink" href="#id1" title="永久链接至标题"></a></h1>
<div class="contents topic" id="contents">
<p class="topic-title first">Contents</p>
<ul class="simple">
<li><a class="reference internal" href="#id1" id="id6">参数设置</a><ul>
<li><a class="reference internal" href="#sgd" id="id7">1. 如何选择SGD算法的学习率</a></li>
<li><a class="reference internal" href="#learning-rate-annealing" id="id8">2. 如何设置学习率退火（learning rate annealing）</a></li>
<li><a class="reference internal" href="#id2" id="id9">3. 如何初始化参数</a></li>
<li><a class="reference internal" href="#id3" id="id10">4. 如何共享参数</a></li>
<li><a class="reference internal" href="#id4" id="id11">5. 如何加载预训练参数</a></li>
<li><a class="reference internal" href="#id5" id="id12">6. 存储的参数格式是什么，如何和明文进行相互转化</a></li>
<li><a class="reference internal" href="#a-protocol-message-was-rejected-because-it-was-too-big" id="id13">7. A protocol message was rejected because it was too big</a></li>
</ul>
</li>
</ul>
</div>
<div class="section" id="sgd">
<h2><a class="toc-backref" href="#id7">1. 如何选择SGD算法的学习率</a><a class="headerlink" href="#sgd" title="永久链接至标题"></a></h2>
<p>在采用sgd/async_sgd进行训练时，一个重要的问题是选择正确的learning_rate。如果learning_rate太大，那么训练有可能不收敛，如果learning_rate太小，那么收敛可能很慢，导致训练时间过长。</p>
<p>通常做法是从一个比较大的learning_rate开始试，如果不收敛，那减少学习率10倍继续试验，直到训练收敛为止。那么如何判断训练不收敛呢？可以估计出如果模型采用不变的输出最小的cost0是多少。</p>
<p>如果训练过程的的cost明显高于这个常数输出的cost，那么我们可以判断为训练不收敛。举一个例子，假如我们是三分类问题，采用multi-class-cross-entropy作为cost，数据中0,1,2三类的比例为 <code class="code docutils literal"><span class="pre">0.2,</span> <span class="pre">0.5,</span> <span class="pre">0.3</span></code> , 那么常数输出所能达到的最小cost是 <code class="code docutils literal"><span class="pre">-(0.2*log(0.2)+0.5*log(0.5)+0.3*log(0.3))=1.03</span></code> 。如果训练一个pass（或者更早）后，cost还大于这个数，那么可以认为训练不收敛，应该降低学习率。</p>
</div>
<div class="section" id="learning-rate-annealing">
<h2><a class="toc-backref" href="#id8">2. 如何设置学习率退火（learning rate annealing）</a><a class="headerlink" href="#learning-rate-annealing" title="永久链接至标题"></a></h2>
<p>在相应的优化算法里设置learning_rate_schedule及相关参数，以使用Adam算法为例，代码如下：</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">Adam</span><span class="p">(</span>
    <span class="n">learning_rate</span><span class="o">=</span><span class="mf">1e-3</span><span class="p">,</span>
    <span class="n">learning_rate_decay_a</span><span class="o">=</span><span class="mf">0.5</span><span class="p">,</span>
    <span class="n">learning_rate_decay_b</span><span class="o">=</span><span class="mf">0.75</span><span class="p">,</span>
    <span class="n">learning_rate_schedule</span><span class="o">=</span><span class="s2">&quot;poly&quot;</span><span class="p">,)</span>
</pre></div>
</div>
<p>PaddlePaddle目前支持8种learning_rate_schedule，这8种learning_rate_schedule及其对应学习率计算方式如下：</p>
<ul>
<li><p class="first">&#8220;constant&#8221;</p>
<p>lr = learning_rate</p>
</li>
<li><p class="first">&#8220;poly&#8221;</p>
<p>lr = learning_rate * pow(1 + learning_rate_decay_a * num_samples_processed, -learning_rate_decay_b)</p>
<p>其中，num_samples_processed为已训练样本数，下同。</p>
</li>
<li><p class="first">&#8220;caffe_poly&#8221;</p>
<p>lr = learning_rate * pow(1.0 - num_samples_processed / learning_rate_decay_a, learning_rate_decay_b)</p>
</li>
<li><p class="first">&#8220;exp&#8221;</p>
<p>lr = learning_rate * pow(learning_rate_decay_a, num_samples_processed / learning_rate_decay_b)</p>
</li>
<li><p class="first">&#8220;discexp&#8221;</p>
<p>lr = learning_rate * pow(learning_rate_decay_a, floor(num_samples_processed / learning_rate_decay_b))</p>
</li>
<li><p class="first">&#8220;linear&#8221;</p>
<p>lr = max(learning_rate - learning_rate_decay_a * num_samples_processed, learning_rate_decay_b)</p>
</li>
<li><p class="first">&#8220;manual&#8221;</p>
<p>这是一种按已训练样本数分段取值的学习率退火方法。使用该learning_rate_schedule时，用户通过参数 <code class="code docutils literal"><span class="pre">learning_rate_args</span></code> 设置学习率衰减因子分段函数，当前的学习率为所设置 <code class="code docutils literal"><span class="pre">learning_rate</span></code> 与当前的衰减因子的乘积。以使用Adam算法为例，代码如下：</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">Adam</span><span class="p">(</span>
    <span class="n">learning_rate</span><span class="o">=</span><span class="mf">1e-3</span><span class="p">,</span>
    <span class="n">learning_rate_schedule</span><span class="o">=</span><span class="s2">&quot;manual&quot;</span><span class="p">,</span>
    <span class="n">learning_rate_args</span><span class="o">=</span><span class="s2">&quot;1000:1.0,2000:0.9,3000:0.8&quot;</span><span class="p">,)</span>
</pre></div>
</div>
<p>在该示例中，当已训练样本数小于等于1000时，学习率为 <code class="code docutils literal"><span class="pre">1e-3</span> <span class="pre">*</span> <span class="pre">1.0</span></code>；当已训练样本数大于1000小于等于2000时，学习率为 <code class="code docutils literal"><span class="pre">1e-3</span> <span class="pre">*</span> <span class="pre">0.9</span></code>；当已训练样本数大于2000时，学习率为 <code class="code docutils literal"><span class="pre">1e-3</span> <span class="pre">*</span> <span class="pre">0.8</span></code>。</p>
</li>
<li><p class="first">&#8220;pass_manual&#8221;</p>
<p>这是一种按已训练pass数分段取值的学习率退火方法。使用该learning_rate_schedule时，用户通过参数 <code class="code docutils literal"><span class="pre">learning_rate_args</span></code> 设置学习率衰减因子分段函数，当前的学习率为所设置 <code class="code docutils literal"><span class="pre">learning_rate</span></code> 与当前的衰减因子的乘积。以使用Adam算法为例，代码如下：</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">optimizer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">Adam</span><span class="p">(</span>
    <span class="n">learning_rate</span><span class="o">=</span><span class="mf">1e-3</span><span class="p">,</span>
    <span class="n">learning_rate_schedule</span><span class="o">=</span><span class="s2">&quot;pass_manual&quot;</span><span class="p">,</span>
