Deploy to GitHub Pages: e4952ca6

ac963590 · Travis CI · f4904247 · ac963590 · ac963590 · ac963590
24 changed file
--- a/doc/_sources/ui/api/trainer_config_helpers/layers.txt
+++ b/doc/_sources/ui/api/trainer_config_helpers/layers.txt
@@ -130,6 +130,12 @@ gru_step_layer
 Recurrent Layer Group
 =====================
+memory
+------
+..  automodule:: paddle.trainer_config_helpers.layers
+    :members: memory
+    :noindex:
 recurrent_group
 ---------------
 ..  automodule:: paddle.trainer_config_helpers.layers

--- a/doc/genindex.html
+++ b/doc/genindex.html
@@ -1823,7 +1823,7 @@ var _hmt = _hmt || [];
  </dt>
-  <dt><a href="source/cuda/rnn/rnn.html#_CPPv222hl_sequence2batch_copyP4realP4realPiiib">hl_sequence2batch_copy (C++ function)</a>
+  <dt><a href="source/cuda/rnn/rnn.html#_CPPv222hl_sequence2batch_copyP4realP4realPKiiib">hl_sequence2batch_copy (C++ function)</a>
  </dt>
@@ -5247,7 +5247,7 @@ var _hmt = _hmt || [];
  </dt>
-  <dt><a href="source/math/matrix/matrix.html#_CPPv2N6paddle9CpuMatrix14copyByRowIndexER6MatrixR7IVector">paddle::CpuMatrix::copyByRowIndex (C++ function)</a>
+  <dt><a href="source/math/matrix/matrix.html#_CPPv2N6paddle9CpuMatrix14copyByRowIndexER6MatrixRK7IVector">paddle::CpuMatrix::copyByRowIndex (C++ function)</a>
  </dt>
@@ -6955,10 +6955,6 @@ var _hmt = _hmt || [];
  </dt>
-  <dt><a href="source/gserver/layers/layer.html#_CPPv2N6paddle11ExpandLayer19cpuExpandStartsPos_E">paddle::ExpandLayer::cpuExpandStartsPos_ (C++ member)</a>
-  </dt>
  <dt><a href="source/gserver/layers/layer.html#_CPPv2N6paddle11ExpandLayer11ExpandLayerERK11LayerConfig">paddle::ExpandLayer::ExpandLayer (C++ function)</a>
  </dt>
@@ -7359,7 +7355,7 @@ var _hmt = _hmt || [];
  </dt>
-  <dt><a href="source/math/matrix/matrix.html#_CPPv2N6paddle9GpuMatrix14copyByRowIndexER6MatrixR7IVector">paddle::GpuMatrix::copyByRowIndex (C++ function)</a>
+  <dt><a href="source/math/matrix/matrix.html#_CPPv2N6paddle9GpuMatrix14copyByRowIndexER6MatrixRK7IVector">paddle::GpuMatrix::copyByRowIndex (C++ function)</a>
  </dt>
@@ -9339,7 +9335,7 @@ var _hmt = _hmt || [];
  </dt>
-  <dt><a href="source/math/matrix/matrix.html#_CPPv2N6paddle6Matrix14copyByRowIndexER6MatrixR7IVector">paddle::Matrix::copyByRowIndex (C++ function)</a>
+  <dt><a href="source/math/matrix/matrix.html#_CPPv2N6paddle6Matrix14copyByRowIndexER6MatrixRK7IVector">paddle::Matrix::copyByRowIndex (C++ function)</a>
  </dt>
@@ -14465,10 +14461,6 @@ var _hmt = _hmt || [];
  </dt>
-  <dt><a href="source/gserver/layers/layer.html#_CPPv2N6paddle25SequenceScatterAgentLayer17cpuInputStartPos_E">paddle::SequenceScatterAgentLayer::cpuInputStartPos_ (C++ member)</a>
-  </dt>
  <dt><a href="source/gserver/layers/layer.html#_CPPv2N6paddle25SequenceScatterAgentLayer7forwardE8PassType">paddle::SequenceScatterAgentLayer::forward (C++ function)</a>
  </dt>

--- a/doc/objects.inv
+++ b/doc/objects.inv
--- a/doc/searchindex.js
+++ b/doc/searchindex.js
--- a/doc/source/cuda/rnn/rnn.html
+++ b/doc/source/cuda/rnn/rnn.html
@@ -1013,8 +1013,8 @@ var _hmt = _hmt || [];
 </dd></dl>
 <dl class="function">
-<dt id="_CPPv222hl_sequence2batch_copyP4realP4realPiiib">
+<dt id="_CPPv222hl_sequence2batch_copyP4realP4realPKiiib">
-<span id="hl_sequence2batch_copy__realP.realP.iP.i.i.b"></span><span class="target" id="paddlehl__sequence_8h_1a6e0b30bd2703b8232ac1d70022306a6a"></span>void <code class="descname">hl_sequence2batch_copy</code><span class="sig-paren">(</span>real *<em>batch</em>, real *<em>sequence</em>, int *<em>batchIndex</em>, int <em>seqWidth</em>, int <em>batchCount</em>, bool <em>seq2batch</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv222hl_sequence2batch_copyP4realP4realPiiib" title="Permalink to this definition">¶</a></dt>
+<span id="hl_sequence2batch_copy__realP.realP.iCP.i.i.b"></span><span class="target" id="paddlehl__sequence_8h_1a13d7f834880527645555849e05278745"></span>void <code class="descname">hl_sequence2batch_copy</code><span class="sig-paren">(</span>real *<em>batch</em>, real *<em>sequence</em>, <em class="property">const</em> int *<em>batchIndex</em>, int <em>seqWidth</em>, int <em>batchCount</em>, bool <em>seq2batch</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv222hl_sequence2batch_copyP4realP4realPKiiib" title="Permalink to this definition">¶</a></dt>
 <dd><p>Memory copy from sequence to batch. </p>
 <p>if seq2batch == true</p>
 <p>copy from sequence to batch: batch[i] = sequence[batchIndex[i]].</p>

--- a/doc/source/gserver/layers/layer.html
+++ b/doc/source/gserver/layers/layer.html
@@ -4623,14 +4623,9 @@ The config file api if gru_step_layer. <dl class="docutils">
 </div>
 <div class="breathe-sectiondef container">
 <p class="breathe-sectiondef-title rubric">Protected Attributes</p>
-<dl class="member">
-<dt id="_CPPv2N6paddle25SequenceScatterAgentLayer17cpuInputStartPos_E">
-<span id="paddle::SequenceScatterAgentLayer::cpuInputStartPos___IVectorPtr"></span><span class="target" id="paddleclasspaddle_1_1SequenceScatterAgentLayer_1a0fd54096dd1552a7e42b9ca1cbbf50ae"></span><a class="reference internal" href="../../math/matrix/matrix.html#_CPPv2N6paddle10IVectorPtrE" title="paddle::IVectorPtr">IVectorPtr</a> <code class="descname">cpuInputStartPos_</code><a class="headerlink" href="#_CPPv2N6paddle25SequenceScatterAgentLayer17cpuInputStartPos_E" title="Permalink to this definition">¶</a></dt>
-<dd></dd></dl>
 <dl class="member">
 <dt id="_CPPv2N6paddle25SequenceScatterAgentLayer14inputStartPos_E">
-<span id="paddle::SequenceScatterAgentLayer::inputStartPos___IVectorPtr"></span><span class="target" id="paddleclasspaddle_1_1SequenceScatterAgentLayer_1a3026312851c50c9952446a7440a07870"></span><a class="reference internal" href="../../math/matrix/matrix.html#_CPPv2N6paddle10IVectorPtrE" title="paddle::IVectorPtr">IVectorPtr</a> <code class="descname">inputStartPos_</code><a class="headerlink" href="#_CPPv2N6paddle25SequenceScatterAgentLayer14inputStartPos_E" title="Permalink to this definition">¶</a></dt>
+<span id="paddle::SequenceScatterAgentLayer::inputStartPos___ICpuGpuVectorPtr"></span><span class="target" id="paddleclasspaddle_1_1SequenceScatterAgentLayer_1acfcf479183ea7b96c05968d2a5ce414e"></span><a class="reference internal" href="../../math/matrix/matrix.html#_CPPv2N6paddle16ICpuGpuVectorPtrE" title="paddle::ICpuGpuVectorPtr">ICpuGpuVectorPtr</a> <code class="descname">inputStartPos_</code><a class="headerlink" href="#_CPPv2N6paddle25SequenceScatterAgentLayer14inputStartPos_E" title="Permalink to this definition">¶</a></dt>
 <dd></dd></dl>
 </div>
@@ -5747,16 +5742,10 @@ sequence is one) to sequence data.&#8221;</p>
 <dd><p>store the ExpandLevel </p>
 </dd></dl>
-<dl class="member">
-<dt id="_CPPv2N6paddle11ExpandLayer19cpuExpandStartsPos_E">
-<span id="paddle::ExpandLayer::cpuExpandStartsPos___IVectorPtr"></span><span class="target" id="paddleclasspaddle_1_1ExpandLayer_1af081e0e3d584c24737be9e9f431914b8"></span><a class="reference internal" href="../../math/matrix/matrix.html#_CPPv2N6paddle10IVectorPtrE" title="paddle::IVectorPtr">IVectorPtr</a> <code class="descname">cpuExpandStartsPos_</code><a class="headerlink" href="#_CPPv2N6paddle11ExpandLayer19cpuExpandStartsPos_E" title="Permalink to this definition">¶</a></dt>
-<dd><p>expanded sequenceStartPositions or subSequenceStartPositions of input[1] </p>
-</dd></dl>
 <dl class="member">
 <dt id="_CPPv2N6paddle11ExpandLayer16expandStartsPos_E">
-<span id="paddle::ExpandLayer::expandStartsPos___IVectorPtr"></span><span class="target" id="paddleclasspaddle_1_1ExpandLayer_1a6528e39055e5de7245e6fcdd83a96c74"></span><a class="reference internal" href="../../math/matrix/matrix.html#_CPPv2N6paddle10IVectorPtrE" title="paddle::IVectorPtr">IVectorPtr</a> <code class="descname">expandStartsPos_</code><a class="headerlink" href="#_CPPv2N6paddle11ExpandLayer16expandStartsPos_E" title="Permalink to this definition">¶</a></dt>
+<span id="paddle::ExpandLayer::expandStartsPos___ICpuGpuVectorPtr"></span><span class="target" id="paddleclasspaddle_1_1ExpandLayer_1a4243425f33452a5d1ac468d257ce111a"></span><a class="reference internal" href="../../math/matrix/matrix.html#_CPPv2N6paddle16ICpuGpuVectorPtrE" title="paddle::ICpuGpuVectorPtr">ICpuGpuVectorPtr</a> <code class="descname">expandStartsPos_</code><a class="headerlink" href="#_CPPv2N6paddle11ExpandLayer16expandStartsPos_E" title="Permalink to this definition">¶</a></dt>
-<dd><p>point to cpuExpandStartsPos_ when useGpu_ is false, copy from cpuExpandStartsPos_ when useGpu_ is true </p>
+<dd><p>expanded sequenceStartPositions or subSequenceStartPositions of input[1] </p>
 </dd></dl>
 </div>

--- a/doc/source/math/matrix/matrix.html
+++ b/doc/source/math/matrix/matrix.html
@@ -1756,8 +1756,8 @@ virtual <span class="target" id="paddleclasspaddle_1_1Matrix_1aa8a2ffb8e06ea97ce
 </dd></dl>
 <dl class="function">
-<dt id="_CPPv2N6paddle6Matrix14copyByRowIndexER6MatrixR7IVector">
+<dt id="_CPPv2N6paddle6Matrix14copyByRowIndexER6MatrixRK7IVector">
-<span id="paddle::Matrix::copyByRowIndex__MatrixR.IVectorR"></span>virtual <span class="target" id="paddleclasspaddle_1_1Matrix_1a4f59ce1c02e1516a53b1dd5247d8d356"></span>void <code class="descname">copyByRowIndex</code><span class="sig-paren">(</span><a class="reference internal" href="#_CPPv2N6paddle6MatrixE" title="paddle::Matrix">Matrix</a> &amp;<em>b</em>, <a class="reference internal" href="#_CPPv2N6paddle7IVectorE" title="paddle::IVector">IVector</a> &amp;<em>rowIndex</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv2N6paddle6Matrix14copyByRowIndexER6MatrixR7IVector" title="Permalink to this definition">¶</a></dt>
+<span id="paddle::Matrix::copyByRowIndex__MatrixR.IVectorCR"></span>virtual <span class="target" id="paddleclasspaddle_1_1Matrix_1ae0eb1e4febf16bff85f1119ba18dedb0"></span>void <code class="descname">copyByRowIndex</code><span class="sig-paren">(</span><a class="reference internal" href="#_CPPv2N6paddle6MatrixE" title="paddle::Matrix">Matrix</a> &amp;<em>b</em>, <em class="property">const</em> <a class="reference internal" href="#_CPPv2N6paddle7IVectorE" title="paddle::IVector">IVector</a> &amp;<em>rowIndex</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv2N6paddle6Matrix14copyByRowIndexER6MatrixRK7IVector" title="Permalink to this definition">¶</a></dt>
 <dd></dd></dl>
 <dl class="function">
@@ -2551,8 +2551,8 @@ where bit(i, j) = ((codes(i) + numClasses) &amp; 2^j) ? 1 : 0
 </dd></dl>
 <dl class="function">
-<dt id="_CPPv2N6paddle9GpuMatrix14copyByRowIndexER6MatrixR7IVector">
+<dt id="_CPPv2N6paddle9GpuMatrix14copyByRowIndexER6MatrixRK7IVector">
-<span id="paddle::GpuMatrix::copyByRowIndex__MatrixR.IVectorR"></span>virtual <span class="target" id="paddleclasspaddle_1_1GpuMatrix_1a46f305110a92a4fbdc98faaba0a35951"></span>void <code class="descname">copyByRowIndex</code><span class="sig-paren">(</span><a class="reference internal" href="#_CPPv2N6paddle6MatrixE" title="paddle::Matrix">Matrix</a> &amp;<em>b</em>, <a class="reference internal" href="#_CPPv2N6paddle7IVectorE" title="paddle::IVector">IVector</a> &amp;<em>rowIndex</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv2N6paddle9GpuMatrix14copyByRowIndexER6MatrixR7IVector" title="Permalink to this definition">¶</a></dt>
+<span id="paddle::GpuMatrix::copyByRowIndex__MatrixR.IVectorCR"></span>virtual <span class="target" id="paddleclasspaddle_1_1GpuMatrix_1a7d757b48a3fcfcd594818c48a88b357c"></span>void <code class="descname">copyByRowIndex</code><span class="sig-paren">(</span><a class="reference internal" href="#_CPPv2N6paddle6MatrixE" title="paddle::Matrix">Matrix</a> &amp;<em>b</em>, <em class="property">const</em> <a class="reference internal" href="#_CPPv2N6paddle7IVectorE" title="paddle::IVector">IVector</a> &amp;<em>rowIndex</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv2N6paddle9GpuMatrix14copyByRowIndexER6MatrixRK7IVector" title="Permalink to this definition">¶</a></dt>
 <dd></dd></dl>
 <dl class="function">
@@ -3111,8 +3111,8 @@ where bit(i, j) = ((codes(i) + numClasses) &amp; 2^j) ? 1 : 0
 <dd></dd></dl>
 <dl class="function">
-<dt id="_CPPv2N6paddle9CpuMatrix14copyByRowIndexER6MatrixR7IVector">
+<dt id="_CPPv2N6paddle9CpuMatrix14copyByRowIndexER6MatrixRK7IVector">
-<span id="paddle::CpuMatrix::copyByRowIndex__MatrixR.IVectorR"></span>virtual <span class="target" id="paddleclasspaddle_1_1CpuMatrix_1a24837d660ae6691384313b5d399ad5fe"></span>void <code class="descname">copyByRowIndex</code><span class="sig-paren">(</span><a class="reference internal" href="#_CPPv2N6paddle6MatrixE" title="paddle::Matrix">Matrix</a> &amp;<em>b</em>, <a class="reference internal" href="#_CPPv2N6paddle7IVectorE" title="paddle::IVector">IVector</a> &amp;<em>rowIndex</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv2N6paddle9CpuMatrix14copyByRowIndexER6MatrixR7IVector" title="Permalink to this definition">¶</a></dt>
+<span id="paddle::CpuMatrix::copyByRowIndex__MatrixR.IVectorCR"></span>virtual <span class="target" id="paddleclasspaddle_1_1CpuMatrix_1a8bfc00af2e2e972193b762434e26d979"></span>void <code class="descname">copyByRowIndex</code><span class="sig-paren">(</span><a class="reference internal" href="#_CPPv2N6paddle6MatrixE" title="paddle::Matrix">Matrix</a> &amp;<em>b</em>, <em class="property">const</em> <a class="reference internal" href="#_CPPv2N6paddle7IVectorE" title="paddle::IVector">IVector</a> &amp;<em>rowIndex</em><span class="sig-paren">)</span><a class="headerlink" href="#_CPPv2N6paddle9CpuMatrix14copyByRowIndexER6MatrixRK7IVector" title="Permalink to this definition">¶</a></dt>
 <dd></dd></dl>
 <dl class="function">

--- a/doc/ui/api/trainer_config_helpers/layers.html
+++ b/doc/ui/api/trainer_config_helpers/layers.html
@@ -901,6 +901,49 @@ will get a warning.</li>
 </div>
 <div class="section" id="recurrent-layer-group">
 <h1>Recurrent Layer Group<a class="headerlink" href="#recurrent-layer-group" title="Permalink to this headline">¶</a></h1>
+<div class="section" id="memory">
+<h2>memory<a class="headerlink" href="#memory" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.trainer_config_helpers.layers.</code><code class="descname">memory</code><span class="sig-paren">(</span><em>name</em>, <em>size</em>, <em>is_seq=False</em>, <em>boot_layer=None</em>, <em>boot_bias=None</em>, <em>boot_bias_active_type=None</em>, <em>boot_with_const_id=None</em><span class="sig-paren">)</span></dt>
+<dd><p>The memory layers is a layer cross each time step. Reference this output
+as previous time step layer <code class="code docutils literal"><span class="pre">name</span></code> &#8216;s output.</p>
+<p>The default memory is zero in first time step, previous time step&#8217;s
+output in the rest time steps.</p>
+<p>If boot_bias, the first time step value is this bias and
+with activation.</p>
+<p>If boot_with_const_id, then the first time stop is a IndexSlot, the
+Arguments.ids()[0] is this <code class="code docutils literal"><span class="pre">cost_id</span></code>.</p>
+<p>If boot_layer is not null, the memory is just the boot_layer&#8217;s output.
+Set <code class="code docutils literal"><span class="pre">is_seq</span></code> is true boot layer is sequence.</p>
+<p>The same name layer in recurrent group will set memory on each time
+step.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
+<li><strong>name</strong> (<em>basestring</em>) &#8211; memory&#8217;s name.</li>
+<li><strong>size</strong> (<em>int</em>) &#8211; size of memory.</li>
+<li><strong>is_seq</strong> (<em>bool</em>) &#8211; is sequence for boot_layer</li>
+<li><strong>boot_layer</strong> (<em>LayerOutput|None</em>) &#8211; boot layer of memory.</li>
+<li><strong>boot_bias</strong> (<em>ParameterAttribute|None</em>) &#8211; boot layer&#8217;s bias</li>
+<li><strong>boot_bias_active_type</strong> (<em>BaseActivation</em>) &#8211; boot layer&#8217;s active type.</li>
+<li><strong>boot_with_const_id</strong> (<em>int</em>) &#8211; boot layer&#8217;s id.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first">LayerOutput object which is a memory.</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">Return type:</th><td class="field-body"><p class="first last">LayerOutput</p>
+</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+</div>
 <div class="section" id="recurrent-group">
 <h2>recurrent_group<a class="headerlink" href="#recurrent-group" title="Permalink to this headline">¶</a></h2>
 <dl class="function">
@@ -2647,6 +2690,7 @@ It is used by recurrent layer group.</p>
 </ul>
 </li>
 <li><a class="reference internal" href="#recurrent-layer-group">Recurrent Layer Group</a><ul>
+<li><a class="reference internal" href="#memory">memory</a></li>
 <li><a class="reference internal" href="#recurrent-group">recurrent_group</a></li>
 <li><a class="reference internal" href="#beam-search">beam_search</a></li>
 <li><a class="reference internal" href="#get-output-layer">get_output_layer</a></li>

--- a/doc/ui/api/trainer_config_helpers/layers_index.html
+++ b/doc/ui/api/trainer_config_helpers/layers_index.html
@@ -114,6 +114,7 @@ var _hmt = _hmt || [];
 </ul>
 </li>
 <li class="toctree-l1"><a class="reference internal" href="layers.html#recurrent-layer-group">Recurrent Layer Group</a><ul>
+<li class="toctree-l2"><a class="reference internal" href="layers.html#memory">memory</a></li>
 <li class="toctree-l2"><a class="reference internal" href="layers.html#recurrent-group">recurrent_group</a></li>
 <li class="toctree-l2"><a class="reference internal" href="layers.html#beam-search">beam_search</a></li>
 <li class="toctree-l2"><a class="reference internal" href="layers.html#get-output-layer">get_output_layer</a></li>

--- a/doc_cn/.buildinfo
+++ b/doc_cn/.buildinfo
 # Sphinx build info version 1
 # This file hashes the configuration used when building these files. When it is not found, a full rebuild will be done.
-config: 70a318b9e7a63a79aedc16f559247671
+config: 8f9e3b6337374f468cc7e48534c4662a
 tags: 645f666f9bcd5a90fca523b33c5a78b7
--- a/doc_cn/_images/graphviz-9be6aad37f57c60f4b971dde0ef44ce27179cf9a.png
+++ b/doc_cn/_images/graphviz-9be6aad37f57c60f4b971dde0ef44ce27179cf9a.png
--- a/doc_cn/_images/graphviz-9be6aad37f57c60f4b971dde0ef44ce27179cf9a.png.map
+++ b/doc_cn/_images/graphviz-9be6aad37f57c60f4b971dde0ef44ce27179cf9a.png.map
+<map id="%3" name="%3">
+</map>
--- a/doc_cn/_sources/algorithm/rnn/hierarchical-layer.txt
+++ b/doc_cn/_sources/algorithm/rnn/hierarchical-layer.txt
+# 支持双层序列作为输入的Layer
+## 概述
+在自然语言处理任务中，序列是一种常见的数据类型。一个独立的词语，可以看作是一个非序列输入，或者，我们称之为一个0层的序列；由词语构成的句子，是一个单层序列；若干个句子构成一个段落，是一个双层的序列。
+双层序列是一个嵌套的序列，它的每一个元素，又是一个单层的序列。这是一种非常灵活的数据组织方式，帮助我们构造一些复杂的输入信息。
+我们可以按照如下层次定义非序列，单层序列，以及双层序列。
+ 0层序列：一个独立的元素，类型可以是PaddlePaddle支持的任意输入数据类型
+ 单层序列：排成一列的多个元素，每个元素是一个0层序列，元素之间的顺序是重要的输入信息
+ 双层序列：排成一列的多个元素，每个元素是一个单层序列，称之为双层序列的一个子序列（subseq），subseq的每个元素是一个0层序列
+在 PaddlePaddle中，下面这些Layer能够接受双层序列作为输入，完成相应的计算。
+## pooling_layer
+pooling_layer的使用示例如下，详细见<a href = "../../../doc/ui/api/trainer_config_helpers/layers.html#pooling-layer">配置API</a>。
+```python
+seq_pool = pooling_layer(input=layer,
+                         pooling_type=AvgPooling(),
+                         agg_level=AggregateLevel.EACH_SEQUENCE)
+```
+- `pooling_type` 目前支持两种，分别是：MaxPooling()和AvgPooling()。
+- `agg_level=AggregateLevel.TIMESTEP`时（默认值）：
+  - 作用：双层序列经过运算变成一个0层序列，或单层序列经过运算变成一个0层序列
+  - 输入：一个双层序列，或一个单层序列
+  - 输出：一个0层序列，即整个输入序列（单层或双层）的平均值（或最大值）
+- `agg_level=AggregateLevel.EACH_SEQUENCE`时：
+  - 作用：一个双层序列经过运算变成一个单层序列
+  - 输入：必须是一个双层序列
+  - 输出：一个单层序列，序列的每个元素是原来双层序列每个subseq元素的平均值（或最大值）
+## last_seq 和 first_seq
+last_seq的使用示例如下（first_seq类似），详细见<a href = "../../../doc/ui/api/trainer_config_helpers/layers.html#last-seq">配置API</a>。
+```python
+last = last_seq(input=layer,
+                agg_level=AggregateLevel.EACH_SEQUENCE)
+```
+- `agg_level=AggregateLevel.TIMESTEP`时（默认值）：
+  - 作用：一个双层序列经过运算变成一个0层序列，或一个单层序列经过运算变成一个0层序列
+  - 输入：一个双层序列或一个单层序列
+  - 输出：一个0层序列，即整个输入序列（双层或者单层）最后一个，或第一个元素。
+- `agg_level=AggregateLevel.EACH_SEQUENCE`时：
+  - 作用：一个双层序列经过运算变成一个单层序列
+  - 输入：必须是一个双层序列
+  - 输出：一个单层序列，其中每个元素是双层序列中每个subseq最后一个（或第一个）元素。
+## expand_layer
+expand_layer的使用示例如下，详细见<a href = "../../../doc/ui/api/trainer_config_helpers/layers.html#expand-layer">配置API</a>。
+```python
+expand = expand_layer(input=layer1,
+                      expand_as=layer2,
+                      expand_level=ExpandLevel.FROM_TIMESTEP)
+```
+- `expand_level=ExpandLevel.FROM_TIMESTEP`时（默认值）：
+  - 作用：一个0层序列经过运算扩展成一个单层序列，或者一个双层序列
+  - 输入：layer1必须是一个0层序列，是待扩展的数据；layer2可以是一个单层序列，或者是一个双层序列，提供扩展的长度信息
+  - 输出：一个单层序列，或一个双层序列，输出序列的类型（双层序列，或单层序列）和序列中含有元素的数目同 layer2一致。若输出是单层序列，单层序列的每个元素（0层序列），都是对layer1元素的拷贝；若输出是双层序列，双层序列每个subseq中每个元素（0层序列），都是对layer1元素的拷贝
+- `expand_level=ExpandLevel.FROM_SEQUENCE`时：
+  - 作用：一个单层序列经过运算扩展成一个双层序列
+  - 输入：layer1必须是一个单层序列，是待扩展的数据；layer2必须是一个双层序列，提供扩展的长度信息
+  - 输出：一个双层序列，序列中含有元素的数目同layer2一致。要求单层序列含有元素的数目（0层序列），和双层序列含有subseq 的数目一致。单层序列第i个元素（0层序列），被扩展为一个单层序列，构成了输出双层序列的第i个subseq。
\ No newline at end of file
--- a/doc_cn/_sources/algorithm/rnn/hierarchical-rnn.txt
+++ b/doc_cn/_sources/algorithm/rnn/hierarchical-rnn.txt
+# 双层RNN配置与示例
+我们在`paddle/gserver/tests/test_RecurrentGradientMachine`单测中，通过多组语义相同的单双层RNN配置，讲解如何使用双层RNN。
+## 示例1：双进双出，subseq间无memory
+配置：单层RNN（`sequence_layer_group`）和双层RNN（`sequence_nest_layer_group`），语义完全相同。
+### 读取双层序列的方法
+首先，我们看一下单双层序列的不同数据组织形式（您也可以采用别的组织形式）：
+- 单层序列的数据（`Sequence/tour_train_wdseg`）如下，一共有10个样本。每个样本由两部分组成，一个label（此处都为2）和一个已经分词后的句子。
+```text
+2  	酒店 有 很 舒适 的 床垫 子 ， 床上用品 也 应该 是 一人 一 换 ， 感觉 很 利落 对 卫生 很 放心 呀 。
+2  	很 温馨 ， 也 挺 干净 的 * 地段 不错 ， 出来 就 有 全家 ， 离 地铁站 也 近 ， 交通 很方便 * 就是 都 不 给 刷牙 的 杯子 啊 ， 就 第一天 给 了 一次性杯子 *
+2  	位置 方便 ， 强烈推荐 ， 十一 出去玩 的 时候 选 的 ， 对面 就是 华润万家 ， 周围 吃饭 的 也 不少 。
+2  	交通便利 ， 吃 很 便利 ， 乾 浄 、 安静 ， 商务 房 有 电脑 、 上网 快 ， 价格 可以 ， 就 早餐 不 好吃 。 整体 是 不错 的 。 適 合 出差 來 住 。
+2  	本来 准备 住 两 晚 ， 第 2 天 一早 居然 停电 ， 且 无 通知 ， 只有 口头 道歉 。 总体来说 性价比 尚可 ， 房间 较 新 ， 还是 推荐 .