    <span class="n">learning_rate_args</span><span class="o">=</span><span class="s2">&quot;1:1.0,2:0.9,3:0.8&quot;</span><span class="p">,)</span>
</pre></div>
</div>
<p>在该示例中，当已训练pass数小于等于1时，学习率为 <code class="code docutils literal"><span class="pre">1e-3</span> <span class="pre">*</span> <span class="pre">1.0</span></code>；当已训练pass数大于1小于等于2时，学习率为 <code class="code docutils literal"><span class="pre">1e-3</span> <span class="pre">*</span> <span class="pre">0.9</span></code>；当已训练pass数大于2时，学习率为 <code class="code docutils literal"><span class="pre">1e-3</span> <span class="pre">*</span> <span class="pre">0.8</span></code>。</p>
</li>
</ul>
</div>
<div class="section" id="id2">
<h2><a class="toc-backref" href="#id9">3. 如何初始化参数</a><a class="headerlink" href="#id2" title="永久链接至标题"></a></h2>
<p>默认情况下，PaddlePaddle使用均值0，标准差为 <span class="math">\(\frac{1}{\sqrt{d}}\)</span> 来初始化参数。其中 <span class="math">\(d\)</span> 为参数矩阵的宽度。这种初始化方式在一般情况下不会产生很差的结果。如果用户想要自定义初始化方式，PaddlePaddle目前提供两种参数初始化的方式:</p>
<ul class="simple">
<li>高斯分布。将 <code class="code docutils literal"><span class="pre">param_attr</span></code> 设置成 <code class="code docutils literal"><span class="pre">param_attr=ParamAttr(initial_mean=0.0,</span> <span class="pre">initial_std=1.0)</span></code></li>
<li>均匀分布。将 <code class="code docutils literal"><span class="pre">param_attr</span></code> 设置成 <code class="code docutils literal"><span class="pre">param_attr=ParamAttr(initial_max=1.0,</span> <span class="pre">initial_min=-1.0)</span></code></li>
</ul>
<p>比如设置一个全连接层的参数初始化方式和bias初始化方式，可以使用如下代码。</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">hidden</span> <span class="o">=</span> <span class="n">fc_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">ipt</span><span class="p">,</span> <span class="n">param_attr</span><span class="o">=</span><span class="n">ParamAttr</span><span class="p">(</span><span class="n">initial_max</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span> <span class="n">initial_min</span><span class="o">=-</span><span class="mf">1.0</span><span class="p">),</span>
                  <span class="n">bias_attr</span><span class="o">=</span><span class="n">ParamAttr</span><span class="p">(</span><span class="n">initial_mean</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span> <span class="n">initial_std</span><span class="o">=</span><span class="mf">0.0</span><span class="p">))</span>
</pre></div>
</div>
<p>上述代码将bias全部初始化为1.0, 同时将参数初始化为 <code class="code docutils literal"><span class="pre">[1.0,</span> <span class="pre">-1.0]</span></code> 的均匀分布。</p>
</div>
<div class="section" id="id3">
<h2><a class="toc-backref" href="#id10">4. 如何共享参数</a><a class="headerlink" href="#id3" title="永久链接至标题"></a></h2>
<p>PaddlePaddle的参数使用名字 <code class="code docutils literal"><span class="pre">name</span></code> 作为参数的ID，相同名字的参数，会共享参数。设置参数的名字，可以使用 <code class="code docutils literal"><span class="pre">ParamAttr(name=&quot;YOUR_PARAM_NAME&quot;)</span></code> 来设置。更方便的设置方式，是使得要共享的参数使用同样的 <code class="code docutils literal"><span class="pre">ParamAttr</span></code> 对象。</p>
<p>简单的全连接网络，参数共享的配置示例为:</p>
<p>这里 <code class="code docutils literal"><span class="pre">hidden_a</span></code> 和 <code class="code docutils literal"><span class="pre">hidden_b</span></code> 使用了同样的parameter和bias。并且softmax层的两个输入也使用了同样的参数 <code class="code docutils literal"><span class="pre">softmax_param</span></code>。</p>
</div>
<div class="section" id="id4">
<h2><a class="toc-backref" href="#id11">5. 如何加载预训练参数</a><a class="headerlink" href="#id4" title="永久链接至标题"></a></h2>
<ul class="simple">
<li>对加载预训练参数的层，设置其参数属性 <code class="code docutils literal"><span class="pre">is_static=True</span></code>，使该层的参数在训练过程中保持不变。以embedding层为例，代码如下：</li>
</ul>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">emb_para</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">attr</span><span class="o">.</span><span class="n">Param</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;emb&#39;</span><span class="p">,</span> <span class="n">is_static</span><span class="o">=</span><span class="bp">True</span><span class="p">)</span>
<span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">embedding</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">word_dim</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">param_attr</span><span class="o">=</span><span class="n">emb_para</span><span class="p">)</span>
</pre></div>
</div>
<ul class="simple">
<li>从模型文件将预训练参数载入 <code class="code docutils literal"><span class="pre">numpy.array</span></code>，在创建parameters后，使用 <code class="code docutils literal"><span class="pre">parameters.set()</span></code> 加载预训练参数。PaddlePaddle保存的模型参数文件前16字节为头信息，用户将参数载入 <code class="code docutils literal"><span class="pre">numpy.array</span></code> 时须从第17字节开始。以embedding层为例，代码如下：</li>
</ul>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">load_parameter</span><span class="p">(</span><span class="n">file_name</span><span class="p">,</span> <span class="n">h</span><span class="p">,</span> <span class="n">w</span><span class="p">):</span>
    <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">file_name</span><span class="p">,</span> <span class="s1">&#39;rb&#39;</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
        <span class="n">f</span><span class="o">.</span><span class="n">read</span><span class="p">(</span><span class="mi">16</span><span class="p">)</span>  <span class="c1"># skip header.</span>
        <span class="k">return</span> <span class="n">np</span><span class="o">.</span><span class="n">fromfile</span><span class="p">(</span><span class="n">f</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">)</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="n">h</span><span class="p">,</span> <span class="n">w</span><span class="p">)</span>