+2  	这个 酒店 去过 很多 次 了 ， 选择 的 主要原因 是 离 客户 最 便宜 相对 又 近 的 酒店
+2  	挺好 的 汉庭 ， 前台 服务 很 热情 ， 卫生 很 整洁 ， 房间 安静 ， 水温 适中 ， 挺好 ！
+2  	HowardJohnson 的 品质 ， 服务 相当 好 的 一 家 五星级 。 房间 不错 、 泳池 不错 、 楼层 安排 很 合理 。 还有 就是 地理位置 ， 简直 一 流 。 就 在 天一阁 、 月湖 旁边 ， 离 天一广场 也 不远 。 下次 来 宁波 还会 住 。
+2  	酒店 很干净 ， 很安静 ， 很 温馨 ， 服务员 服务 好 ， 各方面 都 不错 *
+2  	挺好 的 ， 就是 没 窗户 ， 不过 对 得 起 这 价格
+```
+- 双层序列的数据（`Sequence/tour_train_wdseg.nest`）如下，一共有4个样本。样本间用空行分开，代表不同的双层序列，序列数据和上面的完全一样。每个样本的子句数分别为2,3,2,3。
+```text
+2  	酒店 有 很 舒适 的 床垫 子 ， 床上用品 也 应该 是 一人 一 换 ， 感觉 很 利落 对 卫生 很 放心 呀 。
+2  	很 温馨 ， 也 挺 干净 的 * 地段 不错 ， 出来 就 有 全家 ， 离 地铁站 也 近 ， 交通 很方便 * 就是 都 不 给 刷牙 的 杯子 啊 ， 就 第一天 给 了 一次性杯子 *
+2  	位置 方便 ， 强烈推荐 ， 十一 出去玩 的 时候 选 的 ， 对面 就是 华润万家 ， 周围 吃饭 的 也 不少 。
+2  	交通便利 ， 吃 很 便利 ， 乾 浄 、 安静 ， 商务 房 有 电脑 、 上网 快 ， 价格 可以 ， 就 早餐 不 好吃 。 整体 是 不错 的 。 適 合 出差 來 住 。
+2  	本来 准备 住 两 晚 ， 第 2 天 一早 居然 停电 ， 且 无 通知 ， 只有 口头 道歉 。 总体来说 性价比 尚可 ， 房间 较 新 ， 还是 推荐 .
+2  	这个 酒店 去过 很多 次 了 ， 选择 的 主要原因 是 离 客户 最 便宜 相对 又 近 的 酒店
+2  	挺好 的 汉庭 ， 前台 服务 很 热情 ， 卫生 很 整洁 ， 房间 安静 ， 水温 适中 ， 挺好 ！
+2  	HowardJohnson 的 品质 ， 服务 相当 好 的 一 家 五星级 。 房间 不错 、 泳池 不错 、 楼层 安排 很 合理 。 还有 就是 地理位置 ， 简直 一 流 。 就 在 天一阁 、 月湖 旁边 ， 离 天一广场 也 不远 。 下次 来 宁波 还会 住 。
+2  	酒店 很干净 ， 很安静 ， 很 温馨 ， 服务员 服务 好 ， 各方面 都 不错 *
+2  	挺好 的 ， 就是 没 窗户 ， 不过 对 得 起 这 价格
+```
+其次，我们看一下单双层序列的不同dataprovider（见`sequenceGen.py`）：
+- 单层序列的dataprovider如下：
+  - word_slot是integer_value_sequence类型，代表单层序列。
+  - label是integer_value类型，代表一个向量。
+```python
+def hook(settings, dict_file, **kwargs):
+    settings.word_dict = dict_file
+    settings.input_types = [integer_value_sequence(len(settings.word_dict)), 
+                            integer_value(3)]
+@provider(init_hook=hook)
+def process(settings, file_name):
+    with open(file_name, 'r') as fdata:
+        for line in fdata:
+            label, comment = line.strip().split('\t')
+            label = int(''.join(label.split()))
+            words = comment.split()
+            word_slot = [settings.word_dict[w] for w in words if w in settings.word_dict]
+            yield word_slot, label
+```
+- 双层序列的dataprovider如下：
+  - word_slot是integer_value_sub_sequence类型，代表双层序列。
+  - label是integer_value_sequence类型，代表单层序列，即一个子句一个label。注意：也可以为integer_value类型，代表一个向量，即一个句子一个label。通常根据任务需求进行不同设置。
+  - 关于dataprovider中input_types的详细用法，参见PyDataProvider2。
+```python
+def hook2(settings, dict_file, **kwargs):
+    settings.word_dict = dict_file
+    settings.input_types = [integer_value_sub_sequence(len(settings.word_dict)),
+                            integer_value_sequence(3)]
+@provider(init_hook=hook2)
+def process2(settings, file_name):
+    with open(file_name) as fdata:
+        label_list = []
+        word_slot_list = []
+        for line in fdata:
+            if (len(line)) > 1:
+                label,comment = line.strip().split('\t')
+                label = int(''.join(label.split()))
+                words = comment.split()
+                word_slot = [settings.word_dict[w] for w in words if w in settings.word_dict]
+                label_list.append(label)
+                word_slot_list.append(word_slot)
+            else:
+                yield word_slot_list, label_list
+                label_list = []
+                word_slot_list = []
+```
+### 模型中的配置
+首先，我们看一下单层序列的配置（见`sequence_layer_group.conf`）。注意：batchsize=5表示一次过5句单层序列，因此2个batch就可以完成1个pass。
+```python
+settings(batch_size=5)
+data = data_layer(name="word", size=dict_dim)
+emb = embedding_layer(input=data, size=word_dim)
+# (lstm_input + lstm) is equal to lstmemory 
+with mixed_layer(size=hidden_dim*4) as lstm_input:
+    lstm_input += full_matrix_projection(input=emb)
+lstm = lstmemory_group(input=lstm_input,
+                       size=hidden_dim,
+                       act=TanhActivation(),
+                       gate_act=SigmoidActivation(),
+                       state_act=TanhActivation(),
+                       lstm_layer_attr=ExtraLayerAttribute(error_clipping_threshold=50))
+lstm_last = last_seq(input=lstm)
+with mixed_layer(size=label_dim, 
+                 act=SoftmaxActivation(), 
+                 bias_attr=True) as output:
+    output += full_matrix_projection(input=lstm_last)
+outputs(classification_cost(input=output, label=data_layer(name="label", size=1)))
+```
+其次，我们看一下语义相同的双层序列配置（见`sequence_nest_layer_group.conf`），并对其详细分析：
+- batchsize=2表示一次过2句双层序列。但从上面的数据格式可知，2句双层序列和5句单层序列的数据完全一样。
+- data_layer和embedding_layer不关心数据是否是序列格式，因此两个配置在这两层上的输出是一样的。
+- lstmemory:
+  - 单层序列过了一个mixed_layer和lstmemory_group。
+  - 双层序列在同样的mixed_layer和lstmemory_group外，直接加了一层group。由于这个外层group里面没有memory，表示subseq间不存在联系，即起到的作用仅仅是把双层seq拆成单层，因此双层序列过完lstmemory的输出和单层的一样。
+- last_seq：
+  - 单层序列直接取了最后一个元素
+  - 双层序列首先（last_seq层）取了每个subseq的最后一个元素，将其拼接成一个新的单层序列；接着（expand_layer层）将其扩展成一个新的双层序列，其中第i个subseq中的所有向量均为输入的单层序列中的第i个向量；最后（average_layer层）取了每个subseq的平均值。
+  - 分析得出：第一个last_seq后，每个subseq的最后一个元素就等于单层序列的最后一个元素，而expand_layer和average_layer后，依然保持每个subseq最后一个元素的值不变（这两层仅是为了展示它们的用法，实际中并不需要）。因此单双层序列的输出是一样旳。
+```python
+settings(batch_size=2)
+data = data_layer(name="word", size=dict_dim)
+emb_group = embedding_layer(input=data, size=word_dim)
+# (lstm_input + lstm) is equal to lstmemory 
+def lstm_group(lstm_group_input):
+    with mixed_layer(size=hidden_dim*4) as group_input:
+      group_input += full_matrix_projection(input=lstm_group_input)
+    lstm_output = lstmemory_group(input=group_input,
+                                  name="lstm_group",
+                                  size=hidden_dim,
+                                  act=TanhActivation(),
+                                  gate_act=SigmoidActivation(),
+                                  state_act=TanhActivation(),
+                                  lstm_layer_attr=ExtraLayerAttribute(error_clipping_threshold=50))
+    return lstm_output
+lstm_nest_group = recurrent_group(input=SubsequenceInput(emb_group),
+                                  step=lstm_group,
+                                  name="lstm_nest_group")
+# hasSubseq ->(seqlastins) seq
+lstm_last = last_seq(input=lstm_nest_group, agg_level=AggregateLevel.EACH_SEQUENCE)
+# seq ->(expand) hasSubseq
+lstm_expand = expand_layer(input=lstm_last, expand_as=emb_group, expand_level=ExpandLevel.FROM_SEQUENCE)
+# hasSubseq ->(average) seq
+lstm_average = pooling_layer(input=lstm_expand,
+                             pooling_type=AvgPooling(),
+                             agg_level=AggregateLevel.EACH_SEQUENCE)
+with mixed_layer(size=label_dim, 
+                 act=SoftmaxActivation(), 
+                 bias_attr=True) as output:
+    output += full_matrix_projection(input=lstm_average)
+outputs(classification_cost(input=output, label=data_layer(name="label", size=1)))
+```
+## 示例2：双进双出，subseq间有memory
+配置：单层RNN（`sequence_rnn.conf`），双层RNN（`sequence_nest_rnn.conf`和`sequence_nest_rnn_readonly_memory.conf`），语义完全相同。
+### 读取双层序列的方法
+我们看一下单双层序列的不同数据组织形式和dataprovider（见`rnn_data_provider.py`）
+```python
+data = [
+    [[[1, 3, 2], [4, 5, 2]], 0],
+    [[[0, 2], [2, 5], [0, 1, 2]], 1],
+]
+@provider(input_types=[integer_value_sub_sequence(10),
+                       integer_value(3)])
+def process_subseq(settings, file_name):
+    for d in data:
+        yield d
+@provider(input_types=[integer_value_sequence(10),
+                       integer_value(3)])
+def process_seq(settings, file_name):
+    for d in data:
+        seq = []
+```
+- 单层序列：有两句，分别为[1,3,2,4,5,2]和[0,2,2,5,0,1,2]。
+- 双层序列：有两句，分别为[[1,3,2],[4,5,2]]（2个子句）和[[0,2],[2,5],[0,1,2]]（3个子句）。
+- 单双层序列的label都分别是0和1
+### 模型中的配置
+我们选取单双层序列配置中的不同部分，来对比分析两者语义相同的原因。
+- 单层序列：过了一个很简单的recurrent_group。每一个时间步，当前的输入y和上一个时间步的输出rnn_state做了一个全链接。
+```python
+def step(y):
+    mem = memory(name="rnn_state", size=hidden_dim)
+    return fc_layer(input=[y, mem],
+                    size=hidden_dim,
+                    act=TanhActivation(),
+                    bias_attr=True,
+                    name="rnn_state")
+out = recurrent_group(step=step, input=emb)
+```
+- 双层序列，外层memory是一个元素：
+  - 内层inner_step的recurrent_group和单层序列的几乎一样。除了boot_layer=outer_mem，表示将外层的outer_mem作为内层memory的初始状态。外层outer_step中，outer_mem是一个子句的最后一个向量，即整个双层group是将前一个子句的最后一个向量，作为下一个子句memory的初始状态。
+  - 从输入数据上看，单双层序列的句子是一样的，只是双层序列将其又做了子序列划分。因此双层序列的配置中，必须将前一个子句的最后一个元素，作为boot_layer传给下一个子句的memory，才能保证和单层序列的配置中“每一个时间步都用了上一个时间步的输出结果”一致。
+```python
+def outer_step(x):
+    outer_mem = memory(name="outer_rnn_state", size=hidden_dim)
+    def inner_step(y):
+        inner_mem = memory(name="inner_rnn_state",
+                           size=hidden_dim,
+                           boot_layer=outer_mem)
+        return fc_layer(input=[y, inner_mem],
+                        size=hidden_dim,
+                        act=TanhActivation(),
+                        bias_attr=True,
+                        name="inner_rnn_state")
+    inner_rnn_output = recurrent_group(
+        step=inner_step,
+        input=x)
+    last = last_seq(input=inner_rnn_output, name="outer_rnn_state")
+    return inner_rnn_output
+out = recurrent_group(step=outer_step, input=SubsequenceInput(emb))
+```
+- 双层序列，外层memory是单层序列：
+  - 由于外层每个时间步返回的是一个子句，这些子句的长度往往不等长。因此当外层有is_seq=True的memory时，内层是**无法直接使用**它的，即内层memory的boot_layer不能链接外层的这个memory。
+  - 如果内层memory想**间接使用**这个外层memory，只能通过`pooling_layer`、`last_seq`或`first_seq`这三个layer将它先变成一个元素。但这种情况下，外层memory必须有boot_layer，否则在第0个时间步时，由于外层memory没有任何seq信息，因此上述三个layer的前向会报出“**Check failed: input.sequenceStartPositions**”的错误。
+## 示例3：双进双出，输入不等长
+TBD
+## 示例4：beam_search的生成
+TBD
\ No newline at end of file
--- a/doc_cn/_sources/algorithm/rnn/rnn-tutorial.txt
+++ b/doc_cn/_sources/algorithm/rnn/rnn-tutorial.txt
+# Recurrent Group教程
+## 概述
+序列数据是自然语言处理任务面对的一种主要输入数据类型。
+一句话是由词语构成的序列，多句话进一步构成了段落。因此，段落可以看作是一个嵌套的双层的序列，这个序列的每个元素又是一个序列。
+双层序列是PaddlePaddle支持的一种非常灵活的数据组织方式，帮助我们更好地描述段落、多轮对话等更为复杂的语言数据。基于双层序列输入，我们可以设计搭建一个灵活的、层次化的RNN，分别从词语和句子级别编码输入数据，同时也能够引入更加复杂的记忆机制，更好地完成一些复杂的语言理解任务。
+在PaddlePaddle中，`recurrent_group`是一种任意复杂的RNN单元，用户只需定义RNN在一个时间步内完成的计算，PaddlePaddle负责完成信息和误差在时间序列上的传播。
+更进一步，`recurrent_group`同样可以扩展到双层序列的处理上。通过两个嵌套的`recurrent_group`分别定义子句级别和词语级别上需要完成的运算，最终实现一个层次化的复杂RNN。
+目前，在PaddlePaddle中，能够对双向序列进行处理的有`recurrent_group`和部分Layer，具体可参考文档：<a href = "hierarchical-layer.html">支持双层序列作为输入的Layer</a>。
+## 相关概念
+### 基本原理
+`recurrent_group` 是PaddlePaddle支持的一种任意复杂的RNN单元。使用者只需要关注于设计RNN在一个时间步之内完成的计算，PaddlePaddle负责完成信息和梯度在时间序列上的传播。
+PaddlePaddle中，`recurrent_group`的一个简单调用如下：
+``` python
+recurrent_group(step, input, reverse)
+```
+- step：一个可调用的函数，定义一个时间步之内RNN单元完成的计算
+- input：输入，必须是一个单层序列，或者一个双层序列
+- reverse：是否以逆序处理输入序列
+使用`recurrent_group`的核心是设计step函数的计算逻辑。step函数内部可以自由组合PaddlePaddle支持的各种layer，完成任意的运算逻辑。`recurrent_group` 的输入（即input）会成为step函数的输入，由于step 函数只关注于RNN一个时间步之内的计算，在这里`recurrent_group`替我们完成了原始输入数据的拆分。
+### 输入
+`recurrent_group`处理的输入序列主要分为以下三种类型：
+- **数据输入**：一个双层序列进入`recurrent_group`会被拆解为一个单层序列，一个单层序列进入`recurrent_group`会被拆解为非序列，然后交给step函数，这一过程对用户是完全透明的。可以有以下两种：1）通过data_layer拿到的用户输入；2）其它layer的输出。
+- **只读Memory输入**：`StaticInput` 定义了一个只读的Memory，由`StaticInput`指定的输入不会被`recurrent_group`拆解，`recurrent_group` 循环展开的每个时间步总是能够引用所有输入，可以是一个非序列，或者一个单层序列。
+- **序列生成任务的输入**：`GeneratedInput`只用于在序列生成任务中指定输入数据。
+### 输入示例
+序列生成任务大多遵循encoder-decoer架构，encoder和decoder可以是能够处理序列的任意神经网络单元，而RNN是最流行的选择。
+给定encoder输出和当前词，decoder每次预测产生下一个最可能的词语。在这种结构中，decoder接受两个输入：
+- 要生成的目标序列：是decoder的数据输入，也是decoder循环展开的依据，`recurrent_group`会对这类输入进行拆解。
+- encoder输出，可以是一个非序列，或者一个单层序列：是一个unbounded memory，decoder循环展开的每一个时间步会引用全部结果，不应该被拆解，这种类型的输入必须通过`StaticInput`指定。关于Unbounded Memory的更多讨论请参考论文 [Neural Turning Machine](https://arxiv.org/abs/1410.5401)。
+在序列生成任务中，decoder RNN总是引用上一时刻预测出的词的词向量，作为当前时刻输入。`GeneratedInput`自动完成这一过程。
+### 输出
+`step`函数必须返回一个或多个Layer的输出，这个Layer的输出会作为整个`recurrent_group` 最终的输出结果。在输出的过程中，`recurrent_group` 会将每个时间步的输出拼接，这个过程对用户也是透明的。
+### memory
+memory只能在`recurrent_group`中定义和使用。memory不能独立存在，必须指向一个PaddlePaddle定义的Layer。引用memory得到这layer上一时刻输出，因此，可以将memory理解为一个时延操作。
+可以显示地指定一个layer的输出用于初始化memory。不指定时，memory默认初始化为0。
+## 双层RNN介绍
+`recurrent_group`帮助我们完成对输入序列的拆分，对输出的合并，以及计算逻辑在序列上的循环展开。
+利用这种特性，两个嵌套的`recurrent_group`能够处理双层序列，实现词语和句子两个级别的双层RNN结构。
+- 单层（word-level）RNN：每个状态（state）对应一个词（word）。
+- 双层（sequence-level）RNN：一个双层RNN由多个单层RNN组成，每个单层RNN（即双层RNN的每个状态）对应一个子句（subseq）。
+为了描述方便，下文以NLP任务为例，将含有子句（subseq）的段落定义为一个双层序列，将含有词语的句子定义为一个单层序列，那么0层序列即为一个词语。
+## 双层RNN的使用
+### 训练流程的使用方法
+使用 `recurrent_group`需要遵循以下约定：
+- **单进单出**：输入和输出都是单层序列。
+  - 如果有多个输入，不同输入序列含有的词语数必须严格相等。
+  - 输出一个单层序列，输出序列的词语数和输入序列一致。
+  - memory：在step函数中定义 memory指向一个layer，通过引用memory得到这个layer上一个时刻输出，形成recurrent 连接。memory的is_seq参数必须为false。如果没有定义memory，每个时间步之内的运算是独立的。
+  - boot_layer：memory的初始状态，默认初始状为0，memory的is_seq参数必须为false。
+- **双进双出**：输入和输出都是双层序列。
+  - 如果有多个输入序列，不同输入含有的子句（subseq）数必须严格相等，但子句含有的词语数可以不相等。
+  - 输出一个双层序列，子句（subseq）数、子句的单词数和指定的一个输入序列一致，默认为第一个输入。
+  - memory：在step函数中定义memory，指向一个layer，通过引用memory得到这个layer上一个时刻的输出，形成recurrent连接。定义在外层`recurrent_group` step函数中的memory，能够记录上一个subseq 的状态，可以是一个单层序列（只作为read-only memory），也可以是一个词语。如果没有定义memory，那么 subseq 之间的运算是独立的。
+  - boot_layer：memory 初始状态，可以是一个单层序列（只作为read-only memory）或一个向量。默认不设置，即初始状态为0。
+- **双进单出**：目前还未支持，会报错"In hierachical RNN, all out links should be from sequences now"。
+### 生成流程的使用方法
+使用`beam_search`需要遵循以下约定：
+- 单层RNN：从一个word生成下一个word。
+- 双层RNN：即把单层RNN生成后的subseq给拼接成一个新的双层seq。从语义上看，也不存在一个subseq直接生成下一个subseq的情况。
\ No newline at end of file
--- a/doc_cn/_sources/faq/index.txt
+++ b/doc_cn/_sources/faq/index.txt
+####################
+PaddlePaddle常见问题
+####################
+..  contents::
+1. 如何减少PaddlePaddle的内存占用
+---------------------------------
+神经网络的训练本身是一个非常消耗内存和显存的工作。经常会消耗数十G的内存和数G的显存。
+PaddlePaddle的内存占用主要分为如下几个方面\:
+* DataProvider缓冲池内存 (只针对内存)
+* 神经元激活内存 （针对内存和显存）
+* 参数内存 (针对内存和显存)
+* 其他内存杂项
+这其中，其他内存杂项是指PaddlePaddle本身所用的一些内存，包括字符串分配，临时变量等等，
+这些内存就不考虑如何缩减了。
+其他的内存的减少方法依次为
+减少DataProvider缓冲池内存
++++++++++++++++++++++++++
+PyDataProvider使用的是异步加载，同时在内存里直接随即选取数据来做Shuffle。即
+..  graphviz::
+    digraph {
+        rankdir=LR;
+        数据文件 -> 内存池 -> PaddlePaddle训练
+    }
+所以，减小这个内存池即可减小内存占用，同时也可以加速开始训练前数据载入的过程。但是，这
+个内存池实际上决定了shuffle的粒度。所以，如果将这个内存池减小，又要保证数据是随机的，
+那么最好将数据文件在每次读取之前做一次shuffle。可能的代码为
+..  literalinclude:: reduce_min_pool_size.py
+这样做可以极大的减少内存占用，并且可能会加速训练过程。 详细文档参考 `这里
+<../ui/data_provider/pydataprovider2.html#provider>`_ 。
+神经元激活内存
++++++++++++++
+神经网络在训练的时候，会对每一个激活暂存一些数据，包括激活，參差等等。
+在反向传递的时候，这些数据会被用来更新参数。这些数据使用的内存主要和两个参数有关系，
+一是batch size，另一个是每条序列(Sequence)长度。所以，其实也是和每个mini-batch中包含
+的时间步信息成正比。
+所以，做法可以有两种。他们是
+* 减小batch size。 即在网络配置中 :code:`settings(batch_size=1000)` 设置成一个小一些的值。但是batch size本身是神经网络的超参数，减小batch size可能会对训练结果产生影响。
+* 减小序列的长度，或者直接扔掉非常长的序列。比如，一个数据集大部分序列长度是100-200,
+  但是突然有一个10000长的序列，就很容易导致内存超限。特别是在LSTM等RNN中。
+参数内存
++++++++
+PaddlePaddle支持非常多的优化算法(Optimizer)，不同的优化算法需要使用不同大小的内存。
+例如如果使用 :code:`adadelta` 算法，则需要使用参数规模大约5倍的内存。 如果参数保存下来的
+文件为 :code:`100M`， 那么该优化算法至少需要 :code:`500M` 的内存。
+可以考虑使用一些优化算法，例如 :code:`momentum`。
+2. 如何加速PaddlePaddle的训练速度
+---------------------------------
+PaddlePaddle是神经网络训练平台，加速PaddlePaddle训练有如下几个方面\：
+* 减少数据载入的耗时
+* 加速训练速度
+* 利用更多的计算资源
+减少数据载入的耗时
++++++++++++++++++
+使用 :code:`pydataprovider`时，可以减少缓存池的大小，同时设置内存缓存功能，即可以极大的加速数据载入流程。
+:code:`DataProvider` 缓存池的减小，和之前减小通过减小缓存池来减小内存占用的原理一致。
+..  literalinclude:: reduce_min_pool_size.py
+同时 :code:`@provider` 接口有一个 :code:`cache` 参数来控制缓存方法，将其设置成 :code:`CacheType.CACHE_PASS_IN_MEM` 的话，会将第一个 :code:`pass` (过完所有训练数据即为一个pass)生成的数据缓存在内存里，在之后的 :code:`pass` 中，不会再从 :code:`python` 端读取数据，而是直接从内存的缓存里读取数据。这也会极大减少数据读入的耗时。
+加速训练速度
++++++++++++
+PaddlePaddle支持Sparse的训练，sparse训练需要训练特征是 :code:`sparse_binary_vector` 、 :code:`sparse_vector` 、或者 :code:`integer_value` 的任一一种。同时，与这个训练数据交互的Layer，需要将其Parameter设置成 sparse 更新模式，即设置 :code:`sparse_update=True`
+这里使用简单的 :code:`word2vec` 训练语言模型距离，具体使用方法为\:
+使用一个词前两个词和后两个词，来预测这个中间的词。这个任务的DataProvider为\:
+..  literalinclude:: word2vec_dataprovider.py
+这个任务的配置为\:
+..  literalinclude:: word2vec_config.py
+更多关于sparse训练的内容请参考 `sparse训练的文档 <TBD>`_
+利用更多的计算资源
++++++++++++++++++
+利用更多的计算资源可以分为一下几个方式来进行\:
+* 单机CPU训练
+  * 使用多线程训练。设置命令行参数 :code:`trainer_count`，即可以设置参与训练的线程数量。使用方法为 :code:`paddle train --trainer_count=4`
+* 单机GPU训练
+  * 使用显卡训练。设置命令行参数 :code:`use_gpu`。 使用方法为 :code:`paddle train --use_gpu=true`
+  * 使用多块显卡训练。设置命令行参数 :code:`use_gpu` 和 :code:`trainer_count`。使用 :code:`--use_gpu=True` 开启GPU训练，使用 :code:`trainer_count` 指定显卡数量。使用方法为 :code:`paddle train --use_gpu=true --trainer_count=4`
+* 多机训练
+  * 使用多机训练的方法也比较简单，需要先在每个节点启动 :code:`paddle pserver`，在使用 :code:`paddle train --pservers=192.168.100.1,192.168.100.2` 来指定每个pserver的ip地址
+  * 具体的多机训练方法参考 `多机训练 <TBD>`_ 文档。
+3. 遇到“非法指令”或者是“illegal instruction” 
+--------------------------------------------
+paddle在进行计算的时候为了提升计算性能，使用了avx指令。部分老的cpu型号无法支持这样的指令。通常来说执行下grep avx /proc/cpuinfo看看是否有输出即可知道是否支持。（另：用此方法部分虚拟机可能检测到支持avx指令但是实际运行会挂掉，请当成是不支持，看下面的解决方案）
+解决办法是\:
+* 使用 NO_AVX的 `安装包 <../build_and_install/index.html>`_ 或者 `Docker image <../build_and_install/install/docker_install.html>`_
+* 或者，使用 :code:`-DWITH_AVX=OFF` 重新编译PaddlePaddle。
+4. 如何选择SGD算法的学习率
+--------------------------
+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。
+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。
+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:`0.2, 0.5, 0.3` , 那么常数输出所能达到的最小cost是 :code:`-(0.2*log(0.2)+0.5*log(0.5)+0.3*log(0.3))=1.03` 。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。
+5. 如何初始化参数
+-----------------
+默认情况下，PaddlePaddle使用均值0，标准差为 :math:`\frac{1}{\sqrt{d}}` 来初始化参数。其中 :math:`d` 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\:
+* 高斯分布。将 :code:`param_attr` 设置成 :code:`param_attr=ParamAttr(initial_mean=0.0, initial_std=1.0)`
+* 均匀分布。将 :code:`param_attr` 设置成 :code:`param_attr=ParamAttr(initial_max=1.0, initial_min=-1.0)`
+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。
+..  code-block:: python
+    hidden = fc_layer(input=ipt, param_attr=ParamAttr(initial_max=1.0, initial_min=-1.0), 
+                      bias_attr=ParamAttr(initial_mean=1.0, initial_std=0.0))
+上述代码将bias全部初始化为1.0, 同时将参数初始化为 :code:`[1.0, -1.0]` 的均匀分布。
+6. 如何共享参数
+---------------
+PaddlePaddle的参数使用名字 :code:`name` 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:`ParamAttr(name="YOUR_PARAM_NAME")` 来设置。更方便的设置方式，是想要共享的参数使用同样的 :code:`ParamAttr` 对象。
+简单的全连接网络，参数共享的配置示例为\:
+..  literalinclude:: ../../python/paddle/trainer_config_helpers/tests/configs/shared_fc.py
+这里 :code:`hidden_a` 和 :code:`hidden_b` 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:`softmax_param`。
--- a/doc_cn/_sources/index.txt
+++ b/doc_cn/_sources/index.txt
@@ -3,6 +3,7 @@ PaddlePaddle文档
 使用指南
 --------
 * `快速入门 <demo/quick_start/index.html>`_
 * `编译与安装 <build_and_install/index.html>`_
 * `用户接口 <ui/index.html>`_
@@ -16,4 +17,13 @@ PaddlePaddle文档
 算法教程
 --------
-* `RNN配置 <../doc/algorithm/rnn/rnn.html>`_
+* `Recurrent Group教程 <algorithm/rnn/rnn-tutorial.html>`_
+* `单层RNN示例 <../doc/algorithm/rnn/rnn.html>`_
+* `双层RNN示例 <algorithm/rnn/hierarchical-rnn.html>`_
+* `支持双层序列作为输入的Layer <algorithm/rnn/hierarchical-layer.html>`_
+常见问题
+--------
+* `常见问题 <faq/index.html>`_
--- a/doc_cn/algorithm/rnn/hierarchical-layer.html
+++ b/doc_cn/algorithm/rnn/hierarchical-layer.html
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+  "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+  <head>
+    <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+    <title>支持双层序列作为输入的Layer &#8212; PaddlePaddle  documentation</title>
+    <link rel="stylesheet" href="../../_static/classic.css" type="text/css" />
+    <link rel="stylesheet" href="../../_static/pygments.css" type="text/css" />
+    <script type="text/javascript">
+      var DOCUMENTATION_OPTIONS = {
+        URL_ROOT:    '../../',
+        VERSION:     '',
+        COLLAPSE_INDEX: false,
+        FILE_SUFFIX: '.html',
+        HAS_SOURCE:  true
+      };
+    </script>
+    <script type="text/javascript" src="../../_static/jquery.js"></script>
+    <script type="text/javascript" src="../../_static/underscore.js"></script>
+    <script type="text/javascript" src="../../_static/doctools.js"></script>
+    <script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+    <link rel="index" title="Index" href="../../genindex.html" />
+    <link rel="search" title="Search" href="../../search.html" />
+    <link rel="top" title="PaddlePaddle  documentation" href="../../index.html" /> 
+<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>
+  </head>
+  <body role="document">
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../../genindex.html" title="General Index"
+             accesskey="I">index</a></li>
+        <li class="nav-item nav-item-0"><a href="../../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>  
+    <div class="document">
+      <div class="documentwrapper">
+        <div class="bodywrapper">
+          <div class="body" role="main">
+  <div class="section" id="layer">
+<span id="layer"></span><h1>支持双层序列作为输入的Layer<a class="headerlink" href="#layer" title="Permalink to this headline">¶</a></h1>
+<div class="section" id="">
+<span id="id1"></span><h2>概述<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h2>
+<p>在自然语言处理任务中，序列是一种常见的数据类型。一个独立的词语，可以看作是一个非序列输入，或者，我们称之为一个0层的序列；由词语构成的句子，是一个单层序列；若干个句子构成一个段落，是一个双层的序列。</p>
+<p>双层序列是一个嵌套的序列，它的每一个元素，又是一个单层的序列。这是一种非常灵活的数据组织方式，帮助我们构造一些复杂的输入信息。</p>
+<p>我们可以按照如下层次定义非序列，单层序列，以及双层序列。</p>
+<ul class="simple">
+<li>0层序列：一个独立的元素，类型可以是PaddlePaddle支持的任意输入数据类型</li>
+<li>单层序列：排成一列的多个元素，每个元素是一个0层序列，元素之间的顺序是重要的输入信息</li>
+<li>双层序列：排成一列的多个元素，每个元素是一个单层序列，称之为双层序列的一个子序列（subseq），subseq的每个元素是一个0层序列</li>
+</ul>
+<p>在 PaddlePaddle中，下面这些Layer能够接受双层序列作为输入，完成相应的计算。</p>
+</div>
+<div class="section" id="pooling-layer">
+<span id="pooling-layer"></span><h2>pooling_layer<a class="headerlink" href="#pooling-layer" title="Permalink to this headline">¶</a></h2>
+<p>pooling_layer的使用示例如下，详细见<a href = "../../../doc/ui/api/trainer_config_helpers/layers.html#pooling-layer">配置API</a>。</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">seq_pool</span> <span class="o">=</span> <span class="n">pooling_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">layer</span><span class="p">,</span>
+                         <span class="n">pooling_type</span><span class="o">=</span><span class="n">AvgPooling</span><span class="p">(),</span>
+                         <span class="n">agg_level</span><span class="o">=</span><span class="n">AggregateLevel</span><span class="o">.</span><span class="n">EACH_SEQUENCE</span><span class="p">)</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li><code class="docutils literal"><span class="pre">pooling_type</span></code> 目前支持两种，分别是：MaxPooling()和AvgPooling()。</li>
+<li><code class="docutils literal"><span class="pre">agg_level=AggregateLevel.TIMESTEP</span></code>时（默认值）：<ul>
+<li>作用：双层序列经过运算变成一个0层序列，或单层序列经过运算变成一个0层序列</li>
+<li>输入：一个双层序列，或一个单层序列</li>
+<li>输出：一个0层序列，即整个输入序列（单层或双层）的平均值（或最大值）</li>
+</ul>
+</li>
+<li><code class="docutils literal"><span class="pre">agg_level=AggregateLevel.EACH_SEQUENCE</span></code>时：<ul>
+<li>作用：一个双层序列经过运算变成一个单层序列</li>
+<li>输入：必须是一个双层序列</li>
+<li>输出：一个单层序列，序列的每个元素是原来双层序列每个subseq元素的平均值（或最大值）</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="last-seq-first-seq">
+<span id="last-seq-first-seq"></span><h2>last_seq 和 first_seq<a class="headerlink" href="#last-seq-first-seq" title="Permalink to this headline">¶</a></h2>
+<p>last_seq的使用示例如下（first_seq类似），详细见<a href = "../../../doc/ui/api/trainer_config_helpers/layers.html#last-seq">配置API</a>。</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">last</span> <span class="o">=</span> <span class="n">last_seq</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">layer</span><span class="p">,</span>
+                <span class="n">agg_level</span><span class="o">=</span><span class="n">AggregateLevel</span><span class="o">.</span><span class="n">EACH_SEQUENCE</span><span class="p">)</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li><code class="docutils literal"><span class="pre">agg_level=AggregateLevel.TIMESTEP</span></code>时（默认值）：<ul>
+<li>作用：一个双层序列经过运算变成一个0层序列，或一个单层序列经过运算变成一个0层序列</li>
+<li>输入：一个双层序列或一个单层序列</li>
+<li>输出：一个0层序列，即整个输入序列（双层或者单层）最后一个，或第一个元素。</li>
+</ul>
+</li>
+<li><code class="docutils literal"><span class="pre">agg_level=AggregateLevel.EACH_SEQUENCE</span></code>时：<ul>
+<li>作用：一个双层序列经过运算变成一个单层序列</li>
+<li>输入：必须是一个双层序列</li>
+<li>输出：一个单层序列，其中每个元素是双层序列中每个subseq最后一个（或第一个）元素。</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="expand-layer">
+<span id="expand-layer"></span><h2>expand_layer<a class="headerlink" href="#expand-layer" title="Permalink to this headline">¶</a></h2>
+<p>expand_layer的使用示例如下，详细见<a href = "../../../doc/ui/api/trainer_config_helpers/layers.html#expand-layer">配置API</a>。</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">expand</span> <span class="o">=</span> <span class="n">expand_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">layer1</span><span class="p">,</span>
+                      <span class="n">expand_as</span><span class="o">=</span><span class="n">layer2</span><span class="p">,</span>
+                      <span class="n">expand_level</span><span class="o">=</span><span class="n">ExpandLevel</span><span class="o">.</span><span class="n">FROM_TIMESTEP</span><span class="p">)</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li><code class="docutils literal"><span class="pre">expand_level=ExpandLevel.FROM_TIMESTEP</span></code>时（默认值）：<ul>
+<li>作用：一个0层序列经过运算扩展成一个单层序列，或者一个双层序列</li>
+<li>输入：layer1必须是一个0层序列，是待扩展的数据；layer2可以是一个单层序列，或者是一个双层序列，提供扩展的长度信息</li>
+<li>输出：一个单层序列，或一个双层序列，输出序列的类型（双层序列，或单层序列）和序列中含有元素的数目同 layer2一致。若输出是单层序列，单层序列的每个元素（0层序列），都是对layer1元素的拷贝；若输出是双层序列，双层序列每个subseq中每个元素（0层序列），都是对layer1元素的拷贝</li>
+</ul>
+</li>
+<li><code class="docutils literal"><span class="pre">expand_level=ExpandLevel.FROM_SEQUENCE</span></code>时：<ul>
+<li>作用：一个单层序列经过运算扩展成一个双层序列</li>
+<li>输入：layer1必须是一个单层序列，是待扩展的数据；layer2必须是一个双层序列，提供扩展的长度信息</li>
+<li>输出：一个双层序列，序列中含有元素的数目同layer2一致。要求单层序列含有元素的数目（0层序列），和双层序列含有subseq 的数目一致。单层序列第i个元素（0层序列），被扩展为一个单层序列，构成了输出双层序列的第i个subseq。</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+          </div>
+        </div>
+      </div>
+      <div class="sphinxsidebar" role="navigation" aria-label="main navigation">
+        <div class="sphinxsidebarwrapper">
+  <h3><a href="../../index.html">Table Of Contents</a></h3>
+  <ul>
+<li><a class="reference internal" href="#">支持双层序列作为输入的Layer</a><ul>
+<li><a class="reference internal" href="#">概述</a></li>
+<li><a class="reference internal" href="#pooling-layer">pooling_layer</a></li>
+<li><a class="reference internal" href="#last-seq-first-seq">last_seq 和 first_seq</a></li>
+<li><a class="reference internal" href="#expand-layer">expand_layer</a></li>
+</ul>
+</li>
+</ul>
+  <div role="note" aria-label="source link">
+    <h3>This Page</h3>
+    <ul class="this-page-menu">
+      <li><a href="../../_sources/algorithm/rnn/hierarchical-layer.txt"
+            rel="nofollow">Show Source</a></li>
+    </ul>
+   </div>
+<div id="searchbox" style="display: none" role="search">
+  <h3>Quick search</h3>
+    <form class="search" action="../../search.html" method="get">
+      <div><input type="text" name="q" /></div>
+      <div><input type="submit" value="Go" /></div>
+      <input type="hidden" name="check_keywords" value="yes" />
+      <input type="hidden" name="area" value="default" />
+    </form>
+</div>
+<script type="text/javascript">$('#searchbox').show(0);</script>
+        </div>
+      </div>
+      <div class="clearer"></div>
+    </div>
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../../genindex.html" title="General Index"
+             >index</a></li>
+        <li class="nav-item nav-item-0"><a href="../../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>
+    <div class="footer" role="contentinfo">
+        &#169; Copyright 2016, PaddlePaddle developers.
+      Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.4.8.
+    </div>
+  </body>
+</html>
\ No newline at end of file
--- a/doc_cn/algorithm/rnn/hierarchical-rnn.html
+++ b/doc_cn/algorithm/rnn/hierarchical-rnn.html
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+  "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+  <head>
+    <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+    <title>双层RNN配置与示例 &#8212; PaddlePaddle  documentation</title>
+    <link rel="stylesheet" href="../../_static/classic.css" type="text/css" />
+    <link rel="stylesheet" href="../../_static/pygments.css" type="text/css" />
+    <script type="text/javascript">
+      var DOCUMENTATION_OPTIONS = {
+        URL_ROOT:    '../../',
+        VERSION:     '',
+        COLLAPSE_INDEX: false,
+        FILE_SUFFIX: '.html',
+        HAS_SOURCE:  true
+      };
+    </script>
+    <script type="text/javascript" src="../../_static/jquery.js"></script>
+    <script type="text/javascript" src="../../_static/underscore.js"></script>
+    <script type="text/javascript" src="../../_static/doctools.js"></script>
+    <script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+    <link rel="index" title="Index" href="../../genindex.html" />
+    <link rel="search" title="Search" href="../../search.html" />
+    <link rel="top" title="PaddlePaddle  documentation" href="../../index.html" /> 
+<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>
+  </head>
+  <body role="document">
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../../genindex.html" title="General Index"
+             accesskey="I">index</a></li>
+        <li class="nav-item nav-item-0"><a href="../../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>  
+    <div class="document">
+      <div class="documentwrapper">
+        <div class="bodywrapper">
+          <div class="body" role="main">
+  <div class="section" id="rnn">
+<span id="rnn"></span><h1>双层RNN配置与示例<a class="headerlink" href="#rnn" title="Permalink to this headline">¶</a></h1>
+<p>我们在<code class="docutils literal"><span class="pre">paddle/gserver/tests/test_RecurrentGradientMachine</span></code>单测中，通过多组语义相同的单双层RNN配置，讲解如何使用双层RNN。</p>
+<div class="section" id="subseqmemory">
+<span id="subseqmemory"></span><h2>示例1：双进双出，subseq间无memory<a class="headerlink" href="#subseqmemory" title="Permalink to this headline">¶</a></h2>
+<p>配置：单层RNN（<code class="docutils literal"><span class="pre">sequence_layer_group</span></code>）和双层RNN（<code class="docutils literal"><span class="pre">sequence_nest_layer_group</span></code>），语义完全相同。</p>
+<div class="section" id="">
+<span id="id1"></span><h3>读取双层序列的方法<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>首先，我们看一下单双层序列的不同数据组织形式（您也可以采用别的组织形式）：</p>
+<ul class="simple">
+<li>单层序列的数据（<code class="docutils literal"><span class="pre">Sequence/tour_train_wdseg</span></code>）如下，一共有10个样本。每个样本由两部分组成，一个label（此处都为2）和一个已经分词后的句子。</li>
+</ul>
+<div class="highlight-text"><div class="highlight"><pre><span></span>2   酒店 有 很 舒适 的 床垫 子 ， 床上用品 也 应该 是 一人 一 换 ， 感觉 很 利落 对 卫生 很 放心 呀 。
+2   很 温馨 ， 也 挺 干净 的 * 地段 不错 ， 出来 就 有 全家 ， 离 地铁站 也 近 ， 交通 很方便 * 就是 都 不 给 刷牙 的 杯子 啊 ， 就 第一天 给 了 一次性杯子 *
+2   位置 方便 ， 强烈推荐 ， 十一 出去玩 的 时候 选 的 ， 对面 就是 华润万家 ， 周围 吃饭 的 也 不少 。
+2   交通便利 ， 吃 很 便利 ， 乾 浄 、 安静 ， 商务 房 有 电脑 、 上网 快 ， 价格 可以 ， 就 早餐 不 好吃 。 整体 是 不错 的 。 適 合 出差 來 住 。
+2   本来 准备 住 两 晚 ， 第 2 天 一早 居然 停电 ， 且 无 通知 ， 只有 口头 道歉 。 总体来说 性价比 尚可 ， 房间 较 新 ， 还是 推荐 .
+2   这个 酒店 去过 很多 次 了 ， 选择 的 主要原因 是 离 客户 最 便宜 相对 又 近 的 酒店
+2   挺好 的 汉庭 ， 前台 服务 很 热情 ， 卫生 很 整洁 ， 房间 安静 ， 水温 适中 ， 挺好 ！
+2   HowardJohnson 的 品质 ， 服务 相当 好 的 一 家 五星级 。 房间 不错 、 泳池 不错 、 楼层 安排 很 合理 。 还有 就是 地理位置 ， 简直 一 流 。 就 在 天一阁 、 月湖 旁边 ， 离 天一广场 也 不远 。 下次 来 宁波 还会 住 。
+2   酒店 很干净 ， 很安静 ， 很 温馨 ， 服务员 服务 好 ， 各方面 都 不错 *
+2   挺好 的 ， 就是 没 窗户 ， 不过 对 得 起 这 价格
+</pre></div>
+</div>
+<ul class="simple">
+<li>双层序列的数据（<code class="docutils literal"><span class="pre">Sequence/tour_train_wdseg.nest</span></code>）如下，一共有4个样本。样本间用空行分开，代表不同的双层序列，序列数据和上面的完全一样。每个样本的子句数分别为2,3,2,3。</li>
+</ul>
+<div class="highlight-text"><div class="highlight"><pre><span></span>2   酒店 有 很 舒适 的 床垫 子 ， 床上用品 也 应该 是 一人 一 换 ， 感觉 很 利落 对 卫生 很 放心 呀 。
+2   很 温馨 ， 也 挺 干净 的 * 地段 不错 ， 出来 就 有 全家 ， 离 地铁站 也 近 ， 交通 很方便 * 就是 都 不 给 刷牙 的 杯子 啊 ， 就 第一天 给 了 一次性杯子 *
+2   位置 方便 ， 强烈推荐 ， 十一 出去玩 的 时候 选 的 ， 对面 就是 华润万家 ， 周围 吃饭 的 也 不少 。
+2   交通便利 ， 吃 很 便利 ， 乾 浄 、 安静 ， 商务 房 有 电脑 、 上网 快 ， 价格 可以 ， 就 早餐 不 好吃 。 整体 是 不错 的 。 適 合 出差 來 住 。
+2   本来 准备 住 两 晚 ， 第 2 天 一早 居然 停电 ， 且 无 通知 ， 只有 口头 道歉 。 总体来说 性价比 尚可 ， 房间 较 新 ， 还是 推荐 .
+2   这个 酒店 去过 很多 次 了 ， 选择 的 主要原因 是 离 客户 最 便宜 相对 又 近 的 酒店
+2   挺好 的 汉庭 ， 前台 服务 很 热情 ， 卫生 很 整洁 ， 房间 安静 ， 水温 适中 ， 挺好 ！
+2   HowardJohnson 的 品质 ， 服务 相当 好 的 一 家 五星级 。 房间 不错 、 泳池 不错 、 楼层 安排 很 合理 。 还有 就是 地理位置 ， 简直 一 流 。 就 在 天一阁 、 月湖 旁边 ， 离 天一广场 也 不远 。 下次 来 宁波 还会 住 。
+2   酒店 很干净 ， 很安静 ， 很 温馨 ， 服务员 服务 好 ， 各方面 都 不错 *
+2   挺好 的 ， 就是 没 窗户 ， 不过 对 得 起 这 价格
+</pre></div>
+</div>
+<p>其次，我们看一下单双层序列的不同dataprovider（见<code class="docutils literal"><span class="pre">sequenceGen.py</span></code>）：</p>
+<ul class="simple">
+<li>单层序列的dataprovider如下：<ul>
+<li>word_slot是integer_value_sequence类型，代表单层序列。</li>
+<li>label是integer_value类型，代表一个向量。</li>
+</ul>
+</li>
+</ul>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">hook</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">dict_file</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
+    <span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span> <span class="o">=</span> <span class="n">dict_file</span>
+    <span class="n">settings</span><span class="o">.</span><span class="n">input_types</span> <span class="o">=</span> <span class="p">[</span><span class="n">integer_value_sequence</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span><span class="p">)),</span> 
+                            <span class="n">integer_value</span><span class="p">(</span><span class="mi">3</span><span class="p">)]</span>
+<span class="nd">@provider</span><span class="p">(</span><span class="n">init_hook</span><span class="o">=</span><span class="n">hook</span><span class="p">)</span>
+<span class="k">def</span> <span class="nf">process</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">file_name</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;r&#39;</span><span class="p">)</span> <span class="k">as</span> <span class="n">fdata</span><span class="p">:</span>
+        <span class="k">for</span> <span class="n">line</span> <span class="ow">in</span> <span class="n">fdata</span><span class="p">:</span>
+            <span class="n">label</span><span class="p">,</span> <span class="n">comment</span> <span class="o">=</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">split</span><span class="p">(</span><span class="s1">&#39;</span><span class="se">\t</span><span class="s1">&#39;</span><span class="p">)</span>
+            <span class="n">label</span> <span class="o">=</span> <span class="nb">int</span><span class="p">(</span><span class="s1">&#39;&#39;</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">label</span><span class="o">.</span><span class="n">split</span><span class="p">()))</span>
+            <span class="n">words</span> <span class="o">=</span> <span class="n">comment</span><span class="o">.</span><span class="n">split</span><span class="p">()</span>
+            <span class="n">word_slot</span> <span class="o">=</span> <span class="p">[</span><span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span><span class="p">[</span><span class="n">w</span><span class="p">]</span> <span class="k">for</span> <span class="n">w</span> <span class="ow">in</span> <span class="n">words</span> <span class="k">if</span> <span class="n">w</span> <span class="ow">in</span> <span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span><span class="p">]</span>
+            <span class="k">yield</span> <span class="n">word_slot</span><span class="p">,</span> <span class="n">label</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li>双层序列的dataprovider如下：<ul>
+<li>word_slot是integer_value_sub_sequence类型，代表双层序列。</li>
+<li>label是integer_value_sequence类型，代表单层序列，即一个子句一个label。注意：也可以为integer_value类型，代表一个向量，即一个句子一个label。通常根据任务需求进行不同设置。</li>
+<li>关于dataprovider中input_types的详细用法，参见PyDataProvider2。</li>
+</ul>
+</li>
+</ul>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">hook2</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">dict_file</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
+    <span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span> <span class="o">=</span> <span class="n">dict_file</span>
+    <span class="n">settings</span><span class="o">.</span><span class="n">input_types</span> <span class="o">=</span> <span class="p">[</span><span class="n">integer_value_sub_sequence</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span><span class="p">)),</span>
+                            <span class="n">integer_value_sequence</span><span class="p">(</span><span class="mi">3</span><span class="p">)]</span>
+<span class="nd">@provider</span><span class="p">(</span><span class="n">init_hook</span><span class="o">=</span><span class="n">hook2</span><span class="p">)</span>
+<span class="k">def</span> <span class="nf">process2</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">file_name</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="k">as</span> <span class="n">fdata</span><span class="p">:</span>
+        <span class="n">label_list</span> <span class="o">=</span> <span class="p">[]</span>
+        <span class="n">word_slot_list</span> <span class="o">=</span> <span class="p">[]</span>
+        <span class="k">for</span> <span class="n">line</span> <span class="ow">in</span> <span class="n">fdata</span><span class="p">:</span>
+            <span class="k">if</span> <span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">line</span><span class="p">))</span> <span class="o">&gt;</span> <span class="mi">1</span><span class="p">:</span>
+                <span class="n">label</span><span class="p">,</span><span class="n">comment</span> <span class="o">=</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">split</span><span class="p">(</span><span class="s1">&#39;</span><span class="se">\t</span><span class="s1">&#39;</span><span class="p">)</span>
+                <span class="n">label</span> <span class="o">=</span> <span class="nb">int</span><span class="p">(</span><span class="s1">&#39;&#39;</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">label</span><span class="o">.</span><span class="n">split</span><span class="p">()))</span>
+                <span class="n">words</span> <span class="o">=</span> <span class="n">comment</span><span class="o">.</span><span class="n">split</span><span class="p">()</span>
+                <span class="n">word_slot</span> <span class="o">=</span> <span class="p">[</span><span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span><span class="p">[</span><span class="n">w</span><span class="p">]</span> <span class="k">for</span> <span class="n">w</span> <span class="ow">in</span> <span class="n">words</span> <span class="k">if</span> <span class="n">w</span> <span class="ow">in</span> <span class="n">settings</span><span class="o">.</span><span class="n">word_dict</span><span class="p">]</span>
+                <span class="n">label_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">label</span><span class="p">)</span>
+                <span class="n">word_slot_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">word_slot</span><span class="p">)</span>
+            <span class="k">else</span><span class="p">:</span>
+                <span class="k">yield</span> <span class="n">word_slot_list</span><span class="p">,</span> <span class="n">label_list</span>
+                <span class="n">label_list</span> <span class="o">=</span> <span class="p">[]</span>
+                <span class="n">word_slot_list</span> <span class="o">=</span> <span class="p">[]</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="">
+<span id="id2"></span><h3>模型中的配置<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>首先，我们看一下单层序列的配置（见<code class="docutils literal"><span class="pre">sequence_layer_group.conf</span></code>）。注意：batchsize=5表示一次过5句单层序列，因此2个batch就可以完成1个pass。</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">settings</span><span class="p">(</span><span class="n">batch_size</span><span class="o">=</span><span class="mi">5</span><span class="p">)</span>
+<span class="n">data</span> <span class="o">=</span> <span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;word&quot;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">dict_dim</span><span class="p">)</span>
+<span class="n">emb</span> <span class="o">=</span> <span class="n">embedding_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">data</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="c1"># (lstm_input + lstm) is equal to lstmemory </span>
+<span class="k">with</span> <span class="n">mixed_layer</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="o">*</span><span class="mi">4</span><span class="p">)</span> <span class="k">as</span> <span class="n">lstm_input</span><span class="p">:</span>
+    <span class="n">lstm_input</span> <span class="o">+=</span> <span class="n">full_matrix_projection</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">emb</span><span class="p">)</span>
+<span class="n">lstm</span> <span class="o">=</span> <span class="n">lstmemory_group</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_input</span><span class="p">,</span>
+                       <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">,</span>
+                       <span class="n">act</span><span class="o">=</span><span class="n">TanhActivation</span><span class="p">(),</span>
+                       <span class="n">gate_act</span><span class="o">=</span><span class="n">SigmoidActivation</span><span class="p">(),</span>
+                       <span class="n">state_act</span><span class="o">=</span><span class="n">TanhActivation</span><span class="p">(),</span>
+                       <span class="n">lstm_layer_attr</span><span class="o">=</span><span class="n">ExtraLayerAttribute</span><span class="p">(</span><span class="n">error_clipping_threshold</span><span class="o">=</span><span class="mi">50</span><span class="p">))</span>
+<span class="n">lstm_last</span> <span class="o">=</span> <span class="n">last_seq</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm</span><span class="p">)</span>
+<span class="k">with</span> <span class="n">mixed_layer</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">label_dim</span><span class="p">,</span> 
+                 <span class="n">act</span><span class="o">=</span><span class="n">SoftmaxActivation</span><span class="p">(),</span> 
+                 <span class="n">bias_attr</span><span class="o">=</span><span class="bp">True</span><span class="p">)</span> <span class="k">as</span> <span class="n">output</span><span class="p">:</span>
+    <span class="n">output</span> <span class="o">+=</span> <span class="n">full_matrix_projection</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_last</span><span class="p">)</span>
+<span class="n">outputs</span><span class="p">(</span><span class="n">classification_cost</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">output</span><span class="p">,</span> <span class="n">label</span><span class="o">=</span><span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;label&quot;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">1</span><span class="p">)))</span>
+</pre></div>
+</div>
+<p>其次，我们看一下语义相同的双层序列配置（见<code class="docutils literal"><span class="pre">sequence_nest_layer_group.conf</span></code>），并对其详细分析：</p>
+<ul class="simple">
+<li>batchsize=2表示一次过2句双层序列。但从上面的数据格式可知，2句双层序列和5句单层序列的数据完全一样。</li>
+<li>data_layer和embedding_layer不关心数据是否是序列格式，因此两个配置在这两层上的输出是一样的。</li>
+<li>lstmemory:<ul>
+<li>单层序列过了一个mixed_layer和lstmemory_group。</li>
+<li>双层序列在同样的mixed_layer和lstmemory_group外，直接加了一层group。由于这个外层group里面没有memory，表示subseq间不存在联系，即起到的作用仅仅是把双层seq拆成单层，因此双层序列过完lstmemory的输出和单层的一样。</li>
+</ul>
+</li>
+<li>last_seq：<ul>
+<li>单层序列直接取了最后一个元素</li>
+<li>双层序列首先（last_seq层）取了每个subseq的最后一个元素，将其拼接成一个新的单层序列；接着（expand_layer层）将其扩展成一个新的双层序列，其中第i个subseq中的所有向量均为输入的单层序列中的第i个向量；最后（average_layer层）取了每个subseq的平均值。</li>
+<li>分析得出：第一个last_seq后，每个subseq的最后一个元素就等于单层序列的最后一个元素，而expand_layer和average_layer后，依然保持每个subseq最后一个元素的值不变（这两层仅是为了展示它们的用法，实际中并不需要）。因此单双层序列的输出是一样旳。</li>
+</ul>
+</li>
+</ul>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">settings</span><span class="p">(</span><span class="n">batch_size</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
+<span class="n">data</span> <span class="o">=</span> <span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;word&quot;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">dict_dim</span><span class="p">)</span>
+<span class="n">emb_group</span> <span class="o">=</span> <span class="n">embedding_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">data</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="c1"># (lstm_input + lstm) is equal to lstmemory </span>
+<span class="k">def</span> <span class="nf">lstm_group</span><span class="p">(</span><span class="n">lstm_group_input</span><span class="p">):</span>
+    <span class="k">with</span> <span class="n">mixed_layer</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="o">*</span><span class="mi">4</span><span class="p">)</span> <span class="k">as</span> <span class="n">group_input</span><span class="p">:</span>
+      <span class="n">group_input</span> <span class="o">+=</span> <span class="n">full_matrix_projection</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_group_input</span><span class="p">)</span>
+    <span class="n">lstm_output</span> <span class="o">=</span> <span class="n">lstmemory_group</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">group_input</span><span class="p">,</span>
+                                  <span class="n">name</span><span class="o">=</span><span class="s2">&quot;lstm_group&quot;</span><span class="p">,</span>
+                                  <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">,</span>
+                                  <span class="n">act</span><span class="o">=</span><span class="n">TanhActivation</span><span class="p">(),</span>
+                                  <span class="n">gate_act</span><span class="o">=</span><span class="n">SigmoidActivation</span><span class="p">(),</span>
+                                  <span class="n">state_act</span><span class="o">=</span><span class="n">TanhActivation</span><span class="p">(),</span>
+                                  <span class="n">lstm_layer_attr</span><span class="o">=</span><span class="n">ExtraLayerAttribute</span><span class="p">(</span><span class="n">error_clipping_threshold</span><span class="o">=</span><span class="mi">50</span><span class="p">))</span>
+    <span class="k">return</span> <span class="n">lstm_output</span>
+<span class="n">lstm_nest_group</span> <span class="o">=</span> <span class="n">recurrent_group</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">SubsequenceInput</span><span class="p">(</span><span class="n">emb_group</span><span class="p">),</span>
+                                  <span class="n">step</span><span class="o">=</span><span class="n">lstm_group</span><span class="p">,</span>
+                                  <span class="n">name</span><span class="o">=</span><span class="s2">&quot;lstm_nest_group&quot;</span><span class="p">)</span>
+<span class="c1"># hasSubseq -&gt;(seqlastins) seq</span>
+<span class="n">lstm_last</span> <span class="o">=</span> <span class="n">last_seq</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_nest_group</span><span class="p">,</span> <span class="n">agg_level</span><span class="o">=</span><span class="n">AggregateLevel</span><span class="o">.</span><span class="n">EACH_SEQUENCE</span><span class="p">)</span>
+<span class="c1"># seq -&gt;(expand) hasSubseq</span>
+<span class="n">lstm_expand</span> <span class="o">=</span> <span class="n">expand_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_last</span><span class="p">,</span> <span class="n">expand_as</span><span class="o">=</span><span class="n">emb_group</span><span class="p">,</span> <span class="n">expand_level</span><span class="o">=</span><span class="n">ExpandLevel</span><span class="o">.</span><span class="n">FROM_SEQUENCE</span><span class="p">)</span>
+<span class="c1"># hasSubseq -&gt;(average) seq</span>
+<span class="n">lstm_average</span> <span class="o">=</span> <span class="n">pooling_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_expand</span><span class="p">,</span>
+                             <span class="n">pooling_type</span><span class="o">=</span><span class="n">AvgPooling</span><span class="p">(),</span>