<span class="n">parameters</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">parameters</span><span class="o">.</span><span class="n">create</span><span class="p">(</span><span class="n">my_cost</span><span class="p">)</span>
<span class="n">parameters</span><span class="o">.</span><span class="n">set</span><span class="p">(</span><span class="s1">&#39;emb&#39;</span><span class="p">,</span> <span class="n">load_parameter</span><span class="p">(</span><span class="n">emb_param_file</span><span class="p">,</span> <span class="mi">30000</span><span class="p">,</span> <span class="mi">256</span><span class="p">))</span>
</pre></div>
</div>
</div>
<div class="section" id="id5">
<h2><a class="toc-backref" href="#id12">6. 存储的参数格式是什么，如何和明文进行相互转化</a><a class="headerlink" href="#id5" title="永久链接至标题"></a></h2>
<p>PaddlePaddle保存的模型参数文件内容由16字节头信息和网络参数两部分组成。头信息中，1~4字节表示PaddlePaddle版本信息，请直接填充0；5~8字节表示每个参数占用的字节数，当保存的网络参数为float类型时为4，double类型时为8；9~16字节表示保存的参数总个数。</p>
<p>将PaddlePaddle保存的模型参数还原回明文时，可以使用相应数据类型的 <code class="code docutils literal"><span class="pre">numpy.array</span></code> 加载具体网络参数，此时可以跳过PaddlePaddle模型参数文件的头信息。若在PaddlePaddle编译时，未指定按照double精度编译，默认情况下按照float精度计算，保存的参数也是float类型。这时在使用 <code class="code docutils literal"><span class="pre">numpy.array</span></code> 时，一般设置 <code class="code docutils literal"><span class="pre">dtype=float32</span></code> 。示例如下：</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">read_parameter</span><span class="p">(</span><span class="n">fname</span><span class="p">,</span> <span class="n">width</span><span class="p">):</span>
    <span class="n">s</span> <span class="o">=</span> <span class="nb">open</span><span class="p">(</span><span class="n">fname</span><span class="p">)</span><span class="o">.</span><span class="n">read</span><span class="p">()</span>
    <span class="c1"># skip header</span>
    <span class="n">vec</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">fromstring</span><span class="p">(</span><span class="n">s</span><span class="p">[</span><span class="mi">16</span><span class="p">:],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">)</span>
    <span class="c1"># width is the size of the corresponding layer</span>
    <span class="n">np</span><span class="o">.</span><span class="n">savetxt</span><span class="p">(</span><span class="n">fname</span> <span class="o">+</span> <span class="s2">&quot;.csv&quot;</span><span class="p">,</span> <span class="n">vec</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">),</span>
            <span class="n">fmt</span><span class="o">=</span><span class="s2">&quot;</span><span class="si">%.6f</span><span class="s2">&quot;</span><span class="p">,</span> <span class="n">delimiter</span><span class="o">=</span><span class="s2">&quot;,&quot;</span><span class="p">)</span>
</pre></div>
</div>
<p>将明文参数转化为PaddlePaddle可加载的模型参数时，首先构造头信息，再写入网络参数。下面的代码将随机生成的矩阵转化为可以被PaddlePaddle加载的模型参数。</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">gen_rand_param</span><span class="p">(</span><span class="n">param_file</span><span class="p">,</span> <span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">need_trans</span><span class="p">):</span>
    <span class="n">np</span><span class="o">.</span><span class="n">random</span><span class="o">.</span><span class="n">seed</span><span class="p">()</span>
    <span class="n">header</span> <span class="o">=</span> <span class="n">struct</span><span class="o">.</span><span class="n">pack</span><span class="p">(</span><span class="s2">&quot;iil&quot;</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="n">height</span> <span class="o">*</span> <span class="n">width</span><span class="p">)</span>
    <span class="n">param</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">float32</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">random</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="n">height</span><span class="p">,</span> <span class="n">width</span><span class="p">))</span>
    <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">param_file</span><span class="p">,</span> <span class="s2">&quot;w&quot;</span><span class="p">)</span> <span class="k">as</span> <span class="n">fparam</span><span class="p">:</span>
        <span class="n">fparam</span><span class="o">.</span><span class="n">write</span><span class="p">(</span><span class="n">header</span> <span class="o">+</span> <span class="n">param</span><span class="o">.</span><span class="n">tostring</span><span class="p">())</span>
</pre></div>
</div>
</div>
<div class="section" id="a-protocol-message-was-rejected-because-it-was-too-big">
<h2><a class="toc-backref" href="#id13">7. A protocol message was rejected because it was too big</a><a class="headerlink" href="#a-protocol-message-was-rejected-because-it-was-too-big" title="永久链接至标题"></a></h2>
<p>如果在训练NLP相关模型时，出现以下错误：</p>
<div class="highlight-bash"><div class="highlight"><pre><span></span><span class="o">[</span>libprotobuf ERROR google/protobuf/io/coded_stream.cc:171<span class="o">]</span> A protocol message was rejected because it was too big <span class="o">(</span>more than <span class="m">67108864</span> bytes<span class="o">)</span>.  To increase the limit <span class="o">(</span>or to disable these warnings<span class="o">)</span>, see CodedInputStream::SetTotalBytesLimit<span class="o">()</span> in google/protobuf/io/coded_stream.h.
F1205 <span class="m">14</span>:59:50.295174 <span class="m">14703</span> TrainerConfigHelper.cpp:59<span class="o">]</span> Check failed: m-&gt;conf.ParseFromString<span class="o">(</span>configProtoStr<span class="o">)</span>
</pre></div>
</div>
<p>可能的原因是：传给dataprovider的某一个args过大，一般是由于直接传递大字典导致的。错误的define_py_data_sources2类似：</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">src_dict</span> <span class="o">=</span> <span class="nb">dict</span><span class="p">()</span>
<span class="k">for</span> <span class="n">line_count</span><span class="p">,</span> <span class="n">line</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="nb">open</span><span class="p">(</span><span class="n">src_dict_path</span><span class="p">,</span> <span class="s2">&quot;r&quot;</span><span class="p">)):</span>
   <span class="n">src_dict</span><span class="p">[</span><span class="n">line</span><span class="o">.</span><span class="n">strip</span><span class="p">()]</span> <span class="o">=</span> <span class="n">line_count</span>