+                             <span class="n">agg_level</span><span class="o">=</span><span class="n">AggregateLevel</span><span class="o">.</span><span class="n">EACH_SEQUENCE</span><span class="p">)</span>
+<span class="k">with</span> <span class="n">mixed_layer</span><span class="p">(</span><span class="n">size</span><span class="o">=</span><span class="n">label_dim</span><span class="p">,</span> 
+                 <span class="n">act</span><span class="o">=</span><span class="n">SoftmaxActivation</span><span class="p">(),</span> 
+                 <span class="n">bias_attr</span><span class="o">=</span><span class="bp">True</span><span class="p">)</span> <span class="k">as</span> <span class="n">output</span><span class="p">:</span>
+    <span class="n">output</span> <span class="o">+=</span> <span class="n">full_matrix_projection</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">lstm_average</span><span class="p">)</span>
+<span class="n">outputs</span><span class="p">(</span><span class="n">classification_cost</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">output</span><span class="p">,</span> <span class="n">label</span><span class="o">=</span><span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;label&quot;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">1</span><span class="p">)))</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="subseqmemory">
+<span id="id3"></span><h2>示例2：双进双出，subseq间有memory<a class="headerlink" href="#subseqmemory" title="Permalink to this headline">¶</a></h2>
+<p>配置：单层RNN（<code class="docutils literal"><span class="pre">sequence_rnn.conf</span></code>），双层RNN（<code class="docutils literal"><span class="pre">sequence_nest_rnn.conf</span></code>和<code class="docutils literal"><span class="pre">sequence_nest_rnn_readonly_memory.conf</span></code>），语义完全相同。</p>
+<div class="section" id="">
+<span id="id4"></span><h3>读取双层序列的方法<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>我们看一下单双层序列的不同数据组织形式和dataprovider（见<code class="docutils literal"><span class="pre">rnn_data_provider.py</span></code>）</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">data</span> <span class="o">=</span> <span class="p">[</span>
+    <span class="p">[[[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span> <span class="p">[</span><span class="mi">4</span><span class="p">,</span> <span class="mi">5</span><span class="p">,</span> <span class="mi">2</span><span class="p">]],</span> <span class="mi">0</span><span class="p">],</span>
+    <span class="p">[[[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span> <span class="p">[</span><span class="mi">2</span><span class="p">,</span> <span class="mi">5</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">]],</span> <span class="mi">1</span><span class="p">],</span>
+<span class="p">]</span>
+<span class="nd">@provider</span><span class="p">(</span><span class="n">input_types</span><span class="o">=</span><span class="p">[</span><span class="n">integer_value_sub_sequence</span><span class="p">(</span><span class="mi">10</span><span class="p">),</span>
+                       <span class="n">integer_value</span><span class="p">(</span><span class="mi">3</span><span class="p">)])</span>
+<span class="k">def</span> <span class="nf">process_subseq</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">file_name</span><span class="p">):</span>
+    <span class="k">for</span> <span class="n">d</span> <span class="ow">in</span> <span class="n">data</span><span class="p">:</span>
+        <span class="k">yield</span> <span class="n">d</span>
+<span class="nd">@provider</span><span class="p">(</span><span class="n">input_types</span><span class="o">=</span><span class="p">[</span><span class="n">integer_value_sequence</span><span class="p">(</span><span class="mi">10</span><span class="p">),</span>
+                       <span class="n">integer_value</span><span class="p">(</span><span class="mi">3</span><span class="p">)])</span>
+<span class="k">def</span> <span class="nf">process_seq</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">file_name</span><span class="p">):</span>
+    <span class="k">for</span> <span class="n">d</span> <span class="ow">in</span> <span class="n">data</span><span class="p">:</span>
+        <span class="n">seq</span> <span class="o">=</span> <span class="p">[]</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li>单层序列：有两句，分别为[1,3,2,4,5,2]和[0,2,2,5,0,1,2]。</li>
+<li>双层序列：有两句，分别为[[1,3,2],[4,5,2]]（2个子句）和[[0,2],[2,5],[0,1,2]]（3个子句）。</li>
+<li>单双层序列的label都分别是0和1</li>
+</ul>
+</div>
+<div class="section" id="">
+<span id="id5"></span><h3>模型中的配置<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>我们选取单双层序列配置中的不同部分，来对比分析两者语义相同的原因。</p>
+<ul class="simple">
+<li>单层序列：过了一个很简单的recurrent_group。每一个时间步，当前的输入y和上一个时间步的输出rnn_state做了一个全链接。</li>
+</ul>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">step</span><span class="p">(</span><span class="n">y</span><span class="p">):</span>
+    <span class="n">mem</span> <span class="o">=</span> <span class="n">memory</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;rnn_state&quot;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">)</span>
+    <span class="k">return</span> <span class="n">fc_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="p">[</span><span class="n">y</span><span class="p">,</span> <span class="n">mem</span><span class="p">],</span>
+                    <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">,</span>
+                    <span class="n">act</span><span class="o">=</span><span class="n">TanhActivation</span><span class="p">(),</span>
+                    <span class="n">bias_attr</span><span class="o">=</span><span class="bp">True</span><span class="p">,</span>
+                    <span class="n">name</span><span class="o">=</span><span class="s2">&quot;rnn_state&quot;</span><span class="p">)</span>
+<span class="n">out</span> <span class="o">=</span> <span class="n">recurrent_group</span><span class="p">(</span><span class="n">step</span><span class="o">=</span><span class="n">step</span><span class="p">,</span> <span class="nb">input</span><span class="o">=</span><span class="n">emb</span><span class="p">)</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li>双层序列，外层memory是一个元素：<ul>
+<li>内层inner_step的recurrent_group和单层序列的几乎一样。除了boot_layer=outer_mem，表示将外层的outer_mem作为内层memory的初始状态。外层outer_step中，outer_mem是一个子句的最后一个向量，即整个双层group是将前一个子句的最后一个向量，作为下一个子句memory的初始状态。</li>
+<li>从输入数据上看，单双层序列的句子是一样的，只是双层序列将其又做了子序列划分。因此双层序列的配置中，必须将前一个子句的最后一个元素，作为boot_layer传给下一个子句的memory，才能保证和单层序列的配置中“每一个时间步都用了上一个时间步的输出结果”一致。</li>
+</ul>
+</li>
+</ul>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">outer_step</span><span class="p">(</span><span class="n">x</span><span class="p">):</span>
+    <span class="n">outer_mem</span> <span class="o">=</span> <span class="n">memory</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;outer_rnn_state&quot;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">)</span>
+    <span class="k">def</span> <span class="nf">inner_step</span><span class="p">(</span><span class="n">y</span><span class="p">):</span>
+        <span class="n">inner_mem</span> <span class="o">=</span> <span class="n">memory</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s2">&quot;inner_rnn_state&quot;</span><span class="p">,</span>
+                           <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">,</span>
+                           <span class="n">boot_layer</span><span class="o">=</span><span class="n">outer_mem</span><span class="p">)</span>
+        <span class="k">return</span> <span class="n">fc_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="p">[</span><span class="n">y</span><span class="p">,</span> <span class="n">inner_mem</span><span class="p">],</span>
+                        <span class="n">size</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">,</span>
+                        <span class="n">act</span><span class="o">=</span><span class="n">TanhActivation</span><span class="p">(),</span>
+                        <span class="n">bias_attr</span><span class="o">=</span><span class="bp">True</span><span class="p">,</span>
+                        <span class="n">name</span><span class="o">=</span><span class="s2">&quot;inner_rnn_state&quot;</span><span class="p">)</span>
+    <span class="n">inner_rnn_output</span> <span class="o">=</span> <span class="n">recurrent_group</span><span class="p">(</span>
+        <span class="n">step</span><span class="o">=</span><span class="n">inner_step</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">last</span> <span class="o">=</span> <span class="n">last_seq</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">inner_rnn_output</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s2">&quot;outer_rnn_state&quot;</span><span class="p">)</span>
+    <span class="k">return</span> <span class="n">inner_rnn_output</span>
+<span class="n">out</span> <span class="o">=</span> <span class="n">recurrent_group</span><span class="p">(</span><span class="n">step</span><span class="o">=</span><span class="n">outer_step</span><span class="p">,</span> <span class="nb">input</span><span class="o">=</span><span class="n">SubsequenceInput</span><span class="p">(</span><span class="n">emb</span><span class="p">))</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li>双层序列，外层memory是单层序列：<ul>
+<li>由于外层每个时间步返回的是一个子句，这些子句的长度往往不等长。因此当外层有is_seq=True的memory时，内层是<strong>无法直接使用</strong>它的，即内层memory的boot_layer不能链接外层的这个memory。</li>
+<li>如果内层memory想<strong>间接使用</strong>这个外层memory，只能通过<code class="docutils literal"><span class="pre">pooling_layer</span></code>、<code class="docutils literal"><span class="pre">last_seq</span></code>或<code class="docutils literal"><span class="pre">first_seq</span></code>这三个layer将它先变成一个元素。但这种情况下，外层memory必须有boot_layer，否则在第0个时间步时，由于外层memory没有任何seq信息，因此上述三个layer的前向会报出“<strong>Check failed: input.sequenceStartPositions</strong>”的错误。</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+<div class="section" id="">
+<span id="id6"></span><h2>示例3：双进双出，输入不等长<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h2>
+<p>TBD</p>
+</div>
+<div class="section" id="beam-search">
+<span id="beam-search"></span><h2>示例4：beam_search的生成<a class="headerlink" href="#beam-search" title="Permalink to this headline">¶</a></h2>
+<p>TBD</p>
+</div>
+</div>
+          </div>
+        </div>
+      </div>
+      <div class="sphinxsidebar" role="navigation" aria-label="main navigation">
+        <div class="sphinxsidebarwrapper">
+  <h3><a href="../../index.html">Table Of Contents</a></h3>
+  <ul>
+<li><a class="reference internal" href="#">双层RNN配置与示例</a><ul>
+<li><a class="reference internal" href="#subseqmemory">示例1：双进双出，subseq间无memory</a><ul>
+<li><a class="reference internal" href="#">读取双层序列的方法</a></li>
+<li><a class="reference internal" href="#">模型中的配置</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#subseqmemory">示例2：双进双出，subseq间有memory</a><ul>
+<li><a class="reference internal" href="#">读取双层序列的方法</a></li>
+<li><a class="reference internal" href="#">模型中的配置</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#">示例3：双进双出，输入不等长</a></li>
+<li><a class="reference internal" href="#beam-search">示例4：beam_search的生成</a></li>
+</ul>
+</li>
+</ul>
+  <div role="note" aria-label="source link">
+    <h3>This Page</h3>
+    <ul class="this-page-menu">
+      <li><a href="../../_sources/algorithm/rnn/hierarchical-rnn.txt"
+            rel="nofollow">Show Source</a></li>
+    </ul>
+   </div>
+<div id="searchbox" style="display: none" role="search">
+  <h3>Quick search</h3>
+    <form class="search" action="../../search.html" method="get">
+      <div><input type="text" name="q" /></div>
+      <div><input type="submit" value="Go" /></div>
+      <input type="hidden" name="check_keywords" value="yes" />
+      <input type="hidden" name="area" value="default" />
+    </form>
+</div>
+<script type="text/javascript">$('#searchbox').show(0);</script>
+        </div>
+      </div>
+      <div class="clearer"></div>
+    </div>
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../../genindex.html" title="General Index"
+             >index</a></li>
+        <li class="nav-item nav-item-0"><a href="../../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>
+    <div class="footer" role="contentinfo">
+        &#169; Copyright 2016, PaddlePaddle developers.
+      Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.4.8.
+    </div>
+  </body>
+</html>
\ No newline at end of file
--- a/doc_cn/algorithm/rnn/rnn-tutorial.html
+++ b/doc_cn/algorithm/rnn/rnn-tutorial.html
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+  "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+  <head>
+    <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+    <title>Recurrent Group教程 &#8212; PaddlePaddle  documentation</title>
+    <link rel="stylesheet" href="../../_static/classic.css" type="text/css" />
+    <link rel="stylesheet" href="../../_static/pygments.css" type="text/css" />
+    <script type="text/javascript">
+      var DOCUMENTATION_OPTIONS = {
+        URL_ROOT:    '../../',
+        VERSION:     '',
+        COLLAPSE_INDEX: false,
+        FILE_SUFFIX: '.html',
+        HAS_SOURCE:  true
+      };
+    </script>
+    <script type="text/javascript" src="../../_static/jquery.js"></script>
+    <script type="text/javascript" src="../../_static/underscore.js"></script>
+    <script type="text/javascript" src="../../_static/doctools.js"></script>
+    <script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+    <link rel="index" title="Index" href="../../genindex.html" />
+    <link rel="search" title="Search" href="../../search.html" />
+    <link rel="top" title="PaddlePaddle  documentation" href="../../index.html" /> 
+<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>
+  </head>
+  <body role="document">
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../../genindex.html" title="General Index"
+             accesskey="I">index</a></li>
+        <li class="nav-item nav-item-0"><a href="../../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>  
+    <div class="document">
+      <div class="documentwrapper">
+        <div class="bodywrapper">
+          <div class="body" role="main">
+  <div class="section" id="recurrent-group">
+<span id="recurrent-group"></span><h1>Recurrent Group教程<a class="headerlink" href="#recurrent-group" title="Permalink to this headline">¶</a></h1>
+<div class="section" id="">
+<span id="id1"></span><h2>概述<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h2>
+<p>序列数据是自然语言处理任务面对的一种主要输入数据类型。</p>
+<p>一句话是由词语构成的序列，多句话进一步构成了段落。因此，段落可以看作是一个嵌套的双层的序列，这个序列的每个元素又是一个序列。</p>
+<p>双层序列是PaddlePaddle支持的一种非常灵活的数据组织方式，帮助我们更好地描述段落、多轮对话等更为复杂的语言数据。基于双层序列输入，我们可以设计搭建一个灵活的、层次化的RNN，分别从词语和句子级别编码输入数据，同时也能够引入更加复杂的记忆机制，更好地完成一些复杂的语言理解任务。</p>
+<p>在PaddlePaddle中，<code class="docutils literal"><span class="pre">recurrent_group</span></code>是一种任意复杂的RNN单元，用户只需定义RNN在一个时间步内完成的计算，PaddlePaddle负责完成信息和误差在时间序列上的传播。</p>
+<p>更进一步，<code class="docutils literal"><span class="pre">recurrent_group</span></code>同样可以扩展到双层序列的处理上。通过两个嵌套的<code class="docutils literal"><span class="pre">recurrent_group</span></code>分别定义子句级别和词语级别上需要完成的运算，最终实现一个层次化的复杂RNN。</p>
+<p>目前，在PaddlePaddle中，能够对双向序列进行处理的有<code class="docutils literal"><span class="pre">recurrent_group</span></code>和部分Layer，具体可参考文档：<a href = "hierarchical-layer.html">支持双层序列作为输入的Layer</a>。</p>
+</div>
+<div class="section" id="">
+<span id="id2"></span><h2>相关概念<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="">
+<span id="id3"></span><h3>基本原理<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p><code class="docutils literal"><span class="pre">recurrent_group</span></code> 是PaddlePaddle支持的一种任意复杂的RNN单元。使用者只需要关注于设计RNN在一个时间步之内完成的计算，PaddlePaddle负责完成信息和梯度在时间序列上的传播。</p>
+<p>PaddlePaddle中，<code class="docutils literal"><span class="pre">recurrent_group</span></code>的一个简单调用如下：</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">recurrent_group</span><span class="p">(</span><span class="n">step</span><span class="p">,</span> <span class="nb">input</span><span class="p">,</span> <span class="n">reverse</span><span class="p">)</span>
+</pre></div>
+</div>
+<ul class="simple">
+<li>step：一个可调用的函数，定义一个时间步之内RNN单元完成的计算</li>
+<li>input：输入，必须是一个单层序列，或者一个双层序列</li>
+<li>reverse：是否以逆序处理输入序列</li>
+</ul>
+<p>使用<code class="docutils literal"><span class="pre">recurrent_group</span></code>的核心是设计step函数的计算逻辑。step函数内部可以自由组合PaddlePaddle支持的各种layer，完成任意的运算逻辑。<code class="docutils literal"><span class="pre">recurrent_group</span></code> 的输入（即input）会成为step函数的输入，由于step 函数只关注于RNN一个时间步之内的计算，在这里<code class="docutils literal"><span class="pre">recurrent_group</span></code>替我们完成了原始输入数据的拆分。</p>
+</div>
+<div class="section" id="">
+<span id="id4"></span><h3>输入<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p><code class="docutils literal"><span class="pre">recurrent_group</span></code>处理的输入序列主要分为以下三种类型：</p>
+<ul class="simple">
+<li><strong>数据输入</strong>：一个双层序列进入<code class="docutils literal"><span class="pre">recurrent_group</span></code>会被拆解为一个单层序列，一个单层序列进入<code class="docutils literal"><span class="pre">recurrent_group</span></code>会被拆解为非序列，然后交给step函数，这一过程对用户是完全透明的。可以有以下两种：1）通过data_layer拿到的用户输入；2）其它layer的输出。</li>
+<li><strong>只读Memory输入</strong>：<code class="docutils literal"><span class="pre">StaticInput</span></code> 定义了一个只读的Memory，由<code class="docutils literal"><span class="pre">StaticInput</span></code>指定的输入不会被<code class="docutils literal"><span class="pre">recurrent_group</span></code>拆解，<code class="docutils literal"><span class="pre">recurrent_group</span></code> 循环展开的每个时间步总是能够引用所有输入，可以是一个非序列，或者一个单层序列。</li>
+<li><strong>序列生成任务的输入</strong>：<code class="docutils literal"><span class="pre">GeneratedInput</span></code>只用于在序列生成任务中指定输入数据。</li>
+</ul>
+</div>
+<div class="section" id="">
+<span id="id5"></span><h3>输入示例<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>序列生成任务大多遵循encoder-decoer架构，encoder和decoder可以是能够处理序列的任意神经网络单元，而RNN是最流行的选择。</p>
+<p>给定encoder输出和当前词，decoder每次预测产生下一个最可能的词语。在这种结构中，decoder接受两个输入：</p>
+<ul class="simple">
+<li>要生成的目标序列：是decoder的数据输入，也是decoder循环展开的依据，<code class="docutils literal"><span class="pre">recurrent_group</span></code>会对这类输入进行拆解。</li>
+<li>encoder输出，可以是一个非序列，或者一个单层序列：是一个unbounded memory，decoder循环展开的每一个时间步会引用全部结果，不应该被拆解，这种类型的输入必须通过<code class="docutils literal"><span class="pre">StaticInput</span></code>指定。关于Unbounded Memory的更多讨论请参考论文 <a class="reference external" href="https://arxiv.org/abs/1410.5401">Neural Turning Machine</a>。</li>
+</ul>
+<p>在序列生成任务中，decoder RNN总是引用上一时刻预测出的词的词向量，作为当前时刻输入。<code class="docutils literal"><span class="pre">GeneratedInput</span></code>自动完成这一过程。</p>
+</div>
+<div class="section" id="">
+<span id="id6"></span><h3>输出<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p><code class="docutils literal"><span class="pre">step</span></code>函数必须返回一个或多个Layer的输出，这个Layer的输出会作为整个<code class="docutils literal"><span class="pre">recurrent_group</span></code> 最终的输出结果。在输出的过程中，<code class="docutils literal"><span class="pre">recurrent_group</span></code> 会将每个时间步的输出拼接，这个过程对用户也是透明的。</p>
+</div>
+<div class="section" id="memory">
+<span id="memory"></span><h3>memory<a class="headerlink" href="#memory" title="Permalink to this headline">¶</a></h3>
+<p>memory只能在<code class="docutils literal"><span class="pre">recurrent_group</span></code>中定义和使用。memory不能独立存在，必须指向一个PaddlePaddle定义的Layer。引用memory得到这layer上一时刻输出，因此，可以将memory理解为一个时延操作。</p>
+<p>可以显示地指定一个layer的输出用于初始化memory。不指定时，memory默认初始化为0。</p>
+</div>
+</div>
+<div class="section" id="rnn">
+<span id="rnn"></span><h2>双层RNN介绍<a class="headerlink" href="#rnn" title="Permalink to this headline">¶</a></h2>
+<p><code class="docutils literal"><span class="pre">recurrent_group</span></code>帮助我们完成对输入序列的拆分，对输出的合并，以及计算逻辑在序列上的循环展开。</p>
+<p>利用这种特性，两个嵌套的<code class="docutils literal"><span class="pre">recurrent_group</span></code>能够处理双层序列，实现词语和句子两个级别的双层RNN结构。</p>
+<ul class="simple">
+<li>单层（word-level）RNN：每个状态（state）对应一个词（word）。</li>
+<li>双层（sequence-level）RNN：一个双层RNN由多个单层RNN组成，每个单层RNN（即双层RNN的每个状态）对应一个子句（subseq）。</li>
+</ul>
+<p>为了描述方便，下文以NLP任务为例，将含有子句（subseq）的段落定义为一个双层序列，将含有词语的句子定义为一个单层序列，那么0层序列即为一个词语。</p>
+</div>
+<div class="section" id="rnn">
+<span id="id7"></span><h2>双层RNN的使用<a class="headerlink" href="#rnn" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="">
+<span id="id8"></span><h3>训练流程的使用方法<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>使用 <code class="docutils literal"><span class="pre">recurrent_group</span></code>需要遵循以下约定：</p>
+<ul class="simple">
+<li><strong>单进单出</strong>：输入和输出都是单层序列。<ul>
+<li>如果有多个输入，不同输入序列含有的词语数必须严格相等。</li>
+<li>输出一个单层序列，输出序列的词语数和输入序列一致。</li>
+<li>memory：在step函数中定义 memory指向一个layer，通过引用memory得到这个layer上一个时刻输出，形成recurrent 连接。memory的is_seq参数必须为false。如果没有定义memory，每个时间步之内的运算是独立的。</li>
+<li>boot_layer：memory的初始状态，默认初始状为0，memory的is_seq参数必须为false。</li>
+</ul>
+</li>
+<li><strong>双进双出</strong>：输入和输出都是双层序列。<ul>
+<li>如果有多个输入序列，不同输入含有的子句（subseq）数必须严格相等，但子句含有的词语数可以不相等。</li>
+<li>输出一个双层序列，子句（subseq）数、子句的单词数和指定的一个输入序列一致，默认为第一个输入。</li>
+<li>memory：在step函数中定义memory，指向一个layer，通过引用memory得到这个layer上一个时刻的输出，形成recurrent连接。定义在外层<code class="docutils literal"><span class="pre">recurrent_group</span></code> step函数中的memory，能够记录上一个subseq 的状态，可以是一个单层序列（只作为read-only memory），也可以是一个词语。如果没有定义memory，那么 subseq 之间的运算是独立的。</li>
+<li>boot_layer：memory 初始状态，可以是一个单层序列（只作为read-only memory）或一个向量。默认不设置，即初始状态为0。</li>
+</ul>
+</li>
+<li><strong>双进单出</strong>：目前还未支持，会报错&#8221;In hierachical RNN, all out links should be from sequences now&#8221;。</li>
+</ul>
+</div>
+<div class="section" id="">
+<span id="id9"></span><h3>生成流程的使用方法<a class="headerlink" href="#" title="Permalink to this headline">¶</a></h3>
+<p>使用<code class="docutils literal"><span class="pre">beam_search</span></code>需要遵循以下约定：</p>
+<ul class="simple">
+<li>单层RNN：从一个word生成下一个word。</li>
+<li>双层RNN：即把单层RNN生成后的subseq给拼接成一个新的双层seq。从语义上看，也不存在一个subseq直接生成下一个subseq的情况。</li>
+</ul>
+</div>
+</div>
+</div>
+          </div>
+        </div>
+      </div>
+      <div class="sphinxsidebar" role="navigation" aria-label="main navigation">
+        <div class="sphinxsidebarwrapper">
+  <h3><a href="../../index.html">Table Of Contents</a></h3>
+  <ul>
+<li><a class="reference internal" href="#">Recurrent Group教程</a><ul>
+<li><a class="reference internal" href="#">概述</a></li>
+<li><a class="reference internal" href="#">相关概念</a><ul>
+<li><a class="reference internal" href="#">基本原理</a></li>
+<li><a class="reference internal" href="#">输入</a></li>
+<li><a class="reference internal" href="#">输入示例</a></li>
+<li><a class="reference internal" href="#">输出</a></li>
+<li><a class="reference internal" href="#memory">memory</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#rnn">双层RNN介绍</a></li>
+<li><a class="reference internal" href="#rnn">双层RNN的使用</a><ul>
+<li><a class="reference internal" href="#">训练流程的使用方法</a></li>
+<li><a class="reference internal" href="#">生成流程的使用方法</a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+  <div role="note" aria-label="source link">
+    <h3>This Page</h3>
+    <ul class="this-page-menu">
+      <li><a href="../../_sources/algorithm/rnn/rnn-tutorial.txt"
+            rel="nofollow">Show Source</a></li>
+    </ul>
+   </div>
+<div id="searchbox" style="display: none" role="search">
+  <h3>Quick search</h3>
+    <form class="search" action="../../search.html" method="get">
+      <div><input type="text" name="q" /></div>
+      <div><input type="submit" value="Go" /></div>
+      <input type="hidden" name="check_keywords" value="yes" />
+      <input type="hidden" name="area" value="default" />
+    </form>
+</div>
+<script type="text/javascript">$('#searchbox').show(0);</script>
+        </div>
+      </div>
+      <div class="clearer"></div>
+    </div>
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../../genindex.html" title="General Index"
+             >index</a></li>
+        <li class="nav-item nav-item-0"><a href="../../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>
+    <div class="footer" role="contentinfo">
+        &#169; Copyright 2016, PaddlePaddle developers.
+      Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.4.8.
+    </div>
+  </body>
+</html>
\ No newline at end of file
--- a/doc_cn/faq/index.html
+++ b/doc_cn/faq/index.html
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+  "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+  <head>
+    <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+    <title>PaddlePaddle常见问题 &#8212; PaddlePaddle  documentation</title>
+    <link rel="stylesheet" href="../_static/classic.css" type="text/css" />
+    <link rel="stylesheet" href="../_static/pygments.css" type="text/css" />
+    <script type="text/javascript">
+      var DOCUMENTATION_OPTIONS = {
+        URL_ROOT:    '../',
+        VERSION:     '',
+        COLLAPSE_INDEX: false,
+        FILE_SUFFIX: '.html',
+        HAS_SOURCE:  true
+      };
+    </script>
+    <script type="text/javascript" src="../_static/jquery.js"></script>
+    <script type="text/javascript" src="../_static/underscore.js"></script>
+    <script type="text/javascript" src="../_static/doctools.js"></script>
+    <script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+    <link rel="index" title="Index" href="../genindex.html" />
+    <link rel="search" title="Search" href="../search.html" />
+    <link rel="top" title="PaddlePaddle  documentation" href="../index.html" /> 
+<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>
+  </head>
+  <body role="document">
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../genindex.html" title="General Index"
+             accesskey="I">index</a></li>
+        <li class="nav-item nav-item-0"><a href="../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>  
+    <div class="document">
+      <div class="documentwrapper">
+        <div class="bodywrapper">
+          <div class="body" role="main">
+  <div class="section" id="paddlepaddle">
+<h1><a class="toc-backref" href="#id13">PaddlePaddle常见问题</a><a class="headerlink" href="#paddlepaddle" title="Permalink to this headline">¶</a></h1>
+<div class="contents topic" id="contents">
+<p class="topic-title first">Contents</p>
+<ul class="simple">
+<li><a class="reference internal" href="#paddlepaddle" id="id13">PaddlePaddle常见问题</a><ul>
+<li><a class="reference internal" href="#id1" id="id14">1. 如何减少PaddlePaddle的内存占用</a><ul>
+<li><a class="reference internal" href="#dataprovider" id="id15">减少DataProvider缓冲池内存</a></li>
+<li><a class="reference internal" href="#id3" id="id16">神经元激活内存</a></li>
+<li><a class="reference internal" href="#id4" id="id17">参数内存</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#id5" id="id18">2. 如何加速PaddlePaddle的训练速度</a><ul>
+<li><a class="reference internal" href="#id6" id="id19">减少数据载入的耗时</a></li>
+<li><a class="reference internal" href="#id7" id="id20">加速训练速度</a></li>
+<li><a class="reference internal" href="#id8" id="id21">利用更多的计算资源</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#illegal-instruction" id="id22">3. 遇到“非法指令”或者是“illegal instruction”</a></li>
+<li><a class="reference internal" href="#sgd" id="id23">4. 如何选择SGD算法的学习率</a></li>
+<li><a class="reference internal" href="#id11" id="id24">5. 如何初始化参数</a></li>
+<li><a class="reference internal" href="#id12" id="id25">6. 如何共享参数</a></li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="id1">
+<h2><a class="toc-backref" href="#id14">1. 如何减少PaddlePaddle的内存占用</a><a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h2>
+<p>神经网络的训练本身是一个非常消耗内存和显存的工作。经常会消耗数十G的内存和数G的显存。
+PaddlePaddle的内存占用主要分为如下几个方面:</p>
+<ul class="simple">
+<li>DataProvider缓冲池内存 (只针对内存)</li>
+<li>神经元激活内存 （针对内存和显存）</li>
+<li>参数内存 (针对内存和显存)</li>
+<li>其他内存杂项</li>
+</ul>
+<p>这其中，其他内存杂项是指PaddlePaddle本身所用的一些内存，包括字符串分配，临时变量等等，
+这些内存就不考虑如何缩减了。</p>
+<p>其他的内存的减少方法依次为</p>
+<div class="section" id="dataprovider">
+<h3><a class="toc-backref" href="#id15">减少DataProvider缓冲池内存</a><a class="headerlink" href="#dataprovider" title="Permalink to this headline">¶</a></h3>
+<p>PyDataProvider使用的是异步加载，同时在内存里直接随即选取数据来做Shuffle。即</p>
+<img src="../_images/graphviz-9be6aad37f57c60f4b971dde0ef44ce27179cf9a.png" alt="digraph {
+    rankdir=LR;
+    数据文件 -&gt; 内存池 -&gt; PaddlePaddle训练
+}" />
+<p>所以，减小这个内存池即可减小内存占用，同时也可以加速开始训练前数据载入的过程。但是，这
+个内存池实际上决定了shuffle的粒度。所以，如果将这个内存池减小，又要保证数据是随机的，
+那么最好将数据文件在每次读取之前做一次shuffle。可能的代码为</p>
+<div class="highlight-default"><div class="highlight"><pre><span></span><span class="nd">@provider</span><span class="p">(</span><span class="n">min_pool_size</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span>
+<span class="k">def</span> <span class="nf">process</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">filename</span><span class="p">):</span>
+    <span class="n">os</span><span class="o">.</span><span class="n">system</span><span class="p">(</span><span class="s1">&#39;shuf </span><span class="si">%s</span><span class="s1"> &gt; </span><span class="si">%s</span><span class="s1">.shuf&#39;</span> <span class="o">%</span> <span class="p">(</span><span class="n">filename</span><span class="p">,</span> <span class="n">filename</span><span class="p">))</span>  <span class="c1"># shuffle before.</span>
+    <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="s1">&#39;</span><span class="si">%s</span><span class="s1">.shuf&#39;</span> <span class="o">%</span> <span class="n">filename</span><span class="p">,</span> <span class="s1">&#39;r&#39;</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
+        <span class="k">for</span> <span class="n">line</span> <span class="ow">in</span> <span class="n">f</span><span class="p">:</span>
+            <span class="k">yield</span> <span class="n">get_sample_from_line</span><span class="p">(</span><span class="n">line</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>这样做可以极大的减少内存占用，并且可能会加速训练过程。 详细文档参考 <a class="reference external" href="../ui/data_provider/pydataprovider2.html#provider">这里</a> 。</p>
+</div>
+<div class="section" id="id3">
+<h3><a class="toc-backref" href="#id16">神经元激活内存</a><a class="headerlink" href="#id3" title="Permalink to this headline">¶</a></h3>
+<p>神经网络在训练的时候，会对每一个激活暂存一些数据，包括激活，參差等等。
+在反向传递的时候，这些数据会被用来更新参数。这些数据使用的内存主要和两个参数有关系，
+一是batch size，另一个是每条序列(Sequence)长度。所以，其实也是和每个mini-batch中包含
+的时间步信息成正比。</p>
+<p>所以，做法可以有两种。他们是</p>
+<ul class="simple">
+<li>减小batch size。 即在网络配置中 <code class="code docutils literal"><span class="pre">settings(batch_size=1000)</span></code> 设置成一个小一些的值。但是batch size本身是神经网络的超参数，减小batch size可能会对训练结果产生影响。</li>
+<li>减小序列的长度，或者直接扔掉非常长的序列。比如，一个数据集大部分序列长度是100-200,
+但是突然有一个10000长的序列，就很容易导致内存超限。特别是在LSTM等RNN中。</li>
+</ul>
+</div>
+<div class="section" id="id4">
+<h3><a class="toc-backref" href="#id17">参数内存</a><a class="headerlink" href="#id4" title="Permalink to this headline">¶</a></h3>
+<p>PaddlePaddle支持非常多的优化算法(Optimizer)，不同的优化算法需要使用不同大小的内存。
+例如如果使用 <code class="code docutils literal"><span class="pre">adadelta</span></code> 算法，则需要使用参数规模大约5倍的内存。 如果参数保存下来的
+文件为 <code class="code docutils literal"><span class="pre">100M</span></code>， 那么该优化算法至少需要 <code class="code docutils literal"><span class="pre">500M</span></code> 的内存。</p>
+<p>可以考虑使用一些优化算法，例如 <code class="code docutils literal"><span class="pre">momentum</span></code>。</p>
+</div>
+</div>
+<div class="section" id="id5">
+<h2><a class="toc-backref" href="#id18">2. 如何加速PaddlePaddle的训练速度</a><a class="headerlink" href="#id5" title="Permalink to this headline">¶</a></h2>
+<p>PaddlePaddle是神经网络训练平台，加速PaddlePaddle训练有如下几个方面：</p>
+<ul class="simple">
+<li>减少数据载入的耗时</li>
+<li>加速训练速度</li>
+<li>利用更多的计算资源</li>
+</ul>
+<div class="section" id="id6">
+<h3><a class="toc-backref" href="#id19">减少数据载入的耗时</a><a class="headerlink" href="#id6" title="Permalink to this headline">¶</a></h3>
+<p>使用 <code class="code docutils literal"><span class="pre">pydataprovider`时，可以减少缓存池的大小，同时设置内存缓存功能，即可以极大的加速数据载入流程。</span>
+<span class="pre">:code:`DataProvider</span></code> 缓存池的减小，和之前减小通过减小缓存池来减小内存占用的原理一致。</p>
+<div class="highlight-default"><div class="highlight"><pre><span></span><span class="nd">@provider</span><span class="p">(</span><span class="n">min_pool_size</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span>
+<span class="k">def</span> <span class="nf">process</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">filename</span><span class="p">):</span>
+    <span class="n">os</span><span class="o">.</span><span class="n">system</span><span class="p">(</span><span class="s1">&#39;shuf </span><span class="si">%s</span><span class="s1"> &gt; </span><span class="si">%s</span><span class="s1">.shuf&#39;</span> <span class="o">%</span> <span class="p">(</span><span class="n">filename</span><span class="p">,</span> <span class="n">filename</span><span class="p">))</span>  <span class="c1"># shuffle before.</span>
+    <span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="s1">&#39;</span><span class="si">%s</span><span class="s1">.shuf&#39;</span> <span class="o">%</span> <span class="n">filename</span><span class="p">,</span> <span class="s1">&#39;r&#39;</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
+        <span class="k">for</span> <span class="n">line</span> <span class="ow">in</span> <span class="n">f</span><span class="p">:</span>
+            <span class="k">yield</span> <span class="n">get_sample_from_line</span><span class="p">(</span><span class="n">line</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>同时 <code class="code docutils literal"><span class="pre">&#64;provider</span></code> 接口有一个 <code class="code docutils literal"><span class="pre">cache</span></code> 参数来控制缓存方法，将其设置成 <code class="code docutils literal"><span class="pre">CacheType.CACHE_PASS_IN_MEM</span></code> 的话，会将第一个 <code class="code docutils literal"><span class="pre">pass</span></code> (过完所有训练数据即为一个pass)生成的数据缓存在内存里，在之后的 <code class="code docutils literal"><span class="pre">pass</span></code> 中，不会再从 <code class="code docutils literal"><span class="pre">python</span></code> 端读取数据，而是直接从内存的缓存里读取数据。这也会极大减少数据读入的耗时。</p>
+</div>
+<div class="section" id="id7">
+<h3><a class="toc-backref" href="#id20">加速训练速度</a><a class="headerlink" href="#id7" title="Permalink to this headline">¶</a></h3>
+<p>PaddlePaddle支持Sparse的训练，sparse训练需要训练特征是 <code class="code docutils literal"><span class="pre">sparse_binary_vector</span></code> 、 <code class="code docutils literal"><span class="pre">sparse_vector</span></code> 、或者 <code class="code docutils literal"><span class="pre">integer_value</span></code> 的任一一种。同时，与这个训练数据交互的Layer，需要将其Parameter设置成 sparse 更新模式，即设置 <code class="code docutils literal"><span class="pre">sparse_update=True</span></code></p>
+<p>这里使用简单的 <code class="code docutils literal"><span class="pre">word2vec</span></code> 训练语言模型距离，具体使用方法为:</p>
+<p>使用一个词前两个词和后两个词，来预测这个中间的词。这个任务的DataProvider为:</p>
+<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">DICT_DIM</span><span class="o">=</span><span class="mi">3000</span>
+<span class="nd">@provider</span><span class="p">(</span><span class="n">input_types</span><span class="o">=</span><span class="p">[</span><span class="n">integer_sequence</span><span class="p">(</span><span class="n">DICT_DIM</span><span class="p">),</span> <span class="n">integer_value</span><span class="p">(</span><span class="n">DICT_DIM</span><span class="p">)])</span>
+<span class="k">def</span> <span class="nf">process</span><span class="p">(</span><span class="n">settings</span><span class="p">,</span> <span class="n">filename</span><span class="p">):</span>
+	<span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">filename</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
+		<span class="c1"># yield word ids to predict inner word id</span>
+		<span class="c1"># such as [28, 29, 10, 4], 4</span>
+		<span class="c1"># It means the sentance is  28, 29, 4, 10, 4.</span>
+		<span class="k">yield</span> <span class="n">read_next_from_file</span><span class="p">(</span><span class="n">f</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>这个任务的配置为:</p>
+<div class="highlight-default"><div class="highlight"><pre><span></span><span class="o">...</span> <span class="c1"># the settings and define data provider is omitted.</span>
+<span class="n">DICT_DIM</span><span class="o">=</span><span class="mi">3000</span>  <span class="c1"># dictionary dimension.</span>
+<span class="n">word_ids</span><span class="o">=</span><span class="n">data_layer</span><span class="p">(</span><span class="s1">&#39;word_ids&#39;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">DICT_DIM</span><span class="p">)</span>
+<span class="n">emb</span> <span class="o">=</span> <span class="n">embedding_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">word_ids</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">256</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">sparse_update</span><span class="o">=</span><span class="kc">True</span><span class="p">))</span>
+<span class="n">emb_sum</span> <span class="o">=</span> <span class="n">pooling_layer</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">emb</span><span class="p">,</span> <span class="n">pooling_type</span><span class="o">=</span><span class="n">SumPooling</span><span class="p">())</span>
+<span class="n">predict</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">emb_sum</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">DICT_DIM</span><span class="p">,</span> <span class="n">act</span><span class="o">=</span><span class="n">Softmax</span><span class="p">())</span>
+<span class="n">outputs</span><span class="p">(</span><span class="n">classification_cost</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">predict</span><span class="p">,</span> <span class="n">label</span><span class="o">=</span><span class="n">data_layer</span><span class="p">(</span><span class="s1">&#39;label&#39;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">DICT_DIM</span><span class="p">)))</span> 
+</pre></div>
+</div>
+<p>更多关于sparse训练的内容请参考 <a class="reference external" href="TBD">sparse训练的文档</a></p>
+</div>
+<div class="section" id="id8">
+<h3><a class="toc-backref" href="#id21">利用更多的计算资源</a><a class="headerlink" href="#id8" title="Permalink to this headline">¶</a></h3>
+<p>利用更多的计算资源可以分为一下几个方式来进行:</p>
+<ul class="simple">
+<li>单机CPU训练
+* 使用多线程训练。设置命令行参数 <code class="code docutils literal"><span class="pre">trainer_count</span></code>，即可以设置参与训练的线程数量。使用方法为 <code class="code docutils literal"><span class="pre">paddle</span> <span class="pre">train</span> <span class="pre">--trainer_count=4</span></code></li>
+<li>单机GPU训练
+* 使用显卡训练。设置命令行参数 <code class="code docutils literal"><span class="pre">use_gpu</span></code>。 使用方法为 <code class="code docutils literal"><span class="pre">paddle</span> <span class="pre">train</span> <span class="pre">--use_gpu=true</span></code>
+* 使用多块显卡训练。设置命令行参数 <code class="code docutils literal"><span class="pre">use_gpu</span></code> 和 <code class="code docutils literal"><span class="pre">trainer_count</span></code>。使用 <code class="code docutils literal"><span class="pre">--use_gpu=True</span></code> 开启GPU训练，使用 <code class="code docutils literal"><span class="pre">trainer_count</span></code> 指定显卡数量。使用方法为 <code class="code docutils literal"><span class="pre">paddle</span> <span class="pre">train</span> <span class="pre">--use_gpu=true</span> <span class="pre">--trainer_count=4</span></code></li>
+<li>多机训练
+* 使用多机训练的方法也比较简单，需要先在每个节点启动 <code class="code docutils literal"><span class="pre">paddle</span> <span class="pre">pserver</span></code>，在使用 <code class="code docutils literal"><span class="pre">paddle</span> <span class="pre">train</span> <span class="pre">--pservers=192.168.100.1,192.168.100.2</span></code> 来指定每个pserver的ip地址
+* 具体的多机训练方法参考 <a class="reference external" href="TBD">多机训练</a> 文档。</li>
+</ul>
+</div>
+</div>
+<div class="section" id="illegal-instruction">
+<h2><a class="toc-backref" href="#id22">3. 遇到“非法指令”或者是“illegal instruction”</a><a class="headerlink" href="#illegal-instruction" title="Permalink to this headline">¶</a></h2>
+<p>paddle在进行计算的时候为了提升计算性能，使用了avx指令。部分老的cpu型号无法支持这样的指令。通常来说执行下grep avx /proc/cpuinfo看看是否有输出即可知道是否支持。（另：用此方法部分虚拟机可能检测到支持avx指令但是实际运行会挂掉，请当成是不支持，看下面的解决方案）</p>
+<p>解决办法是:</p>
+<ul class="simple">
+<li>使用 NO_AVX的 <a class="reference external" href="../build_and_install/index.html">安装包</a> 或者 <a class="reference external" href="../build_and_install/install/docker_install.html">Docker image</a></li>
+<li>或者，使用 <code class="code docutils literal"><span class="pre">-DWITH_AVX=OFF</span></code> 重新编译PaddlePaddle。</li>
+</ul>
+</div>
+<div class="section" id="sgd">
+<h2><a class="toc-backref" href="#id23">4. 如何选择SGD算法的学习率</a><a class="headerlink" href="#sgd" title="Permalink to this headline">¶</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="id11">
+<h2><a class="toc-backref" href="#id24">5. 如何初始化参数</a><a class="headerlink" href="#id11" title="Permalink to this headline">¶</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="id12">
+<h2><a class="toc-backref" href="#id25">6. 如何共享参数</a><a class="headerlink" href="#id12" title="Permalink to this headline">¶</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，是想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>
+<div class="highlight-default"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">paddle.trainer_config_helpers</span> <span class="k">import</span> <span class="o">*</span>
+<span class="n">settings</span><span class="p">(</span>
+    <span class="n">learning_rate</span><span class="o">=</span><span class="mi">1</span><span class="n">e</span><span class="o">-</span><span class="mi">4</span><span class="p">,</span>
+    <span class="n">batch_size</span><span class="o">=</span><span class="mi">1000</span>
+<span class="p">)</span>
+<span class="n">a</span> <span class="o">=</span> <span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;feature_a&#39;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">200</span><span class="p">)</span>
+<span class="n">b</span> <span class="o">=</span> <span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;feature_b&#39;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">200</span><span class="p">)</span>
+<span class="n">fc_param</span> <span class="o">=</span> <span class="n">ParamAttr</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;fc_param&#39;</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_param</span> <span class="o">=</span> <span class="n">ParamAttr</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;bias_param&#39;</span><span class="p">,</span> <span class="n">initial_mean</span><span class="o">=</span><span class="mf">0.0</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>
+<span class="n">softmax_param</span> <span class="o">=</span> <span class="n">ParamAttr</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;softmax_param&#39;</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">hidden_a</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">a</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">200</span><span class="p">,</span> <span class="n">param_attr</span><span class="o">=</span><span class="n">fc_param</span><span class="p">,</span> <span class="n">bias_attr</span><span class="o">=</span><span class="n">bias_param</span><span class="p">)</span>
+<span class="n">hidden_b</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">b</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">200</span><span class="p">,</span> <span class="n">param_attr</span><span class="o">=</span><span class="n">fc_param</span><span class="p">,</span> <span class="n">bias_attr</span><span class="o">=</span><span class="n">bias_param</span><span class="p">)</span>