<span class="n">define_py_data_sources2</span><span class="p">(</span>
   <span class="n">train_list</span><span class="p">,</span>
   <span class="n">test_list</span><span class="p">,</span>
   <span class="n">module</span><span class="o">=</span><span class="s2">&quot;dataprovider&quot;</span><span class="p">,</span>
   <span class="n">obj</span><span class="o">=</span><span class="s2">&quot;process&quot;</span><span class="p">,</span>
   <span class="n">args</span><span class="o">=</span><span class="p">{</span><span class="s2">&quot;src_dict&quot;</span><span class="p">:</span> <span class="n">src_dict</span><span class="p">})</span>
</pre></div>
</div>
<p>解决方案是：将字典的地址作为args传给dataprovider，然后在dataprovider里面根据该地址加载字典。即define_py_data_sources2应改为：</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">define_py_data_sources2</span><span class="p">(</span>
   <span class="n">train_list</span><span class="p">,</span>
   <span class="n">test_list</span><span class="p">,</span>
   <span class="n">module</span><span class="o">=</span><span class="s2">&quot;dataprovider&quot;</span><span class="p">,</span>
   <span class="n">obj</span><span class="o">=</span><span class="s2">&quot;process&quot;</span><span class="p">,</span>
   <span class="n">args</span><span class="o">=</span><span class="p">{</span><span class="s2">&quot;src_dict_path&quot;</span><span class="p">:</span> <span class="n">src_dict_path</span><span class="p">})</span>
</pre></div>
</div>
<p>完整源码可参考 <a class="reference external" href="https://github.com/PaddlePaddle/Paddle/tree/develop/demo/seqToseq">seqToseq</a> 示例。</p>
</div>
</div>


           </div>
          </div>
          <footer>
  
    <div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
      
        <a href="../local/index_cn.html" class="btn btn-neutral float-right" title="本地训练与预测" accesskey="n">Next <span class="fa fa-arrow-circle-right"></span></a>
      
      
        <a href="../model/index_cn.html" class="btn btn-neutral" title="模型配置" accesskey="p"><span class="fa fa-arrow-circle-left"></span> Previous</a>
      
    </div>
  

  <hr/>

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

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

</footer>

        </div>
      </div>

    </section>

  </div>
  


  

    <script type="text/javascript">
        var DOCUMENTATION_OPTIONS = {
            URL_ROOT:'../../',
            VERSION:'',
            COLLAPSE_INDEX:false,
            FILE_SUFFIX:'.html',
            HAS_SOURCE:  true,
            SOURCELINK_SUFFIX: ".txt",
        };
    </script>
      <script type="text/javascript" src="../../_static/jquery.js"></script>
      <script type="text/javascript" src="../../_static/underscore.js"></script>
      <script type="text/javascript" src="../../_static/doctools.js"></script>
      <script type="text/javascript" src="../../_static/translations.js"></script>
      <script type="text/javascript" src="https://cdn.bootcss.com/mathjax/2.7.0/MathJax.js"></script>
       
  

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

</body>
</html>