+<span class="n">predict</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="p">[</span><span class="n">hidden_a</span><span class="p">,</span> <span class="n">hidden_b</span><span class="p">],</span> <span class="n">param_attr</span><span class="o">=</span><span class="p">[</span><span class="n">softmax_param</span><span class="p">,</span> <span class="n">softmax_param</span><span class="p">],</span>
+                   <span class="n">bias_attr</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">10</span><span class="p">,</span> <span class="n">act</span><span class="o">=</span><span class="n">SoftmaxActivation</span><span class="p">())</span>
+<span class="n">outputs</span><span class="p">(</span><span class="n">classification_cost</span><span class="p">(</span><span class="nb">input</span><span class="o">=</span><span class="n">predict</span><span class="p">,</span> <span class="n">label</span><span class="o">=</span><span class="n">data_layer</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="s1">&#39;label&#39;</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="mi">10</span><span class="p">)))</span>
+</pre></div>
+</div>
+<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>
+          </div>
+        </div>
+      </div>
+      <div class="sphinxsidebar" role="navigation" aria-label="main navigation">
+        <div class="sphinxsidebarwrapper">
+  <h3><a href="../index.html">Table Of Contents</a></h3>
+  <ul>
+<li><a class="reference internal" href="#">PaddlePaddle常见问题</a><ul>
+<li><a class="reference internal" href="#id1">1. 如何减少PaddlePaddle的内存占用</a><ul>
+<li><a class="reference internal" href="#dataprovider">减少DataProvider缓冲池内存</a></li>
+<li><a class="reference internal" href="#id3">神经元激活内存</a></li>
+<li><a class="reference internal" href="#id4">参数内存</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#id5">2. 如何加速PaddlePaddle的训练速度</a><ul>
+<li><a class="reference internal" href="#id6">减少数据载入的耗时</a></li>
+<li><a class="reference internal" href="#id7">加速训练速度</a></li>
+<li><a class="reference internal" href="#id8">利用更多的计算资源</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#illegal-instruction">3. 遇到“非法指令”或者是“illegal instruction”</a></li>
+<li><a class="reference internal" href="#sgd">4. 如何选择SGD算法的学习率</a></li>
+<li><a class="reference internal" href="#id11">5. 如何初始化参数</a></li>
+<li><a class="reference internal" href="#id12">6. 如何共享参数</a></li>
+</ul>
+</li>
+</ul>
+  <div role="note" aria-label="source link">
+    <h3>This Page</h3>
+    <ul class="this-page-menu">
+      <li><a href="../_sources/faq/index.txt"
+            rel="nofollow">Show Source</a></li>
+    </ul>
+   </div>
+<div id="searchbox" style="display: none" role="search">
+  <h3>Quick search</h3>
+    <form class="search" action="../search.html" method="get">
+      <div><input type="text" name="q" /></div>
+      <div><input type="submit" value="Go" /></div>
+      <input type="hidden" name="check_keywords" value="yes" />
+      <input type="hidden" name="area" value="default" />
+    </form>
+</div>
+<script type="text/javascript">$('#searchbox').show(0);</script>
+        </div>
+      </div>
+      <div class="clearer"></div>
+    </div>
+    <div class="related" role="navigation" aria-label="related navigation">
+      <h3>Navigation</h3>
+      <ul>
+        <li class="right" style="margin-right: 10px">
+          <a href="../genindex.html" title="General Index"
+             >index</a></li>
+        <li class="nav-item nav-item-0"><a href="../index.html">PaddlePaddle  documentation</a> &#187;</li> 
+      </ul>
+    </div>
+    <div class="footer" role="contentinfo">
+        &#169; Copyright 2016, PaddlePaddle developers.
+      Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.4.8.
+    </div>
+  </body>
+</html>
\ No newline at end of file
--- a/doc_cn/index.html
+++ b/doc_cn/index.html
@@ -78,7 +78,16 @@ var _hmt = _hmt || [];
 <div class="section" id="id9">
 <h2>算法教程<a class="headerlink" href="#id9" title="Permalink to this headline">¶</a></h2>
 <ul class="simple">
-<li><a class="reference external" href="../doc/algorithm/rnn/rnn.html">RNN配置</a></li>
+<li><a class="reference external" href="algorithm/rnn/rnn-tutorial.html">Recurrent Group教程</a></li>
+<li><a class="reference external" href="../doc/algorithm/rnn/rnn.html">单层RNN示例</a></li>
+<li><a class="reference external" href="algorithm/rnn/hierarchical-rnn.html">双层RNN示例</a></li>
+<li><a class="reference external" href="algorithm/rnn/hierarchical-layer.html">支持双层序列作为输入的Layer</a></li>
+</ul>
+</div>
+<div class="section" id="id12">
+<h2>常见问题<a class="headerlink" href="#id12" title="Permalink to this headline">¶</a></h2>
+<ul class="simple">
+<li><a class="reference external" href="faq/index.html">常见问题</a></li>
 </ul>
 </div>
 </div>
@@ -95,6 +104,7 @@ var _hmt = _hmt || [];
 <li><a class="reference internal" href="#id1">使用指南</a></li>
 <li><a class="reference internal" href="#id8">开发指南</a></li>
 <li><a class="reference internal" href="#id9">算法教程</a></li>
+<li><a class="reference internal" href="#id12">常见问题</a></li>
 </ul>
 </li>
 </ul>

--- a/doc_cn/objects.inv
+++ b/doc_cn/objects.inv
--- a/doc_cn/searchindex.js
+++ b/doc_cn/searchindex.js
-Search.setIndex({envversion:49,filenames:["build/docker/build_docker_image","build_and_install/cmake/compile_options","build_and_install/cmake/index","build_and_install/cmake/install_deps","build_and_install/cmake/make_and_install","build_and_install/index","build_and_install/install/docker_install","build_and_install/install/ubuntu_install","cluster/index","demo/index","demo/quick_start/index","index","ui/cmd/dump_config","ui/cmd/index","ui/cmd/make_diagram","ui/cmd/merge_model","ui/cmd/paddle_pserver","ui/cmd/paddle_train","ui/cmd/paddle_version","ui/data_provider/index","ui/data_provider/pydataprovider2","ui/data_provider/write_new_dataprovider","ui/index","ui/predict/swig_py_paddle"],objects:{},objnames:{},objtypes:{},terms:{"0000x":10,"000\u5f20\u7070\u5ea6\u56fe\u7247\u7684\u6570\u5b57\u5206\u7c7b\u6570\u636e\u96c6":20,"00186201e":23,"04\u4e2d\u6b63\u786e":7,"08823112e":23,"0b1":7,"10\u4ee5\u4e0a\u7684linux":6,"10\u7ef4\u7684\u6574\u6570\u503c":20,"10gbe":6,"10m":0,"12194102e":23,"12\u7248\u672c\u6d4b\u8bd5\u901a\u8fc7":0,"12\u7248\u672c\u7684\u60c5\u51b5\u4e0b\u5e76\u6ca1\u6709\u6d4b\u8bd5":0,"15501715e":23,"15mb":10,"16mb":10,"1\u7684\u8bdd":20,"252kb":10,"25639710e":23,"25k":10,"27787406e":23,"28\u7684\u50cf\u7d20\u7070\u5ea6\u503c":20,"28\u7684\u7a20\u5bc6\u5411\u91cf\u548c\u4e00\u4e2a":20,"2\u8fdb\u884c\u8fdb\u4e00\u6b65\u6f14\u5316":10,"32777140e":23,"36540484e":23,"40gbe":6,"43630644e":23,"48565123e":23,"48684503e":23,"49316648e":23,"50k":0,"51111044e":23,"53018653e":23,"56gbe":6,"5\u5230\u672c\u5730\u73af\u5883\u4e2d":7,"70634608e":23,"72296313e":23,"85625684e":23,"93137714e":23,"96644767e":23,"99982715e":23,"9\u7684\u6570\u5b57":20,"\u4e00\u4e2a\u6587\u4ef6":20,"\u4e00\u4e2alogging\u5bf9\u8c61":20,"\u4e00\u4e2apass\u8868\u793a\u8fc7\u4e00\u904d\u6240\u6709\u8bad\u7ec3\u6837\u672c":10,"\u4e00\u6761":20,"\u4e00\u81f4\u7684\u7279\u5f81":20,"\u4e00\u822c\u60c5\u51b5\u4e0b":19,"\u4e00\u822c\u63a8\u8350\u8bbe\u7f6e\u6210true":20,"\u4e00\u884c\u4e3a\u4e00\u4e2a\u6837\u672c":10,"\u4e00\u884c\u5bf9\u5e94\u4e00\u4e2a\u6570\u636e\u6587\u4ef6":19,"\u4e0a\u7684\u6548\u679c":10,"\u4e0b\u8f7d\u8fdb\u7a0b\u4f1a\u91cd\u542f":0,"\u4e0b\u8ff0\u5185\u5bb9\u5c06\u5206\u4e3a\u5982\u4e0b\u51e0\u4e2a\u7c7b\u522b\u63cf\u8ff0":6,"\u4e0b\u975e\u5e38\u5c11\u7684\u53d8\u91cf\u5f15\u7528":20,"\u4e0b\u9762dataprovid":10,"\u4e0d\u4e00\u5b9a\u548c\u65f6\u95f4\u6709\u5173\u7cfb":20,"\u4e0d\u4f1a\u6267\u884c\u6d4b\u8bd5\u64cd\u4f5c":19,"\u4e0d\u5305\u542blabel":23,"\u4e0d\u540c\u7684\u6570\u636e\u7c7b\u578b\u548c\u5e8f\u5217\u6a21\u5f0f\u8fd4\u56de\u7684\u683c\u5f0f\u4e0d\u540c":20,"\u4e0d\u652f\u6301avx\u6307\u4ee4\u96c6\u7684cpu\u4e5f\u53ef\u4ee5\u8fd0\u884c":6,"\u4e0d\u7f13\u5b58\u4efb\u4f55\u6570\u636e":20,"\u4e0d\u9700\u8981avx\u6307\u4ee4\u96c6\u7684cpu\u4e5f\u53ef\u4ee5\u8fd0\u884c":6,"\u4e0e\u8bad\u7ec3\u7f51\u7edc\u914d\u7f6e\u4e0d\u540c\u7684\u662f":10,"\u4e14\u5e8f\u5217\u7684\u6bcf\u4e00\u4e2a\u5143\u7d20\u8fd8\u662f\u4e00\u4e2a\u65f6\u95f4\u5e8f\u5217":20,"\u4e24\u4e2a\u6587\u6863":6,"\u4e24\u7c7b":10,"\u4e25\u91cd\u7684\u95ee\u9898":20,"\u4e2a":10,"\u4e2ayield":20,"\u4e2d":10,"\u4e2d\u5b9a\u4e49\u4f7f\u7528\u54ea\u79cddataprovider\u53ca\u5176\u53c2\u6570":19,"\u4e2d\u5b9a\u4e49\u7684\u987a\u5e8f\u4e00\u81f4":20,"\u4e2d\u5bfb\u627e\u8fd9\u4e9bblas\u7684\u5b9e\u73b0":1,"\u4e2d\u7684":20,"\u4e2d\u7684\u4e8c\u8fdb\u5236\u4f7f\u7528\u4e86":6,"\u4e2d\u7684set":20,"\u4e2d\u914d\u7f6e":20,"\u4e3a":20,"\u4e3a\u4e86\u8fd0\u884cpaddlepaddle\u7684docker\u955c\u50cf":6,"\u4e3a\u4f8b\u8fdb\u884c\u9884\u6d4b":10,"\u4e3b\u8981\u51fd\u6570\u662fprocess\u51fd\u6570":20,"\u4e3b\u8981\u5206\u4e3a\u4ee5\u4e0b\u51e0\u4e2a\u6b65\u9aa4":23,"\u4e3b\u8981\u5305\u62ec\u4e24\u90e8\u5206":10,"\u4e3b\u8981\u662f\u589e\u52a0\u4e86\u521d\u59cb\u5316\u673a\u5236":20,"\u4e3b\u8981\u6b65\u9aa4\u4e3a":23,"\u4e3b\u8981\u7531\u4e8e\u65e7\u7248\u672c":0,"\u4e3b\u8981\u7684\u8f6f\u4ef6\u5305\u4e3apy_paddl":23,"\u4e5f\u4f1a\u6254\u5230\u8fd9\u6761\u6570\u636e":20,"\u4e5f\u4f1a\u8bfb\u53d6\u76f8\u5173\u8def\u5f84\u53d8\u91cf\u6765\u8fdb\u884c\u641c\u7d22":1,"\u4e5f\u53ef\u4ee5":20,"\u4e5f\u53ef\u4ee5\u4f7f\u7528":20,"\u4e5f\u53ef\u4ee5\u548cpaddl":13,"\u4e5f\u53ef\u4ee5\u76f4\u63a5\u6267\u884c":6,"\u4e5f\u53ef\u4ee5\u901a\u8fc7\u5982\u4e0b\u65b9\u5f0f\u9884\u6d4b":10,"\u4e5f\u53ef\u4ee5\u901a\u8fc7save":10,"\u4e5f\u53ef\u4ee5\u9884\u6d4b\u6ca1\u6709label\u7684\u6d4b\u8bd5\u96c6":10,"\u4e5f\u5c31\u662f\u5c06\u8bcd\u5411\u91cf\u6a21\u578b\u989d\u6b65":10,"\u4e5f\u5c31\u662f\u76f4\u63a5\u5199\u5185\u5b58\u7684float\u6570\u7ec4":23,"\u4e5f\u9700\u8981\u4e24\u6b21\u968f\u673a\u9009\u62e9\u5230\u540c\u6837\u7684generator\u7684\u65f6\u5019":20,"\u4e86\u975e\u5e38\u65b9\u4fbf\u7684\u4e8c\u8fdb\u5236\u5206\u53d1\u624b\u6bb5":6,"\u4e8c\u6b21\u5f00\u53d1\u53ef\u4ee5":6,"\u4eba\u5458\u7b49\u7b49":0,"\u4ec5\u4ec5\u9700\u8981":20,"\u4ecb\u7ecdpaddlepaddle\u4f7f\u7528\u6d41\u7a0b\u548c\u5e38\u7528\u7684\u7f51\u7edc\u57fa\u7840\u5355\u5143\u7684\u914d\u7f6e\u65b9\u6cd5":10,"\u4ece\u6587\u4ef6\u4e2d\u8bfb\u53d6\u6bcf\u4e00\u6761\u6570\u636e":20,"\u4ece\u6587\u672c\u6587\u4ef6\u4e2d\u8bfb\u53d6":20,"\u4ece\u800c\u4e0d\u80fd\u5728\u8fd0\u884c\u7f16\u8bd1\u547d\u4ee4\u7684\u65f6\u5019\u63a5\u53d7\u53c2\u6570":0,"\u4ece\u800c\u751f\u6210\u591a\u4e2agener":20,"\u4ece\u800c\u9632\u6b62\u8fc7\u62df\u5408":19,"\u4ed6\u4eec\u662f":[6,7,13,20],"\u4ed6\u4eec\u7684imag":6,"\u4ed6\u53ef\u4ee5\u5c06\u67d0\u4e00\u4e2a\u51fd\u6570\u6807\u8bb0\u6210\u4e00\u4e2apydataprovid":20,"\u4ee4\u884c\u53c2\u6570\u6587\u6863":10,"\u4ee5\u592a\u7f51\u5361":6,"\u4ee5\u76f8\u5bf9\u8def\u5f84\u5f15\u7528":19,"\u4efb\u610f\u4e00\u79cdcblas\u5b9e\u73b0":1,"\u4f1a\u62a5\u5bfb\u627e\u4e0d\u5230\u8fd9\u4e9b\u52a8\u6001\u5e93":7,"\u4f1a\u6839\u636e\u547d\u4ee4\u884c\u53c2\u6570\u6307\u5b9a\u7684\u6d4b\u8bd5\u65b9\u5f0f":19,"\u4f1a\u6839\u636einput_types\u68c0\u67e5\u6570\u636e\u7684\u5408\u6cd5\u6027":20,"\u4f1a\u751f\u6210\u591a\u4e2agener":20,"\u4f1a\u9884\u5148\u8bfb\u53d6\u5168\u90e8\u6570\u636e\u5230\u5185\u5b58\u4e2d":20,"\u4f20\u5165\u547d\u4ee4\u884c\u53c2\u6570\u521d\u59cb\u5316":23,"\u4f20\u5165\u7684\u662f\u4e00\u4e2a\u51fd\u6570":20,"\u4f20\u5165\u7684\u914d\u7f6e\u53c2\u6570\u5305\u62ec":0,"\u4f20\u5165\u8fd9\u4e2a\u53d8\u91cf\u7684\u65b9\u5f0f\u4e3a":20,"\u4f46\u662f":[0,20],"\u4f46\u662f\u5728":20,"\u4f46\u662f\u5982\u679c\u4f7f\u7528\u4e86\u9ad8\u6027\u80fd\u7684\u7f51\u5361":6,"\u4f46\u662f\u65b9\u4fbf\u8c03\u8bd5\u548cbenchmark":1,"\u4f46\u662f\u6709\u65f6\u4e3a\u4e86\u8ba1\u7b97\u5747\u8861\u6027":20,"\u4f46\u7406\u8bba\u4e0a\u652f\u6301\u5176\u4ed6\u7684":7,"\u4f46\u9700\u8981\u6ce8\u610f\u7684\u662f\u7f16\u8bd1\u548c":1,"\u4f4e\u4e8edocker":0,"\u4f53\u53ef\u4ee5\u53c2\u8003":20,"\u4f7f\u5728python\u73af\u5883\u4e0b\u7684\u9884\u6d4b\u63a5\u53e3\u66f4\u52a0\u7b80\u5355":23,"\u4f7f\u7528":[1,6,23],"\u4f7f\u7528\u5982\u4e0b\u811a\u672c\u53ef\u4ee5\u786e\u5b9a\u672c\u673a\u7684cpu\u77e5\u5426\u652f\u6301":6,"\u4f7f\u7528\u7684\u547d\u4ee4\u4e5f\u662f":1,"\u4f7f\u7528\u8be5\u63a5\u53e3\u7528\u6237\u53ef\u4ee5\u53ea\u5173\u6ce8\u5982\u4f55":20,"\u4f7f\u7528\u8be5dockerfile\u6784\u5efa\u51fa\u955c\u50cf":6,"\u4f7f\u7528\u8fd9\u4e2a\u5173\u952e\u8bcd":20,"\u4f7f\u7528deb\u5305\u5728ubuntu\u4e0a\u5b89\u88c5paddlepaddl":5,"\u4f7f\u7528dockerfile\u6784\u5efa\u4e00\u4e2a\u5168\u65b0\u7684dock":6,"\u4f7f\u7528mnist\u624b\u5199\u8bc6\u522b\u4f5c\u4e3a\u6837\u4f8b":23,"\u4f7f\u7528ssh\u8bbf\u95eepaddlepaddle\u955c\u50cf":6,"\u4f8b\u5982":[1,10,20],"\u4f8b\u5982\u6587\u4ef6\u540d\u662f":20,"\u4f8b\u5982rdma\u7f51\u5361":6,"\u4f8b\u5982sigmoid\u53d8\u6362":10,"\u4f9d\u6b21\u8fd4\u56de\u4e86\u6587\u4ef6\u4e2d\u7684\u6bcf\u6761\u6570\u636e":20,"\u4fe1\u606f":6,"\u5173\u95edcontain":6,"\u5176\u4e2d":[0,6,7,19,20,23],"\u5176\u4e2d\u6587\u672c\u8f93\u5165\u7c7b\u578b\u5b9a\u4e49\u4e3a\u6574\u6570\u65f6\u5e8f\u7c7b\u578binteg":10,"\u5176\u4e2d\u7b2c\u4e00\u884c\u662f\u5f15\u5165paddlepaddle\u7684pydataprovider2\u5305":20,"\u5176\u4ed6\u53c2\u6570\u5747\u4f7f\u7528kei":20,"\u5176\u4ed6\u53c2\u6570\u8bf7\u53c2\u8003":10,"\u5176\u4ed6\u53c2\u6570\u90fd\u4f7f\u7528kei":20,"\u5176\u4f5c\u7528\u662f\u5c06\u8bad\u7ec3\u6570\u636e\u4f20\u5165\u5185\u5b58\u6216\u8005\u663e\u5b58":19,"\u5176\u5b83\u90e8\u5206\u548c\u903b\u8f91\u56de\u5f52\u7f51\u7edc\u7ed3\u6784\u4e00\u81f4":10,"\u5176\u6570\u636e\u4f7f\u7528":20,"\u5176\u7b2c\u4e00\u884c\u8bf4\u660e\u4e86paddle\u7684\u7248\u672c":18,"\u5177":20,"\u5177\u4f53\u53ef\u4ee5\u8bbe\u7f6e\u6210\u4ec0\u4e48\u5176\u4ed6\u683c":20,"\u5177\u4f53\u6709\u54ea\u4e9b\u683c\u5f0f":20,"\u5177\u4f53\u8bf7\u53c2\u8003\u6ce8\u610f\u4e8b\u9879\u4e2d\u7684":6,"\u5177\u4f53dataprovider\u8fd8\u5177\u6709\u4ec0\u4e48\u529f\u80fd":20,"\u5177\u6709\u4e24\u4e2a\u53c2\u6570":20,"\u5177\u6709\u548c\u5bbf\u4e3b\u673a\u76f8\u8fd1\u7684\u8fd0\u884c\u6548\u7387":6,"\u5177\u6709\u7684\u5c5e\u6027\u6709":20,"\u5178\u578b\u7684\u8f93\u51fa\u7ed3\u679c\u4e3a":23,"\u5178\u578b\u7684\u9884\u6d4b\u4ee3\u7801\u5982\u4e0b":23,"\u5185\u5b58\u4e0d\u591f\u7528\u7684\u60c5\u51b5":19,"\u518d\u6307\u5b9a":1,"\u5199\u5165train":20,"\u5199\u5728train":19,"\u51c6\u5907\u6570\u636e":23,"\u51fd\u6570":20,"\u51fd\u6570\u4e2d":20,"\u51fd\u6570\u4e2d\u4f7f\u7528":20,"\u51fd\u6570\u4e2d\u7684":20,"\u51fd\u6570\u662f\u4f7f\u7528":20,"\u51fd\u6570\u6765\u4fdd\u8bc1\u517c\u5bb9\u6027":20,"\u51fd\u6570\u67e5\u8be2\u6587\u6863":23,"\u5206\u5e03\u5f0f\u8bad\u7ec3":10,"\u5206\u7c7b\u6210\u6b63\u9762\u60c5\u7eea\u548c":20,"\u5217\u8868\u5982\u4e0b":20,"\u5219\u53ef\u4ee5\u4f7f\u7528":6,"\u5219\u53ef\u4ee5\u9009\u62e9\u4e0a\u8868\u4e2d\u7684avx\u7248\u672cpaddlepaddl":6,"\u5219\u5b57\u4e0e\u5b57\u4e4b\u95f4\u7528\u7a7a\u683c\u5206\u9694":10,"\u5219\u9700\u8981\u4f7f\u7528":7,"\u5219\u9700\u8981\u5148\u5c06":6,"\u5219\u9700\u8981\u8fdb\u884c\u4e00\u5b9a\u7684\u4e8c\u6b21\u5f00\u53d1":6,"\u521b\u5efa\u4e00\u4e2a":23,"\u521b\u5efagener":20,"\u5220\u9664contain":6,"\u5229\u7528\u5355\u8bcdid\u67e5\u627e\u5bf9\u5e94\u7684\u8be5\u5355\u8bcd\u7684\u8fde\u7eed\u8868\u793a\u5411\u91cf":10,"\u5229\u7528\u903b\u8f91\u56de\u5f52\u6a21\u578b\u5bf9\u8be5\u5411\u91cf\u8fdb\u884c\u5206\u7c7b":10,"\u522b":10,"\u5269\u4e0b\u7684pass\u4f1a\u76f4\u63a5\u4ece\u5185\u5b58\u91cc":20,"\u52a0\u4e86l2\u6b63\u5219\u548c\u68af\u5ea6\u622a\u65ad":10,"\u52a0\u8f7d\u6570\u636e":10,"\u5305":6,"\u5305\u548c":6,"\u5305\u62ec":10,"\u5305\u62ec\u7b80\u5355\u7684rnn\u6a21\u578b":10,"\u5305\u62ecdocker\u955c\u50cf":5,"\u5305\u62ecpaddle\u7684\u4e8c\u8fdb\u5236":6,"\u5305\u62ecpaddle\u8fd0\u884cdemo\u6240\u9700\u8981\u7684\u4f9d\u8d56":6,"\u5373":[6,10],"\u5373\u4e0d\u5728\u4e4e\u5185\u5b58\u6682\u5b58\u591a\u5c11\u6761\u6570\u636e":20,"\u5373\u4e0d\u662f\u4e00\u6761\u5e8f\u5217":20,"\u5373\u4ece\u5355\u8bcd\u5b57\u7b26\u4e32\u5230\u5355\u8bcdid\u7684\u5b57\u5178":20,"\u5373\u4f1a\u751f\u6210100\u4e2agener":20,"\u5373\u4f7f\u5728check\u4e2d\u6570\u636e\u4e0d\u5408\u6cd5":20,"\u5373\u4f7f\u5728process\u91cc\u9762\u53ea\u4f1a\u6709\u4e00":20,"\u5373\u5305\u542b\u65f6\u95f4\u6b65\u4fe1\u606f":20,"\u5373\u53ef":20,"\u5373\u53ef\u4ee5\u4f7f\u7528ssh\u8bbf\u95ee\u5bbf\u4e3b\u673a\u76848022\u7aef\u53e3":6,"\u5373\u53ef\u542f\u52a8\u548c\u8fdb\u5165paddlepaddle\u7684contain":6,"\u5373\u53ef\u5728\u672c\u5730\u7f16\u8bd1\u51fapaddlepaddle\u7684\u955c\u50cf":0,"\u5373\u53ef\u6253\u5370\u51fapaddlepaddle\u7684\u7248\u672c\u548c\u6784\u5efa":6,"\u5373\u5927\u90e8\u5206\u503c\u4e3a0":20,"\u5373\u5982\u679ctrain":20,"\u5373\u5bf9\u7b2c3\u6b65\u8fdb\u884c\u66ff\u6362":10,"\u5373\u662f\u4e00\u6761\u65f6\u95f4\u5e8f\u5217":20,"\u5373train":20,"\u5377\u79ef\u7f51\u7edc\u662f\u4e00\u79cd\u7279\u6b8a\u7684\u4ece\u8bcd\u5411\u91cf\u8868\u793a\u5230\u53e5\u5b50\u8868\u793a\u7684\u65b9\u6cd5":10,"\u53c2\u6570":0,"\u53c2\u6570\u6570\u91cf":10,"\u53c2\u8003":19,"\u53c2\u89c1":[3,4],"\u53d1\u884c\u7248":7,"\u53d6\u51b3\u4e8e\u662f\u5426\u5bfb\u627e\u5230gflags":1,"\u53d6\u51b3\u4e8e\u662f\u5426\u5bfb\u627e\u5230glog":1,"\u53d6\u51b3\u4e8e\u662f\u5426\u5bfb\u627e\u5230gtest":1,"\u53d6\u51b3\u4e8e\u662f\u5426\u627e\u5230swig":1,"\u53d8\u4e3a3\u4e2a\u65b0\u7684\u5b50\u6b65\u9aa4":10,"\u53d8\u4f1a\u62a5\u8fd9\u4e2a\u9519\u8bef":7,"\u53d8\u91cf":20,"\u53e5\u5b50\u8868\u793a\u7684\u8ba1\u7b97\u66f4\u65b0\u4e3a2\u6b65":10,"\u53ea\u5305\u62ecpaddle\u7684\u4e8c\u8fdb\u5236":6,"\u53ea\u662f\u5c06\u53e5\u5b50\u5229\u7528\u8fde\u7eed\u5411\u91cf\u8868\u793a\u66ff\u6362\u7a00\u758f":10,"\u53ea\u662f\u8bf4\u660e\u6570\u636e\u7684\u987a\u5e8f\u662f\u91cd\u8981\u7684":20,"\u53ea\u80fd\u591f\u8fd4\u56delist\u6216\u8005tupl":20,"\u53ea\u9700\u8981\u4f7f\u7528\u4e00\u884c\u4ee3\u7801\u5373\u53ef\u4ee5\u8bbe\u7f6e\u8bad\u7ec3\u5f15\u7528\u8fd9\u4e2adataprovid":20,"\u53ea\u9700\u8981\u5728":20,"\u53ea\u9700\u8981\u77e5\u9053\u8fd9\u53ea\u662f\u4e00\u4e2a\u6807\u8bb0\u5c5e\u6027\u7684\u65b9\u6cd5\u5c31\u53ef\u4ee5\u4e86":20,"\u53ef\u4ee5\u4e3a\u4e00\u4e2a\u6570\u636e\u6587\u4ef6\u8fd4\u56de\u591a\u6761\u8bad\u7ec3\u6837\u672c":20,"\u53ef\u4ee5\u4f20\u516510k":0,"\u53ef\u4ee5\u4f7f\u7528":0,"\u53ef\u4ee5\u4f7f\u7528\u547d\u4ee4":7,"\u53ef\u4ee5\u4f7f\u7528\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u8bc4\u4f30\u5e26\u6709label\u7684\u9a8c\u8bc1\u96c6":10,"\u53ef\u4ee5\u4f7f\u7528graphviz\u5bf9paddlepaddle\u7684\u7f51\u7edc\u6a21\u578b\u8fdb\u884c\u7ed8\u5236":13,"\u53ef\u4ee5\u4f7f\u7528paddl":13,"\u53ef\u4ee5\u4f7f\u7528python\u7684":23,"\u53ef\u4ee5\u53c2\u8003":10,"\u53ef\u4ee5\u5728\u4e00\u4e2a\u51fd\u6570\u91cc":20,"\u53ef\u4ee5\u5728cmake\u7684\u547d\u4ee4\u884c\u8bbe\u7f6e":1,"\u53ef\u4ee5\u5c06\u4e00\u6761\u6570\u636e\u8bbe\u7f6e\u6210\u591a\u4e2abatch":20,"\u53ef\u4ee5\u5c06paddlepaddle\u7684\u6a21\u578b\u548c\u914d\u7f6e\u6253\u5305\u6210\u4e00\u4e2a\u6587\u4ef6":13,"\u53ef\u4ee5\u5c06paddlepaddle\u7684\u8bad\u7ec3\u6a21\u578b\u4ee5proto":13,"\u53ef\u4ee5\u65b9\u4fbf\u5d4c\u5165\u5f0f\u5de5\u4f5c":1,"\u53ef\u4ee5\u6839\u636e\u4e0d\u540c\u7684\u6570\u636e\u914d\u7f6e\u4e0d\u540c\u7684\u8f93\u5165\u7c7b\u578b":20,"\u53ef\u4ee5\u8fd4\u56de\u4e00\u4e2adict":20,"\u53ef\u4ee5\u901a\u8fc7show":10,"\u53ef\u7528\u5728\u6d4b\u8bd5\u6216\u8bad\u7ec3\u65f6\u6307\u5b9a\u521d\u59cb\u5316\u6a21\u578b":10,"\u53ef\u80fd\u7684\u5185\u5b58\u6cc4\u9732\u95ee\u9898":19,"\u53ef\u80fd\u7684\u8f93\u51fa\u4e3a":7,"\u53ef\u9009":20,"\u5404\u79cd\u53c2\u6570\u548c\u7ef4\u62a4":0,"\u540c\u65f6":[0,1,20],"\u540c\u65f6\u4f1a\u8ba1\u7b97\u5206\u7c7b\u51c6\u786e\u7387":10,"\u540c\u65f6\u6b22\u8fce\u8d21\u732e\u66f4\u591a\u7684\u5b89\u88c5\u5305":5,"\u540d\u79f0":10,"\u540e\u9762\u8ddf\u7740\u4e00\u7cfb\u5217\u7f16\u8bd1\u53c2\u6570":18,"\u5411\u91cf\u8868\u793a":10,"\u5426":1,"\u5426\u5219":19,"\u5426\u5219\u9700\u8981\u9009\u62e9\u975eavx\u7684paddlepaddl":6,"\u547d\u4ee4":0,"\u547d\u4ee4\u4e3a":6,"\u547d\u4ee4\u5373\u53ef\u5b8c\u6210\u5b89\u88c5":7,"\u547d\u4ee4\u6307\u5b9a\u7684\u53c2\u6570\u4f1a\u4f20\u5165\u7f51\u7edc\u914d\u7f6e\u4e2d":10,"\u547d\u4ee4\u8fd0\u884c\u955c\u50cf":6,"\u547d\u4ee4\u9884\u5148\u4e0b\u8f7d\u955c\u50cf":6,"\u548c":[1,6,20],"\u548c\u4e09\u79cd\u5e8f\u5217\u6a21\u5f0f":20,"\u548c\u5dee\u8bc4":10,"\u548c\u6587\u672c\u4fe1\u606f\u7528tab\u95f4\u9694":10,"\u548c\u6d4b\u8bd5\u6587\u4ef6\u5217\u8868":19,"\u548c\u7528\u6237\u4f20\u5165\u7684\u53c2\u6570":20,"\u548c\u9884\u5904\u7406\u811a\u672c":10,"\u548ccudnn":7,"\u548cinitalizer\u91cc\u5b9a\u4e49\u987a\u5e8f\u4e00\u81f4":10,"\u5668":10,"\u56db\u4e2a\u7248\u672c":7,"\u56db\u79cd\u6570\u636e\u7c7b\u578b\u662f":20,"\u56fe\u50cf\u5206\u7c7b":9,"\u5728":[1,7,20],"\u5728\u58f0\u660edataprovider\u7684\u65f6\u5019\u4f20\u5165\u4e86dictionary\u4f5c\u4e3a\u53c2\u6570":20,"\u5728\u5b8c\u6210\u4e86\u6570\u636e\u548c\u7f51\u7edc\u7ed3\u6784\u642d\u5efa\u4e4b\u540e":10,"\u5728\u672c\u95ee\u9898\u4e2d":10,"\u5728\u6a21\u578b\u914d\u7f6e\u4e2d\u5229\u7528":10,"\u5728\u6b64\u4e3a\u65b9\u4fbf\u5bf9\u6bd4\u4e0d\u540c\u7f51\u7edc\u7ed3\u6784":10,"\u5728\u6bcf\u4e2a\u7ef4\u5ea6\u4e0a\u53d6\u51fa\u5728\u8be5\u53e5\u8bdd\u65b0\u7684\u5411\u91cf\u96c6\u5408\u4e0a\u8be5\u7ef4\u5ea6\u7684\u6700\u5927\u503c\u4f5c\u4e3a\u6700\u540e\u7684\u53e5\u5b50\u8868\u793a\u5411\u91cf":10,"\u5728\u7a0b\u5e8f\u5f00\u59cb\u9636\u6bb5":23,"\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u8fdb\u884c\u6d4b\u8bd5":19,"\u5728\u8bad\u7ec3\u914d\u7f6e\u91cc":20,"\u5728\u8fd9\u4e2a\u51fd\u6570\u4e2d":20,"\u5728\u914d\u7f6e\u4e2d\u8bfb\u53d6\u4e86\u5b57\u5178":20,"\u5728cmake\u914d\u7f6e\u65f6\u53ef\u4ee5\u4f7f\u7528":1,"\u5728pydataprovider\u4e2d":20,"\u5728python\u73af\u5883\u4e0b\u9884\u6d4b\u7ed3\u679c":23,"\u57fa\u672c\u4e0a\u4e0d\u80fd\u6574\u4f53\u4fee\u6b63":20,"\u57fa\u672c\u7684\u5904\u7406\u903b\u8f91\u4e5f\u548cmnist\u903b\u8f91\u4e00\u81f4":20,"\u57fa\u672c\u7684pydataprovider\u4f7f\u7528\u4ecb\u7ecd\u5b8c\u6bd5\u4e86":20,"\u591a\u4e2ainput\u4ee5list\u65b9\u5f0f\u8f93\u5165":10,"\u591a\u6b21\u8fd4\u56de\u53d8\u91cf":20,"\u591a\u7ebf\u7a0b\u4e0b\u8f7d\u8fc7\u7a0b\u4e2d":0,"\u591a\u7ebf\u7a0b\u6570\u636e\u8bfb\u53d6":20,"\u5927":20,"\u597d\u8bc4":10,"\u5982\u679c":[7,20],"\u5982\u679c\u4e0d\u4e86\u89e3":20,"\u5982\u679c\u4e0d\u4f7f\u7528\u5219\u4f1a\u4f7f\u7528\u4e00\u4e2a\u7b80\u5316\u7248\u7684\u547d\u4ee4\u884c\u53c2\u6570\u89e3\u6790":1,"\u5982\u679c\u4e0d\u4f7f\u7528\u5219\u4f1a\u4f7f\u7528\u4e00\u4e2a\u7b80\u5316\u7248\u7684\u65e5\u5fd7\u5b9e\u73b0":1,"\u5982\u679c\u4e0d\u5207\u8bcd":10,"\u5982\u679c\u4e0d\u8bbe\u7f6e\u7684\u8bdd":20,"\u5982\u679c\u4f7f\u7528gpu\u7248\u672c\u7684paddlepaddl":7,"\u5982\u679c\u5728":7,"\u5982\u679c\u5728\u7b2c\u4e00\u6b21cmake\u4e4b\u540e\u60f3\u8981\u91cd\u65b0\u8bbe":1,"\u5982\u679c\u5728\u8bad\u7ec3\u65f6":20,"\u5982\u679c\u5c0f\u4e8e\u8fd9\u4e2a\u4e0b\u8f7d\u901f\u5ea6":0,"\u5982\u679c\u60a8\u4f7f\u7528":6,"\u5982\u679c\u60f3\u8981\u5728\u5916\u90e8\u673a\u5668\u8bbf\u95ee\u8fd9\u4e2acontain":6,"\u5982\u679c\u662ffalse\u7684\u8bdd":20,"\u5982\u679c\u662ftrue\u7684\u8bdd":20,"\u5982\u679c\u6709\u66f4\u590d\u6742\u7684\u4f7f\u7528":19,"\u5982\u679c\u7528\u6237\u4e0d\u6307\u5b9a\u8fd4\u56de\u6570\u636e\u7684\u5bf9\u5e94\u5173\u7cfb":20,"\u5982\u679c\u8bbe\u7f6e\u6210true\u7684\u8bdd":20,"\u5982\u679c\u8f93\u51fa":6,"\u5982\u679c\u8fd0\u884cgpu\u7248\u672c\u7684paddlepaddl":6,"\u5982\u679ctest":19,"\u5b83\u5305\u542b\u7684\u53c2\u6570\u6709":20,"\u5b89\u88c5\u5305\u5728ubuntu":7,"\u5b89\u88c5\u5305\u7684\u4e0b\u8f7d\u5730\u5740\u662f":7,"\u5b89\u88c5\u597d\u7684paddlepaddle\u811a\u672c\u5305\u62ec\u591a\u6761\u547d\u4ee4":13,"\u5b89\u88c5\u5b8c\u6210\u540e":7,"\u5b89\u88c5\u5b8c\u6210\u7684paddlepaddle\u4e3b\u4f53\u5305\u62ec\u4e09\u4e2a\u90e8\u5206":6,"\u5b89\u88c5\u5b8c\u6210paddlepaddle\u540e":7,"\u5b89\u88c5\u6559\u7a0b":10,"\u5b89\u88c5\u65b9\u6cd5\u8bf7\u53c2\u8003":6,"\u5b89\u88c5\u7f16\u8bd1\u4f9d\u8d56":3,"\u5b89\u88c5\u7f16\u8bd1paddlepaddle\u9700\u8981\u7684\u4f9d\u8d56":2,"\u5b89\u88c5docker\u9700\u8981\u60a8\u7684\u673a\u5668":6,"\u5b89\u88c5paddlepaddl":10,"\u5b89\u88c5paddlepaddle\u7684docker\u955c\u50cf":5,"\u5b8c\u6210\u591a\u673a\u8bad\u7ec3":13,"\u5b8c\u6574\u4ee3\u7801\u89c1":23,"\u5b9a\u4e49\u6587\u672c\u4fe1\u606f":10,"\u5b9e\u73b0\u4e86\u6253\u5f00\u6587\u672c\u6587\u4ef6":20,"\u5bc6\u7801\u4e5f\u662froot":6,"\u5bf9\u4e8e\u7528\u6237\u6765\u8bf4":20,"\u5bf9\u4e8e\u7ed9\u5b9a\u7684\u4e00\u6761\u6587\u672c":10,"\u5bf9\u4e8ecuda\u7684toolkit\u6709\u65ad\u70b9\u7eed\u4f20\u548c\u4f20\u8f93\u901f\u5ea6\u8fc7\u5c0f\u91cd\u542f\u4e0b\u8f7d\u7684":0,"\u5bf9\u4e8emnist\u800c\u8a00":20,"\u5bf9\u8be5\u8868\u793a\u8fdb\u884c\u975e\u7ebf\u6027\u53d8\u6362":10,"\u5bf9\u8c61":20,"\u5bf9\u8c61convert":23,"\u5c06\u4f1a\u6d4b\u8bd5\u914d\u7f6e\u6587\u4ef6\u4e2dtest":10,"\u5c06\u5b57\u5178\u5b58\u5165\u4e86set":20,"\u5c06\u5bbf\u4e3b\u673a\u76848022\u7aef\u53e3\u6620\u5c04\u5230container\u768422\u7aef\u53e3\u4e0a":6,"\u5c06\u6570\u636e\u7ec4\u5408\u6210batch\u8bad\u7ec3":20,"\u5c06\u6587\u4ef6\u7684\u7edd\u5bf9\u8def\u5f84\u6216\u76f8\u5bf9\u8def\u5f84":19,"\u5c06\u8bc4\u8bba\u5206\u4e3a\u597d\u8bc4":10,"\u5c06\u8be5\u53e5\u8bdd\u5305\u542b\u7684\u6240\u6709\u5355\u8bcd\u5411\u91cf\u6c42\u5e73\u5747\u5f97\u5230\u53e5\u5b50\u7684\u8868\u793a":10,"\u5c06ssh\u88c5\u5165\u7cfb\u7edf\u5185\u5e76\u5f00\u542f\u8fdc\u7a0b\u8bbf\u95ee":6,"\u5c31":20,"\u5c31\u50cf\u8fd9\u4e2a\u6837\u4f8b\u4e00\u6837":20,"\u5c31\u662f\u5c06\u8fd9\u4e9b\u52a8\u6001\u5e93\u52a0\u5230\u73af\u5883\u53d8\u91cf\u91cc\u9762":7,"\u5c5e\u6027":20,"\u5dee\u8bc4":10,"\u5e38\u89c1\u7684\u8f93\u51fa\u683c\u5f0f\u4e3a":18,"\u5e76\u4e14":20,"\u5e76\u4e14\u4f7f\u7528\u5173\u952e\u8bcd":20,"\u5e76\u4e14\u5220\u9664container\u4e2d\u7684\u6570\u636e":6,"\u5e76\u4e14\u5728\u5185\u5b58\u8db3\u591f":20,"\u5e76\u4e14\u6807\u8bb0process\u51fd\u6570\u662f\u4e00\u4e2adataprovid":20,"\u5e76\u4f7f\u7528\u4e86dropout":10,"\u5e76\u572823\u884c\u8fd4\u56de\u7ed9paddlepaddle\u8fdb\u7a0b":20,"\u5e76\u5c06\u6bcf\u884c\u8f6c\u6362\u6210\u548c":20,"\u5e76\u63d0\u4f9b":6,"\u5e76\u63d0\u4f9b\u4e86\u7b80\u5355\u7684cache\u529f\u80fd":20,"\u5e76\u8bbe\u7f6e\u597d\u5bf9\u5e94\u7684\u73af\u5883\u53d8\u91cf":7,"\u5e76\u9010\u6e10\u5c55\u793a\u66f4\u52a0\u6df1\u5165\u7684\u529f\u80fd":10,"\u5e8f\u5217\u6a21\u578b\u6570\u636e\u63d0\u4f9b":19,"\u5e8f\u5217\u6a21\u578b\u662f\u6307\u6570\u636e\u7684\u67d0\u4e00\u7ef4\u5ea6\u662f\u4e00\u4e2a\u5e8f\u5217\u5f62\u5f0f":20,"\u5e93\u7684\u8bdd":7,"\u5f0f":20,"\u5f15\u7528\u7684dataprovider\u662f":20,"\u5f53\u51fd\u6570\u8fd4\u56de\u7684\u65f6\u5019":20,"\u5f53\u524dlog_period\u4e2abatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747\u5206\u7c7b\u9519\u8bef\u7387":10,"\u5f53\u524dlog_period\u4e2abatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747cost":10,"\u5f53\u7136":19,"\u5f53\u8c03":20,"\u5f88":10,"\u5f97\u5230\u7ed3\u679c":7,"\u5feb\u901f\u5165\u95e8":11,"\u5ff5\u662f":20,"\u60a8\u53ef\u4ee5\u4f7f\u7528":6,"\u60a8\u5c31\u53ef\u4ee5\u8fdc\u7a0b\u7684\u4f7f\u7528paddlepaddle\u5566":6,"\u60a8\u9700\u8981\u5728\u673a\u5668\u4e2d\u5b89\u88c5\u597ddocker":6,"\u60a8\u9700\u8981\u8fdb\u5165\u955c\u50cf\u8fd0\u884cpaddlepaddl":6,"\u60c5\u611f\u5206\u6790":9,"\u60f3\u8981\u8fd0\u884cpaddlepaddl":6,"\u6210\u4e3a\u7ef4\u5ea6\u4e3ahidden":10,"\u6211\u4eec\u4ece\u63d0\u524d\u7ed9\u5b9a\u7684\u7c7b\u522b\u96c6\u5408\u4e2d\u9009\u62e9\u5176\u6240\u5c5e\u7c7b":10,"\u6211\u4eec\u4ee5\u6587\u672c\u5206\u7c7b\u95ee\u9898\u4f5c\u4e3a\u80cc\u666f":10,"\u6211\u4eec\u4f7f\u7528":10,"\u6211\u4eec\u5728\u6b64\u603b":10,"\u6211\u4eec\u5c06\u4ee5\u57fa\u672c\u7684\u903b\u8f91\u56de\u5f52\u7f51\u7edc\u4f5c\u4e3a\u8d77\u70b9":10,"\u6211\u4eec\u5c06\u5728\u540e\u9762\u4ecb\u7ecd\u8bad\u7ec3\u548c\u9884\u6d4b\u7684\u6d41\u7a0b\u7684\u811a\u672c":10,"\u6211\u4eec\u5c06\u8bad\u7ec3\u7684\u547d\u4ee4\u884c\u4fdd\u5b58\u5728\u4e86":10,"\u6211\u4eec\u63a8\u8350\u4f7f\u7528docker\u955c\u50cf\u6765\u90e8\u7f72\u73af\u5883":5,"\u6211\u4eec\u63d0\u4f9b\u4e8612\u4e2a":6,"\u6211\u4eec\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5de5\u5177\u7c7bdataproviderconvert":23,"\u6211\u4eec\u8fdb\u5165\u5230\u8bad\u7ec3\u90e8\u5206":10,"\u6211\u4eec\u91c7\u7528\u5355\u5c42lstm\u6a21\u578b":10,"\u6211\u4eec\u968f\u65f6\u603b\u7ed3\u4e86\u5404\u4e2a\u7f51\u7edc\u7684\u590d\u6742\u5ea6\u548c\u6548\u679c":10,"\u6216\u5176\u4ed6":10,"\u6216\u8005":6,"\u6216\u800510g\u8fd9\u6837\u7684\u5355\u4f4d":0,"\u6216\u8005\u4f7f\u7528\u4e0b\u9762\u4e00\u6761\u547d\u4ee4\u5b89\u88c5":7,"\u6216\u8005\u5728python":20,"\u6216\u8005\u6570\u636e\u5e93\u8fde\u63a5\u5730\u5740\u7b49\u7b49":19,"\u6216\u8005\u8bbe\u7f6e\u4e3anone":19,"\u6216\u8005\u9700\u8981\u66f4\u9ad8\u7684\u6548\u7387":19,"\u6216\u8005\u9ad8\u6027\u80fd\u7684":6,"\u6240\u4ee5":[20,23],"\u6240\u4ee5\u5728cpu\u7684\u8fd0\u7b97\u6027\u80fd\u4e0a\u5e76\u4e0d\u4f1a\u6709\u4e25\u91cd\u7684\u5f71\u54cd":6,"\u6240\u4ee5\u5982\u679c\u5bf9\u4e8e\u5185\u5b58\u6bd4\u8f83\u5c0f\u7684\u673a\u5668":20,"\u6240\u4ee5\u5982\u679c\u60f3\u8981\u5728\u540e\u53f0\u542f\u7528ssh":6,"\u6240\u4ee5\u5c06":20,"\u6240\u4ee5\u63a8\u8350\u4f7f\u7528\u663e\u5f0f\u6307\u5b9a\u8fd4\u56de\u503c\u548c\u6570\u636e\u5bf9\u5e94\u5173\u7cfb":20,"\u6240\u4ee5\u6700\u4f73\u5b9e\u8df5\u63a8\u8350\u4e0d\u8981\u5c06\u6bcf\u4e00\u4e2a\u6837\u672c\u90fd\u653e\u5165train":20,"\u6240\u4ee5\u7528\u4e8e\u9884\u6d4b\u7684\u914d\u7f6e\u6587\u4ef6\u8981\u505a\u76f8\u5e94\u7684\u4fee\u6539":23,"\u6240\u4ee5\u8f93\u51fa\u7684value\u5305\u542b\u4e24\u4e2a\u5411\u91cf":23,"\u6240\u4ee5gpu\u5728\u8fd0\u7b97\u6027\u80fd\u4e0a\u4e5f\u4e0d\u4f1a\u6709\u4e25\u91cd\u7684\u5f71\u54cd":6,"\u6240\u4ee5init_hook\u5c3d\u91cf\u4f7f\u7528":20,"\u6240\u6709\u5b57\u7b26\u90fd\u5c06\u8f6c\u6362\u4e3a\u8fde\u7eed\u6574\u6570\u8868\u793a\u7684id\u4f20\u7ed9\u6a21\u578b":10,"\u6240\u6709\u6587\u4ef6\u5217\u8868":20,"\u6240\u6709\u7684":1,"\u6240\u6709\u7684\u4e0b\u8f7d\u7ebf\u7a0b\u5173\u95ed\u65f6":0,"\u6240\u6709\u914d\u7f6e\u5728":10,"\u6240\u8c13\u65f6\u95f4\u6b65\u4fe1\u606f":20,"\u624d\u4f1a\u91ca\u653e\u8be5\u6bb5\u5185\u5b58":20,"\u624d\u4f1astop":20,"\u6253\u5370\u7684\u65e5\u5fd7\u53d8\u591a":1,"\u6267\u884c":0,"\u6267\u884c\u5982\u4e0b\u547d\u4ee4\u5373\u53ef\u4ee5\u5173\u95ed\u8fd9\u4e2acontain":6,"\u6267\u884c\u65b9\u6cd5\u5982\u4e0b":6,"\u62a5\u9519":7,"\u62fc\u63a5\u6210\u4e00\u4e2a\u65b0\u7684\u5411\u91cf\u8868\u793a":10,"\u6307\u4ee4\u96c6":6,"\u6307\u5b9a\u521d\u59cb\u5316\u6a21\u578b\u8def\u5f84":10,"\u6307\u5b9a\u751f\u6210\u6570\u636e\u7684\u51fd\u6570":10,"\u6307\u5b9a\u8bad\u7ec3":10,"\u6307\u5b9abatch":10,"\u6307\u5b9aoutputs\u8f93\u51fa\u6982\u7387\u5c42":10,"\u6389\u7f16\u8bd1\u76ee\u5f55\u540e":1,"\u63a5\u4e0b\u6765\u4f7f\u7528":23,"\u63a5\u53e3\u4f7f\u7528\u591a\u7ebf\u7a0b\u8bfb\u53d6\u6570\u636e":20,"\u63a8\u8350\u4f7f\u7528\u5c06\u672c\u5730\u7f51\u5361":6,"\u63a8\u8350\u4f7f\u7528\u6700\u65b0\u7248\u672c\u7684cudnn":1,"\u63a8\u8350\u6e05\u7406":1,"\u63a8\u8350\u76f4\u63a5\u653e\u7f6e\u5230\u8bad\u7ec3\u76ee\u5f55":19,"\u63a8\u8350\u8bbe\u7f6e":20,"\u63cf\u8ff0":1,"\u63cf\u8ff0\u4e86docker":0,"\u6548\u679c\u4e00\u81f4":20,"\u6548\u679c\u603b\u7ed3":10,"\u6559\u7a0b":10,"\u6570\u636e":20,"\u6570\u636e\u4f20\u8f93\u65e0\u9700label\u6570\u636e":10,"\u6570\u636e\u5904\u7406python\u6587\u4ef6\u540d":10,"\u6570\u636e\u5982\u4f55\u5b58\u50a8\u7b49\u7b49":20,"\u6570\u636e\u63d0\u4f9b":19,"\u6570\u636e\u6587\u4ef6\u5b58\u653e\u5728\u672c\u5730\u78c1\u76d8\u4e2d":19,"\u6570\u636e\u662f\u7ed9\u4e00\u6bb5\u82f1\u6587\u6587\u672c":20,"\u6570\u636e\u683c\u5f0f\u548c\u8be6\u7ec6\u6587\u6863\u8bf7\u53c2\u8003":10,"\u6587\u4ef6":20,"\u6587\u4ef6\u4e2d":10,"\u6587\u672c\u4e2d\u7684\u5355\u8bcd\u7528\u7a7a\u683c\u5206\u9694":10,"\u6587\u672c\u4fe1\u606f\u5c31\u662f\u4e00\u4e2a\u5e8f\u5217":20,"\u6587\u672c\u5206\u7c7b\u95ee\u9898":10,"\u6587\u672c\u5377\u79ef\u5206\u4e3a\u4e09\u4e2a\u6b65\u9aa4":10,"\u6587\u672c\u751f\u6210":9,"\u65b0\u5199layer":11,"\u65b9\u4fbf\u8c03\u8bd5\u4f7f\u7528":13,"\u65b9\u4fbf\u90e8\u7f72\u5206\u53d1":13,"\u65e0\u9700label\u76f8\u5173\u7684\u5c42":10,"\u65f6\u5e8f\u6a21\u578b\u5373\u4e3arnn\u6a21\u578b":10,"\u65f6\u5e8f\u6a21\u578b\u5747\u4f7f\u7528\u8be5\u811a\u672c":10,"\u662f":1,"\u662f\u4e00\u4e2abatch":20,"\u662f\u4e00\u4e2apython\u7684":20,"\u662f\u4e00\u4e2aswig\u5c01\u88c5\u7684paddlepaddle\u5305":6,"\u662f\u4e0d\u662f\u5f88\u7b80\u5355\u5462":20,"\u662f\u4e2adataprovider\u662f\u4e0d\u662f\u8981\u505ashuffl":20,"\u662f\u4ec0\u4e48\u4e5f\u6ca1\u5173\u7cfb":20,"\u662f\u4ece\u8bad\u7ec3\u914d\u7f6e\u4f20\u5165\u7684dict\u5bf9\u8c61":20,"\u662f\u51e0\u4e4e\u4e0d\u5360\u5185\u5b58\u7684":20,"\u662f\u521d\u59cb\u5316\u65f6\u8c03\u7528\u7684\u51fd\u6570":20,"\u662f\u540c\u4e00\u4e2a\u5bf9\u8c61":20,"\u662f\u5426\u4f7f\u7528\u53cc\u7cbe\u5ea6\u6d6e\u70b9\u6570":1,"\u662f\u5426\u4f7f\u7528\u8fd0\u884c\u65f6\u52a8\u6001\u52a0\u8f7dcuda\u52a8\u6001\u5e93":1,"\u662f\u5426\u4f7f\u7528gflags":1,"\u662f\u5426\u4f7f\u7528glog":1,"\u662f\u5426\u5185\u5d4cpython\u89e3\u91ca\u5668":1,"\u662f\u5426\u5bfb\u627e\u5230cuda\u5de5\u5177\u94fe":1,"\u662f\u5426\u5f00\u542f\u5355\u5143\u6d4b\u8bd5":1,"\u662f\u5426\u5f00\u542f\u8ba1\u65f6\u529f\u80fd\u5f00\u542f\u8ba1\u65f6\u529f\u80fd\u4f1a\u5bfc\u81f4\u8fd0\u884c\u7565\u6162":1,"\u662f\u5426\u5f00\u542fgpu\u529f\u80fd":0,"\u662f\u5426\u5f00\u542frdma\u652f\u6301":1,"\u662f\u5426\u7f16\u8bd1\u4e2d\u6587\u6587\u6863":1,"\u662f\u5426\u7f16\u8bd1\u542b\u6709avx\u6307\u4ee4\u96c6\u7684paddlepaddle\u4e8c\u8fdb\u5236":1,"\u662f\u5426\u7f16\u8bd1\u65f6\u8fdb\u884c\u4ee3\u7801\u98ce\u683c\u68c0\u67e5":1,"\u662f\u5426\u7f16\u8bd1\u82f1\u6587\u6587\u6863":1,"\u662f\u5426\u7f16\u8bd1gpu\u652f\u6301":1,"\u662f\u5426\u7f16\u8bd1python\u7684swig\u63a5\u53e3":1,"\u662f\u5728\u8fd0\u884c\u65f6\u6267\u884c\u7684":20,"\u662f\u6570\u636e\u7f13\u5b58\u7684\u7b56\u7565":20,"\u662f\u6570\u636e\u8f93\u5165\u683c\u5f0f":20,"\u662f\u8bbe\u7f6e\u8fd9\u4e2adataprovider\u8fd4\u56de\u4ec0\u4e48\u6837\u7684\u6570\u636e":20,"\u662f\u8bbe\u7f6edataprovider\u5728\u5185\u5b58\u4e2d\u6682\u5b58\u7684\u6570\u636e\u6761\u6570":20,"\u662f\u8bbe\u7f6edataprovider\u5728\u5185\u5b58\u4e2d\u6700\u5c0f\u6682\u5b58\u7684\u6570\u636e\u6761\u6570":20,"\u662fpaddlepaddle\u8d1f\u8d23\u63d0\u4f9b\u6570\u636e\u7684\u6a21\u5757":19,"\u662fpython\u7684\u4e00\u4e2a\u5173\u952e\u8bcd":20,"\u663e":10,"\u666e\u901a\u7528\u6237\u8bf7\u8d70\u5b89\u88c5\u6d41\u7a0b":5,"\u66f4\u8be6\u7ec6\u7528\u4f8b\u8bf7\u53c2\u8003\u6587\u6863":10,"\u66f4\u8be6\u7ec6\u7684\u4ecb\u7ecd\u8bf7\u53c2\u8003\u5404\u4e2a\u547d\u4ee4\u7684\u547d\u4ee4\u884c\u53c2\u6570\u6587\u6863":13,"\u66f4\u8be6\u7ec6\u7684\u7f51\u7edc\u914d\u7f6e":10,"\u6700\u4f4e\u7ebf\u7a0b\u7684\u4e0b\u8f7d\u901f\u5ea6":0,"\u6700\u540e\u4f7f\u7528":23,"\u6709100\u4e2a\u8bad\u7ec3\u6587\u4ef6":20,"\u6709\u503c\u7684\u4f4d\u7f6e\u53ea\u80fd\u53d61":20,"\u6709\u503c\u7684\u90e8\u5206\u53ef\u4ee5\u662f\u4efb\u4f55\u6d6e\u70b9\u6570":20,"\u6709\u90e8\u5206\u53c2\u6570\u662fpaddle\u81ea\u52a8\u751f\u6210\u7684":20,"\u672c\u8282\u6211\u4eec\u5c06\u4e13\u6ce8\u4e8e\u7f51\u7edc\u7ed3\u6784\u7684\u4ecb\u7ecd":10,"\u6765\u5b89\u88c5":7,"\u6765\u5f15\u7528\u8fd9\u4e2aimag":6,"\u6765\u63a5\u53d7\u4e0d\u4f7f\u7528\u7684":20,"\u6765\u786e\u5b9a\u5bf9\u5e94\u5173\u7cfb":20,"\u6765\u81ea\u5b9a\u4e49\u4f20\u6570\u636e\u7684\u8fc7\u7a0b":19,"\u6765\u8bf4\u660e\u7b80\u5355\u7684pydataprovider\u5982\u4f55\u4f7f\u7528":20,"\u6765\u8fdb\u884c\u8bad\u7ec3":6,"\u6765\u914d\u7f6ecudnn\u7684\u5b89\u88c5\u8def\u5f84":1,"\u6784\u9020gradientmachin":23,"\u6790\u597d\u7684\u914d\u7f6e\u521b\u5efa\u795e\u7ecf\u7f51\u7edc":23,"\u67e5\u770b\u5b89\u88c5\u540e\u7684paddl":7,"\u6807\u7b7e\u662f0":20,"\u6837\u4f8b\u6570\u636e\u4e3a":20,"\u6837\u4f8b\u6570\u636e\u5982\u4e0b":20,"\u6837\u672c":20,"\u6839\u636e\u4e0a\u4e00\u6b65\u89e3":23,"\u6839\u636e\u6a21\u578b\u914d\u7f6e\u6587\u4ef6\u4e2d":20,"\u683c\u5f0f\u5982\u4e0b":10,"\u68d2":10,"\u6a21\u578b\u5b58\u50a8\u8def\u5f84":10,"\u6a21\u578b\u8bad\u7ec3\u4f1a\u770b\u5230\u8fd9\u6837\u7684\u65e5\u5fd7":10,"\u6a21\u578b\u914d\u7f6e":11,"\u6a2a\u5411\u5305\u62ec\u4e09\u4e2a\u7248\u672c":6,"\u6b63\u5e38\u7684docker":6,"\u6b63\u6837\u672c":10,"\u6bcf\u4e00\u4e2a\u4efb\u52a1\u6d41\u7a0b\u90fd\u53ef\u4ee5\u5206\u4e3a\u5982\u4e0b5\u4e2a\u57fa\u7840\u90e8\u5206":10,"\u6bcf\u4e00\u6761\u8bad\u7ec3\u6570\u636e\u90fd\u662f\u4e00\u4e2a\u6587\u4ef6":20,"\u6bcf\u4e00\u884c":20,"\u6bcf\u4e2a\u5c42\u90fd\u6709\u4e00\u4e2a\u6216\u591a\u4e2ainput":10,"\u6bcf\u4e2agenerator\u5728\u6ca1\u6709\u8c03\u7528\u7684\u65f6\u5019":20,"\u6bcf\u4e2apass\u7684\u7b2c0\u4e2abatch\u5230\u5f53\u524dbatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747\u5206\u7c7b\u9519\u8bef\u7387":10,"\u6bcf\u4e2apass\u7684\u7b2c0\u4e2abatch\u5230\u5f53\u524dbatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747cost":10,"\u6bcf\u6b21\u90fd\u4f1a\u4ecepython\u7aef\u8bfb\u53d6\u6570\u636e":20,"\u6bcf\u884c\u4fdd\u5b58\u4e00\u6761\u6837\u672c":10,"\u6bcf\u9694\u591a\u5c11batch\u6253\u5370\u4e00\u6b21\u65e5\u5fd7":10,"\u6bd4\u5982\u901a\u8fc7\u7528\u6237\u5bf9\u7535\u5b50\u5546\u52a1\u7f51\u7ad9\u8bc4\u8bba":10,"\u6bd4\u8f83\u53ef\u80fd\u7684\u547d\u4ee4\u5982\u4e0b":7,"\u6ca1\u6709\u4f5c\u7528":20,"\u6ca1\u6709\u5b89\u88c5":7,"\u6ca1\u6709\u8bbe\u7f6e":7,"\u6ce8\u610f":[0,20],"\u6d41\u7a0b\u5982\u4e0b":10,"\u6d4b\u8bd5\u6570\u636e":10,"\u6d4b\u8bd5\u7684\u65f6\u5019\u9ed8\u8ba4\u4e0dshuffl":20,"\u6d4b\u8bd5\u811a\u672c\u5982\u4e0b":10,"\u6e90\u7801":10,"\u6e90\u7801\u6839\u76ee\u5f55":0,"\u6fc0\u6d3b\u51fd\u6570\u7c7b\u578b":10,"\u7136\u540e\u6267\u884c\u5982\u4e0b":7,"\u7136\u540e\u8fd0\u884c\u8fd9\u4e2acontainer\u5373\u53ef":6,"\u7248\u672c":7,"\u751f\u6210\u5404\u4e2a\u5e73\u53f0\u7684makefil":1,"\u75280\u548c1\u8868\u793a":20,"\u7528\u4e86\u4e24\u4e2a\u6708\u4e4b\u540e\u8fd9\u4e2a\u663e\u793a\u5668\u5c4f\u5e55\u788e\u4e86":10,"\u7528\u4e8e\u4e0d\u652f\u6301avx\u6307\u4ee4\u96c6\u7684cpu":7,"\u7528\u6237\u4e5f\u53ef\u4ee5\u5728c":19,"\u7528\u6237\u4e5f\u53ef\u4ee5\u663e\u5f0f\u6307\u5b9a\u8fd4\u56de\u7684\u6570\u636e\u5bf9\u5e94\u5173\u7cfb":20,"\u7528\u6237\u53ef\u4ee5\u4f7f\u7528python\u7684":19,"\u7528\u6237\u53ef\u4ee5\u6839\u636e\u8bad\u7ec3log\u9009\u62e9test\u7ed3\u679c\u6700\u597d\u7684\u6a21\u578b\u6765\u9884\u6d4b":10,"\u7528\u6237\u53ef\u4ee5\u9009\u62e9\u5bf9\u5e94\u7248\u672c\u7684docker":6,"\u7528\u6237\u540d\u4e3a":6,"\u7528\u6237\u5728dataprovider\u4e2d\u9700\u8981\u5b9e\u73b0\u5982\u4f55\u8bbf\u95ee\u5176\u4e2d\u6bcf\u4e00\u4e2a\u6587\u4ef6":19,"\u7528\u6237\u5b9a\u4e49\u7684\u53c2\u6570\u4f7f\u7528args\u5728\u8bad\u7ec3\u914d\u7f6e\u4e2d\u8bbe\u7f6e":20,"\u7528\u6237\u63a5\u53e3":11,"\u7528\u6237\u9700\u8981\u5148\u5c06paddlepaddle\u5b89\u88c5\u5305\u4e0b\u8f7d\u5230\u672c\u5730":7,"\u7528\u6765\u505a\u9884\u6d4b\u548c\u7b80\u5355\u7684\u5b9a\u5236\u5316":6,"\u7528\u8fc7\u4e00\u6b21\u7684\u65f6\u5019":20,"\u7531\u4e8e\u6570\u636e\u662f\u4e24\u6761":23,"\u7531\u4e8edocker\u662f\u57fa\u4e8e\u5bb9\u5668\u7684\u8f7b\u91cf\u5316\u865a\u62df\u65b9\u6848":6,"\u7531\u4e8epaddlepaddle\u7684docker\u955c\u50cf\u5e76\u4e0d\u5305\u542b\u4efb\u4f55\u9884\u5b9a\u4e49\u7684\u8fd0\u884c\u547d\u4ee4":6,"\u7531\u6613\u5230\u96be\u5c55\u793a4\u79cd\u4e0d\u540c\u7684\u7f51\u7edc\u914d\u7f6e":10,"\u7684":10,"\u7684\u540d\u5b57":20,"\u7684\u5b89\u88c5\u6587\u6863":6,"\u7684\u60c5\u51b5\u4e0b\u8d8a\u5927\u8d8a\u597d":20,"\u7684\u6587\u6863":20,"\u7684\u65f6\u5019\u5982\u679c\u62a5\u4e00\u4e9b\u4f9d\u8d56\u672a\u627e\u5230\u7684\u9519\u8bef\u662f\u6b63\u5e38\u7684":7,"\u7684\u662f":20,"\u7684\u673a\u5668\u4e0a\u8fdb\u884c":0,"\u7684\u7f51\u6865\u6765\u8fdb\u884c\u7f51\u7edc\u901a\u4fe1":6,"\u7684demo\u5b66\u4e60\u5982\u4f55\u8fdb\u884c\u591a\u673a\u8bad\u7ec3":10,"\u7684docker\u53ef\u80fd\u7f3a\u4e4f":0,"\u7684matrix":23,"\u7684python\u5305\u662fpaddlepaddle\u7684\u8bad\u7ec3\u4e3b\u8981\u7a0b\u5e8f":6,"\u7684python\u5305\u6765\u505a\u914d\u7f6e\u6587\u4ef6\u89e3\u6790\u7b49\u5de5\u4f5c":6,"\u7684python\u9884\u6d4b\u8fc7\u7a0b":10,"\u76ee\u5f55":10,"\u76ee\u5f55\u4e0b":0,"\u76f4\u63a5\u63d0\u53d6\u51fa\u795e\u7ecf\u7f51\u7edcoutput\u5c42\u7684\u8f93\u51fa\u7ed3\u679c":23,"\u76f8\u5173\u547d\u4ee4\u4e3a":6,"\u76f8\u5173\u7684\u6982":20,"\u76f8\u5bf9\u4e8epaddlepaddle\u7a0b\u5e8f\u8fd0\u884c\u65f6\u7684\u8def\u5f84":19,"\u77e5\u9053\u5982\u4f55\u4ece":20,"\u793a":10,"\u7a0b\u5e8f\u6216\u8005\u81ea\u5b9a\u4e49\u4e00\u4e2a\u542b\u6709\u542f\u52a8\u811a\u672c\u7684imag":6,"\u7aef\u81ea\u5b9a\u4e49\u4e00\u4e2a":19,"\u7b2c\u4e00\u4e2a\u53c2\u6570\u662f":20,"\u7b2c\u4e00\u4e2apass\u4f1a\u4ecepython\u7aef\u8bfb\u53d6\u6570\u636e":20,"\u7b2c\u4e00\u6bb5\u6570\u636e\u4e3a\u8fd9\u5f20\u56fe\u7247\u7684label":20,"\u7b2c\u4e8c\u4e2a\u53c2\u6570\u662ffilenam":20,"\u7b2c\u4e8c\u6bb5\u6570\u636e\u4e3a\u8fd9\u4e2a\u56fe\u7247\u7684\u50cf\u7d20\u503c":20,"\u7b80\u5355\u4f18\u5316":0,"\u7b80\u5355\u7684\u4f7f\u7528":19,"\u7b80\u5355\u7684\u4f7f\u7528\u573a\u666f":19,"\u7b80\u5355\u7684\u4f7f\u7528\u6837\u4f8b\u4e3a":0,"\u7b80\u5355\u7684\u542b\u6709ssh\u7684dockerfile\u5982\u4e0b":6,"\u7b80\u5355\u7684pydataprovider\u6837\u4f8b\u5c31\u8bf4\u660e\u5b8c\u6bd5\u4e86":20,"\u7c7b\u522bid":10,"\u7c7b\u522bid\u7684\u6570\u636e\u7c7b\u578b":10,"\u7c7b\u578b\u6765\u8bbe\u7f6e":20,"\u7eb5\u5411\u5305\u62ec\u56db\u4e2a\u7248\u672c":6,"\u7ec3":13,"\u7ed3\u4e0a\u8ff0\u7f51\u7edc\u7ed3\u6784\u5728amazon":10,"\u7ee7\u7eed\u8bad\u7ec3":20,"\u7ef4\u5ea6\u4e3aword":10,"\u7ef4\u5ea6\u662f\u7c7b\u522b\u4e2a\u6570":10,"\u7ef4\u5ea6\u662f\u8bcd\u5178\u5927\u5c0f":10,"\u7f13\u5b58\u8bad\u7ec3\u6570\u636e\u5230\u5185\u5b58":20,"\u7f16\u8bd1\u53c2\u6570\u9009\u9879\u6587\u4ef6":18,"\u7f16\u8bd1\u73af\u5883\u548c\u6e90\u4ee3\u7801":6,"\u7f16\u8bd1\u9009\u9879":1,"\u7f16\u8bd1\u9009\u9879\u4e3b\u8981\u63a8\u8350\u9ad8\u7ea7\u7528\u6237\u67e5\u770b":5,"\u7f16\u8bd1\u9009\u9879\u5217\u8868\u5982\u4e0b":1,"\u7f16\u8bd1paddlepaddle\u7684gpu\u7248\u672c\u5e76\u4e0d\u9700\u8981\u4e00\u5b9a\u5728\u5177\u6709gpu":0,"\u7f51\u7edc\u540d\u79f0":10,"\u7f51\u7edc\u914d\u7f6e":10,"\u7f6e\u8fd9\u4e9b\u53d8\u91cf":1,"\u800c":6,"\u800c\u4e09\u79cd\u5e8f\u5217\u6a21\u5f0f\u4e3a":20,"\u800c\u4e0d\u4f7f\u7528docker":6,"\u800c\u4e0d\u7528\u5173\u5fc3\u6570\u636e\u5982\u4f55\u4f20\u8f93\u7ed9paddlepaddl":20,"\u800c\u4e14\u9884\u6d4b\u7f51\u7edc\u901a\u5e38\u76f4\u63a5\u8f93\u51fa\u6700\u540e\u4e00\u5c42\u7684\u7ed3\u679c\u800c\u4e0d\u662f\u50cf\u8bad\u7ec3\u65f6\u4e00\u6837\u4ee5cost":23,"\u800c\u5728":[1,20],"\u800c\u5982\u679c\u6309\u987a\u5e8f\u8c03\u7528\u8fd9\u4e9bgenerator\u5c31\u4e0d\u4f1a\u51fa\u73b0\u8fd9\u4e2a\u95ee\u9898":20,"\u800c\u662f\u5c06\u6837\u672c\u7684\u5730\u5740\u653e\u5165\u53e6\u4e00\u4e2a\u6587\u672c":20,"\u800c\u6ca1\u6709\u6d4b\u8bd5\u6570\u636e":20,"\u800c\u7279\u5f81\u5373\u4e3a":20,"\u800c\u8fd9\u4e2a\u4e00\u822c\u8bf4\u660epaddlepaddle\u5df2\u7ecf\u5b89\u88c5\u5b8c\u6bd5\u4e86":7,"\u800c\u8fd9\u4e2a\u53d8\u91cf\u63a8\u8350\u5927\u4e8e\u8bad\u7ec3\u7684batch":20,"\u800c\u8fd9\u4e2acontext\u53ef\u80fd\u4f1a\u975e\u5e38":20,"\u800c\u975e\u9759\u6001\u52a0\u8f7dcuda\u52a8\u6001\u5e93":1,"\u800cgpu\u7684\u9a71\u52a8\u548c\u8bbe\u5907\u5168\u90e8\u6620\u5c04\u5230\u4e86\u5bb9\u5668\u5185":6,"\u800cpaddlepaddle\u8fdb\u7a0b\u5e2e\u52a9\u7528\u6237\u505a\u4e86":20,"\u811a\u672c":6,"\u811a\u672c\u53ef\u4ee5\u542f\u52a8paddlepaddle\u7684\u8bad\u7ec3\u8fdb\u7a0b\u548cpserv":6,"\u811a\u672c\u548c":6,"\u811a\u672c\u7c7b\u4f3c\u4e8e":6,"\u81ea\u5b9a\u4e49\u4e00\u4e2adataprovid":19,"\u81f3\u5c11\u5177\u67093":6,"\u81f3\u6b64":[6,20],"\u83b7\u53d6\u5229\u7528one":10,"\u83b7\u53d6\u6bcf\u4e2a\u5355\u8bcd\u5de6\u53f3\u5404k\u4e2a\u8fd1\u90bb":10,"\u83b7\u53d6\u8be5\u6761\u6837\u672c\u7c7b\u522bid":10,"\u8868\u793a\u6574\u6570\u6807\u7b7e":20,"\u8868\u793a\u662f\u5426\u5141\u8bb8paddle\u6682\u5b58\u7565\u5fae\u591a\u4f59pool_size\u7684\u6570\u636e":20,"\u8868\u793a\u7a00\u758f\u7684\u5411\u91cf":20,"\u8868\u793a\u7a00\u758f\u7684\u96f6\u4e00\u5411\u91cf":20,"\u8868\u793a\u7a20\u5bc6\u7684\u6d6e\u70b9\u6570\u5411\u91cf":20,"\u8868\u793a\u8fc7\u4e8620\u4e2abatch":10,"\u8868\u793a\u8fc7\u4e862560\u4e2a\u6837\u672c":10,"\u8868\u793a\u8fd9\u4e2adataprovider\u662f\u8bad\u7ec3\u7528\u7684dataprovider\u6216\u8005\u6d4b\u8bd5\u7528\u7684":20,"\u89e3\u51b3\u529e\u6cd5\u662f\u5c06cuda":7,"\u89e3\u51b3\u65b9\u6cd5\u5f88\u7b80\u5355":7,"\u89e3\u6790\u8bad\u7ec3\u65f6\u7684\u914d\u7f6e\u6587\u4ef6":23,"\u89e3\u91ca":10,"\u8ba9\u795e\u7ecf\u7f51\u7edc\u53ef\u4ee5\u8fdb\u884c\u8bad\u7ec3":19,"\u8bad\u7ec3":6,"\u8bad\u7ec3\u6570\u636e\u975e\u5e38\u591a\u7684\u60c5\u51b5\u4e0b":20,"\u8bad\u7ec3\u6587\u4ef6\u5217\u8868":19,"\u8bad\u7ec3\u65f6\u6240\u9700\u8bbe\u7f6e\u7684\u4e3b\u8981\u53c2\u6570\u5982\u4e0b":10,"\u8bad\u7ec3\u7684\u65f6\u5019\u9ed8\u8ba4shuffl":20,"\u8bad\u7ec3\u811a\u672c":10,"\u8bad\u7ec3\u811a\u672c\u5728":10,"\u8bad\u7ec3\u8f6e\u6b21":10,"\u8bb2\u6570\u636e\u53d1\u9001\u7ed9paddlepaddl":20,"\u8bbe\u7f6e\u4e0b\u5217\u7f16\u8bd1\u9009\u9879\u65f6":1,"\u8bbe\u7f6e\u6210":20,"\u8bbe\u7f6e\u6210\u4e86\u5e8f\u5217":20,"\u8bbe\u7f6e\u6210true\u7684\u8bdd":20,"\u8bbe\u7f6e\u8f93\u5165\u7c7b\u578b":20,"\u8bc4\u4f30\u4ea7\u54c1\u7684\u8d28\u91cf":10,"\u8bcd\u6027\u6807\u6ce8":9,"\u8be5\u5c42\u795e\u7ecf\u5143\u4e2a\u6570":10,"\u8be5\u6570\u636e":20,"\u8be5\u6784\u5efa\u811a\u672c\u5145\u5206\u8003\u8651\u4e86\u7f51\u7edc\u4e0d\u7a33\u5b9a\u7684\u60c5\u51b5":0,"\u8be5\u6a21\u578b\u4f9d\u7136\u662f\u4f7f\u7528\u903b\u8f91\u56de\u5f52\u5206\u7c7b\u7f51\u7edc\u7684\u6846\u67b6":10,"\u8be5\u76ee\u5f55\u4e0b\u6709\u4e24\u4e2a\u6587\u4ef6":0,"\u8be5\u811a\u672c\u7684\u4f7f\u7528\u65b9\u6cd5\u662f":0,"\u8be5image\u57fa\u4e8eubuntu":0,"\u8be5image\u7684\u6784\u5efa\u5728dock":0,"\u8be6\u60c5\u8bf7\u53c2\u8003":23,"\u8be6\u7ec6\u7684\u53c2\u6570\u89e3\u91ca\u5982\u4e0b\u9762\u8868\u683c":10,"\u8be6\u7ec6\u7684\u547d\u4ee4\u884c\u53c2\u6570\u8bf7\u53c2\u8003":23,"\u8be6\u7ec6\u7684cmake\u4f7f\u7528\u65b9\u6cd5\u53ef\u4ee5\u53c2\u8003":1,"\u8bf4\u660e":1,"\u8bf4\u660e\u547d\u4ee4\u884c\u53c2\u6570":7,"\u8bf7\u53c2\u8003":[6,20],"\u8bf7\u53c2\u8003\u4e0b\u8282refer":20,"\u8bf7\u53c2\u8003\u4e0b\u8ff0\u6587\u7ae0":19,"\u8bf7\u5b89\u88c5cuda":7,"\u8bfb\u5165\u89e3\u6790\u8bad\u7ec3\u914d\u7f6e":23,"\u8bfb\u53d6\u6570\u636e":20,"\u8c03\u7528":1,"\u8c03\u7528\u4e00\u6b21":20,"\u8c03\u7528\u7b2c\u4e8c\u6b21\u7684\u65f6\u5019":20,"\u8d1f\u6837\u672c":10,"\u8d1f\u8d23\u591a\u673a\u8bad\u7ec3\u4e2d\u7684\u53c2\u6570\u805a\u5408\u5de5\u4f5c":13,"\u8d1f\u9762\u60c5\u7eea\u4e24\u7c7b":20,"\u8def\u5f84\u53d8\u91cf\u4e3a":1,"\u8f93\u5165n\u4e2a\u5355\u8bcd":10,"\u8f93\u51fa\u4e3an\u4e2aword":10,"\u8fd0\u884c":[6,7],"\u8fd0\u884c\u4f7f\u7528\u7684cudnn\u5c3d\u91cf\u662f\u540c\u4e00\u4e2a\u7248\u672c":1,"\u8fd0\u884c\u8fd9\u4e2acontain":6,"\u8fd0\u884cpaddlepaddle\u7684gpu\u7248\u672c\u4e00\u5b9a\u8981\u5728\u5177\u6709cuda\u7684\u673a\u5668\u4e0a\u8fd0\u884c":0,"\u8fd4\u56de0":20,"\u8fd4\u56de\u4e00\u4e2alist\u6216\u8005tupl":20,"\u8fd4\u56de\u6570\u636e\u5728paddlepaddle\u4e2d\u662f\u4ec5\u4ec5\u8fd4\u56de\u4e00\u6761\u5b8c\u6574\u7684\u8bad\u7ec3\u6837\u672c":20,"\u8fd4\u56de\u7684\u987a\u5e8f\u9700\u8981\u548c":20,"\u8fd4\u56debatch_size\u7684\u5927\u5c0f":20,"\u8fd9":10,"\u8fd93\u4e2a\u5b50\u6b65\u9aa4\u53ef\u914d\u7f6e\u4e3a":10,"\u8fd9\u4e2a\u4e5f\u662fpaddlepaddle\u6240\u80fd\u591f\u4fdd\u8bc1\u7684shuffle\u7c92\u5ea6":20,"\u8fd9\u4e2a\u51fd\u6570\u4ee5\u4e00\u6761\u6570\u636e\u4e3a\u53c2\u6570":20,"\u8fd9\u4e2a\u51fd\u6570\u4f1a\u5728":20,"\u8fd9\u4e2a\u51fd\u6570\u5728\u521d\u59cb\u5316\u7684\u65f6\u5019\u4f1a\u88ab\u8c03\u7528":20,"\u8fd9\u4e2a\u51fd\u6570\u7684\u53c2\u6570\u662f":20,"\u8fd9\u4e2a\u521d\u59cb\u5316\u51fd\u6570\u5177\u6709\u5982\u4e0b\u53c2\u6570":20,"\u8fd9\u4e2a\u53c2\u6570\u5728\u8fd9\u4e2a\u6837\u4f8b\u91cc\u6ca1\u6709\u4f7f\u7528":20,"\u8fd9\u4e2a\u53c2\u6570\u88abpaddlepaddle\u8fdb\u7a0b\u4f20\u5165":20,"\u8fd9\u4e2a\u548c\u5728":20,"\u8fd9\u4e2a\u58f0\u660e\u57fa\u672c\u4e0a\u548cmnist\u7684\u6837\u4f8b\u4e00\u81f4":20,"\u8fd9\u4e2a\u5b57\u5178\u53ef\u4ee5\u5728":20,"\u8fd9\u4e2a\u5bf9\u5e94\u5173\u7cfb\u53ef\u80fd\u4e0d\u6b63\u786e":20,"\u8fd9\u4e2a\u5bf9\u8c61\u548cprocess\u7684\u7b2c\u4e00\u4e2a\u53c2\u6570\u4e00\u81f4":20,"\u8fd9\u4e2a\u5de5\u5177\u7c7b\u63a5\u6536\u548cpydataprovider2\u4e00\u6837\u7684\u8f93\u5165\u6570\u636e":23,"\u8fd9\u4e2a\u5e8f\u5217\u6a21\u578b\u6bd4\u8f83\u590d\u6742":20,"\u8fd9\u4e2a\u63a5\u53e3\u5e76\u4e0d\u7528\u6237\u53cb\u597d":23,"\u8fd9\u4e2a\u663e\u793a\u5668\u5f88\u68d2":10,"\u8fd9\u4e2a\u672c\u8eab\u4e0d\u662f\u4e00\u4e2a\u5f88":20,"\u8fd9\u4e2a\u6a21\u5757\u4e2d\u7684":20,"\u8fd9\u4e2a\u8bbe\u7f6e\u4e3a":20,"\u8fd9\u4e2a\u8f6f\u4ef6\u5305\u6587\u6863\u76f8\u5bf9\u5b8c\u5584":23,"\u8fd9\u4e2a\u95ee\u9898\u662fpydataprovider\u8bfb\u6570\u636e\u65f6\u5019\u7684\u903b\u8f91\u95ee\u9898":20,"\u8fd9\u4e9b\u53c2\u6570\u5305\u62ecpaddle\u5b9a\u4e49\u7684\u53c2\u6570":20,"\u8fd9\u4e9b\u53d8":1,"\u8fd9\u4e9b\u53d8\u91cf\u53ea\u5728\u7b2c\u4e00\u6b21cmake\u7684\u65f6\u5019\u6709\u6548":1,"\u8fd9\u4e9b\u53d8\u91cf\u5747\u53ef\u4ee5\u4f7f\u7528":1,"\u8fd9\u4e9b\u6d41\u7a0b\u4e2d\u7684\u6570\u636e\u4e0b\u8f7d":10,"\u8fd9\u6837\u505a\u53ef\u4ee5\u907f\u514d\u5f88\u591a\u6b7b\u9501\u95ee\u9898":20,"\u8fd9\u884c\u7684\u4f5c\u7528\u662f\u8bbe\u7f6edataprovider\u7684\u4e00\u4e9b\u5c5e\u6027":20,"\u8fd9\u91cc":20,"\u8fd9\u91cc\u4e3e\u4f8b\u7684\u6570\u636e\u662f\u82f1\u6587\u60c5\u611f\u5206\u7c7b\u7684\u6570\u636e":20,"\u8fd9\u91cc\u4ee5":10,"\u8fd9\u91cc\u4ee5mnist\u624b\u5199\u8bc6\u522b\u4e3a\u4f8b":20,"\u8fd9\u91cc\u53ef\u4ee5\u53c2\u8003paddle\u7684":18,"\u8fd9\u91cc\u6211\u4eec\u4f7f\u7528\u7b80\u5355\u7684\u6587\u672c\u6587\u4ef6\u8868\u793amnist\u56fe\u7247":20,"\u8fd9\u91cc\u6307\u5b9a\u8bcd\u5178":10,"\u8fd9\u91cc\u6ca1\u6709\u4ecb\u7ecd\u591a\u673a\u5206\u5e03\u5f0f\u8bad\u7ec3":10,"\u8fd9\u91cc\u7684":20,"\u8fd9\u91cc\u7684\u8f93\u5165\u7279\u5f81\u662f\u8bcdid\u7684\u5e8f\u5217":20,"\u8fd9\u91cc\u8981\u6ce8\u610f\u9884\u6d4b\u6570\u636e\u901a\u5e38":23,"\u8fd9\u91cc\u8bbe\u7f6e\u7684\u662f\u8fd4\u56de\u4e00\u4e2a":20,"\u8fd9\u91cc\u8bf4\u660e\u4e86\u8bad\u7ec3\u6570\u636e\u662f":20,"\u8fd9\u91cc\u91c7\u7528adam\u4f18\u5316\u65b9\u6cd5":10,"\u8fdb\u5165\u8be5\u6e90\u7801\u76ee\u5f55":0,"\u8fdb\u5165docker":6,"\u8fdc\u7a0b\u8bbf\u95ee":6,"\u8fde\u63a5\u8bf7\u53c2\u8003":10,"\u9009\u62e9\u666e\u901acpu\u7248\u672c\u7684devel\u7248\u672c\u7684imag":6,"\u9009\u9879":1,"\u901a\u8fc7\u7f16\u8bd1\u65f6\u6307\u5b9a\u8def\u5f84\u6765\u5b9e\u73b0\u5f15\u7528\u5404\u79cdbla":1,"\u903b\u8f91\u56de\u5f52":10,"\u90a3\u4e48":20,"\u90a3\u4e48\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d":19,"\u90a3\u4e48\u5bf9\u5e94\u7684dataprovider\u65e2\u4e3a":20,"\u90a3\u4e48\u8fd9\u4e2a\u4e0b\u8f7d\u7ebf\u7a0b\u5c06\u4f1a\u5173\u95ed":0,"\u90a3\u4e48paddlepaddle\u4f1a\u7c97\u7565\u7684\u6839\u636elayer\u7684\u58f0\u660e\u987a\u5e8f":20,"\u90fd\u4f20\u9012\u7ed9process\u51fd\u6570":20,"\u914d\u7f6e\u4e86":20,"\u914d\u7f6e\u53c2\u6570\u914d\u7f6e\u7ed9dataprovider\u7684":20,"\u914d\u7f6e\u6587\u4ef6":10,"\u91cc\u4f1a\u7ee7\u7eed\u5b89\u88c5":7,"\u91cc\u63d0\u4f9b\u4e86\u6570\u636e\u4e0b\u8f7d\u811a\u672c":10,"\u91cc\u9762\u8bfb\u53d6":20,"\u91cf\u4e5f\u53ef\u4ee5\u901a\u8fc7\u8c03\u7528cmake\u547d\u4ee4\u524d\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u6307\u5b9a":1,"\u9488\u5bf9\u672c\u95ee\u9898":10,"\u94fe\u63a5\u4f55\u79cdblas\u7b49\u7b49":1,"\u9519\u8bef\u7387":10,"\u95f4\u9694":20,"\u9664\u4e86":20,"\u9664\u8fc7data\u5c42":10,"\u9700\u8981\u53c2\u8003":6,"\u9700\u8981\u652f\u6301avx\u6307\u4ee4\u96c6\u7684cpu":6,"\u9700\u8981\u6ce8\u610f":20,"\u9700\u8981\u6ce8\u610f\u7684\u662f":[1,7],"\u9884\u6d4b\u6570\u636e\u6307\u5b9atest":10,"\u9884\u6d4b\u7ed3\u679c\u4ee5\u6587\u672c\u7684\u5f62\u5f0f\u4fdd\u5b58\u5728":10,"\u9884\u6d4b\u811a\u672c":10,"\u9884\u6d4bid":10,"\u989d\u5916\u7684\u53c2\u6570":10,"\u9996\u5148\u5217\u4e3e\u903b\u8f91\u56de\u5f52\u7f51\u7edc":10,"\u9996\u5148\u6211\u4eec\u5c06\u8fd9\u4e2a\u6570\u636e\u6587\u4ef6":20,"\u9996\u5148\u8bf7\u53c2\u8003":10,"\u9aa43":10,"\u9ed8\u8ba4\u4e00\u4e2apass\u4fdd\u5b58\u4e00\u6b21\u6a21\u578b":10,"\u9ed8\u8ba4\u503c":1,"\u9ed8\u8ba4\u5355\u4f4d\u662fbyte":0,"\u9ed8\u8ba4\u60c5\u51b5\u4e0b\u4e00\u6761\u6570\u636e":20,"adamax\u7b49":10,"amazon\u7535\u5b50\u4ea7\u54c1\u8bc4\u8bba\u6570\u636e":10,"api\u9884\u6d4b":10,"argument\u4f20\u5165":20,"argument\u5f62\u5f0f\u4f20\u5165":20,"atlas\u5e93\u7684\u8def\u5f84":1,"batches\u8bbe\u7f6e\u6bcf\u9694\u591a\u5c11batch\u4fdd\u5b58\u4e00\u6b21\u6a21\u578b":10,"bool\u53c2\u6570":20,"case":[10,22],"cd\u5230\u542b\u6709dockerfile\u7684\u8def\u5f84\u4e2d":6,"check\u662ffalse\u7684\u8bdd":20,"cmake\u53ef\u4ee5\u5c06cmake\u9879\u76ee\u6587\u4ef6":1,"cmake\u662f\u4e00\u4e2a\u8de8\u5e73\u53f0\u7684\u7f16\u8bd1\u811a\u672c":1,"cmake\u7684\u5b98\u65b9\u6587\u6863":1,"cmake\u7f16\u8bd1\u65f6\u4f1a\u9996\u5148\u5728\u7cfb\u7edf\u8def\u5f84":1,"container\u540e":6,"cpu\u7248\u672c":6,"cuda\u76f8\u5173\u7684driver\u548c\u8bbe\u5907\u6620\u5c04\u8fdbcontainer\u4e2d":6,"d\u547d\u4ee4\u5373\u53ef":1,"d\u547d\u4ee4\u6307\u5b9a":1,"dataprovider\u521b\u5efa\u7684\u65f6\u5019\u6267\u884c":20,"dataprovider\u53ef\u4ee5\u662f":20,"dataprovider\u63d0\u4f9b\u4e86\u4e24\u79cd\u7b80\u5355\u7684cache\u7b56\u7565":20,"dataprovider\u7684\u5177\u4f53\u7528\u6cd5\u548c\u5982\u4f55\u5b9e\u73b0\u4e00\u4e2a\u65b0\u7684dataprovid":19,"devel\u548cdemo":6,"dim\u7684\u65b0\u7684\u5411\u91cf":10,"dim\u7ef4\u5ea6\u5411\u91cf":10,"docker\u662f\u4e00\u4e2a\u57fa\u4e8e\u5bb9\u5668\u7684\u8f7b\u91cf\u7ea7\u865a\u62df\u73af\u5883":6,"docker\u7684\u5b98\u65b9\u6587\u6863":6,"dockerfile\u548cbuild":0,"dockerfile\u662fdock":0,"dockerfile\u7684\u6587\u6863":6,"dockerfile\u7684\u6700\u4f73\u5b9e\u8df5":6,"driver\u6dfb\u52a0\u5230ld_library_path\u4e2d":7,"elec\u6d4b\u8bd5\u96c6":10,"embedding\u6a21\u578b\u9700\u8981\u7a0d\u5fae\u6539\u53d8\u6570\u636e\u63d0\u4f9b\u7684\u811a\u672c":10,"export":[1,6,7],"f\u4ee3\u8868\u4e00\u4e2a\u6d6e\u70b9\u6570":20,"float":20,"generator\u4fbf\u4f1a\u5b58\u4e0b\u5f53\u524d\u7684\u4e0a\u4e0b\u6587":20,"generator\u7684\u4e0a\u4e0b\u6587\u4e2d\u5c3d\u91cf\u7559":20,"generator\u81f3\u5c11\u8c03\u7528\u4e24\u6b21\u624d\u4f1a\u77e5\u9053\u662f\u5426\u505c\u6b62":20,"gpu\u53cc\u7f13\u5b58":20,"gpu\u7248\u672c":6,"gpu\u7248\u672c\u4e8c\u8fdb\u5236":1,"gru\u6a21\u578b":10,"gru\u6a21\u578b\u914d\u7f6e":10,"i\u4ee3\u8868\u4e00\u4e2a\u6574\u6570":20,"id\u4e3a0\u7684\u6982\u7387":10,"id\u4e3a1\u7684\u6982\u7387":10,"image\u6784\u5efa\u6e90\u7801\u653e\u7f6e\u5728":0,"image\u7684\u4e3b\u8981\u63cf\u8ff0\u6587\u4ef6":0,"image\u7684\u4e3b\u8981\u6784\u5efa\u6b65\u9aa4":0,"image\u7684\u6784\u5efa\u6b65\u9aa4":0,"import":[10,20,23],"include\u4e0b\u9700\u8981\u5305\u542bcbla":1,"include\u4e0b\u9700\u8981\u5305\u542bmkl":1,"init_hook\u53ef\u4ee5\u4f20\u5165\u4e00\u4e2a\u51fd\u6570":20,"int":[10,20],"key\u662fdata_layer\u7684\u540d\u5b57":20,"layer\u4f5c\u4e3a\u8f93\u51fa":23,"layer\u6587\u6863":10,"ld_library_path\u7b49\u7b49":7,"ld_library_path\u91cc\u9762\u627e\u4e0d\u5230\u8fd9\u4e9b\u52a8\u6001":7,"lib\u4e0b\u9700\u8981\u5305\u542bcblas\u548catlas\u4e24\u4e2a\u5e93":1,"lib\u4e0b\u9700\u8981\u5305\u542bcblas\u5e93":1,"lib\u4e0b\u9700\u8981\u5305\u542bopenblas\u5e93":1,"lib\u76ee\u5f55\u4e0b\u9700\u8981\u5305\u542b":1,"list\u4e0d\u8bbe\u7f6e":19,"list\u4e2d":[19,20],"list\u4e2d\u7684\u4e00\u884c":20,"list\u4e2d\u7684\u6bcf\u4e00\u884c":20,"list\u4e3a\u7eaf\u6587\u672c\u6587\u4ef6":19,"list\u4e5f\u53ef\u4ee5\u653e\u7f6ehdfs\u6587\u4ef6\u8def\u5f84":19,"list\u5199\u5165\u90a3\u4e2a\u6587\u672c\u6587\u4ef6\u7684\u5730\u5740":20,"list\u5373\u4e3a":20,"list\u548ctest":19,"list\u5747\u4e3a\u672c\u5730\u7684\u4e24\u4e2a\u6587\u4ef6":19,"list\u6307\u5b9a\u7684\u6570\u636e":10,"list\u7684\u4f4d\u7f6e":10,"list\u82e5\u5e72\u6570\u636e\u6587\u4ef6\u8def\u5f84\u7684\u67d0\u4e00\u4e2a\u8def\u5f84":20,"lstm\u6a21\u578b\u7b49":10,"lstm\u6a21\u578b\u914d\u7f6e":10,"make\u548cmak":2,"mkl\u7684\u8def\u5f84":1,"mkl_sequential\u548cmkl_intel_lp64\u4e09\u4e2a\u5e93":1,"mnist\u662f\u4e00\u4e2a\u5305\u542b\u6709":20,"movielens\u6570\u636e\u96c6":9,"movielens\u8bc4\u5206\u56de\u5f52":9,"name\u90fd\u662f":6,"osx\u6216\u8005\u662fwindows\u673a\u5668":6,"osx\u7684\u5b89\u88c5\u6587\u6863":6,"paddle\u5b9a\u4e49\u7684\u53c2\u6570\u5305\u62ec":20,"paddle\u7684":7,"paddlepaddle\u4f7f\u7528\u8fd0\u884c\u65f6\u52a8\u6001\u8fde\u63a5cuda\u7684so":7,"paddlepaddle\u4fdd\u7559\u6dfb\u52a0\u53c2\u6570\u7684\u6743\u529b":20,"paddlepaddle\u53ef\u4ee5\u4f7f\u7528":1,"paddlepaddle\u53ef\u4ee5\u8bfb\u53d6python\u5199\u7684\u4f20\u8f93\u6570\u636e\u811a\u672c":10,"paddlepaddle\u5728\u8fd0\u884c\u65f6\u627e\u4e0d\u5230\u5bf9\u5e94\u7684config\u6587\u4ef6":7,"paddlepaddle\u5c06train":20,"paddlepaddle\u63a8\u8350\u4f7f\u7528docker\u8fdb\u884cpaddlepaddle\u7684\u90e8\u7f72\u548c":6,"paddlepaddle\u63d0\u4f9b\u4e86docker\u7684\u4f7f\u7528\u955c\u50cf":6,"paddlepaddle\u63d0\u4f9b\u6570\u4e2a\u9884\u7f16\u8bd1\u7684\u4e8c\u8fdb\u5236\u6765\u8fdb\u884c\u5b89\u88c5":5,"paddlepaddle\u63d0\u4f9b\u7684\u955c\u50cf\u5e76\u4e0d\u5305\u542b\u4efb\u4f55\u547d\u4ee4\u8fd0\u884c":6,"paddlepaddle\u7684\u6570\u636e\u5305\u62ec\u56db\u79cd\u4e3b\u8981\u7c7b\u578b":20,"paddlepaddle\u7684\u7f16\u8bd1\u9009\u9879\u53ef\u4ee5\u5728\u8c03\u7528cmake\u7684\u65f6\u5019\u8bbe\u7f6e":1,"paddlepaddle\u7684\u7f16\u8bd1\u9009\u9879\u662f\u53ef\u4ee5\u63a7\u5236paddlepaddle\u751f\u6210cpu":1,"paddlepaddle\u7684dock":0,"paddlepaddle\u7684python\u9884\u6d4b\u63a5\u53e3":22,"paddlepaddle\u7684ubuntu\u5b89\u88c5\u5305\u5206\u4e3a\u56db\u4e2a\u7248\u672c":7,"paddlepaddle\u76ee\u524d\u4f7f\u7528swig\u5bf9\u5176\u5e38\u7528\u7684\u9884\u6d4b\u63a5\u53e3\u8fdb\u884c\u4e86\u5c01\u88c5":23,"paddlepaddle\u76ee\u524d\u652f\u6301\u4f7f\u7528deb\u5305\u5b89\u88c5":7,"paddlepaddle\u8fd0\u884c\u65f6\u5982\u679c\u6ca1\u6709\u5bfb\u627e\u5230cuda\u7684driv":7,"paddlepaddle\u9700\u8981\u7528\u6237\u5728\u7f51\u7edc\u914d\u7f6e":19,"period\u8bbe\u7f6e\u6253\u5370\u53c2\u6570\u4fe1\u606f\u7b49":10,"process\u51fd\u6570":20,"process\u51fd\u6570\u662f\u5b9e\u73b0\u6570\u636e\u8f93\u5165\u7684\u4e3b\u51fd\u6570":20,"process\u51fd\u6570\u8c03\u7528\u591a\u6b21":20,"pserver\u4e3apaddlepaddle\u7684paramet":13,"pserver\u7684\u547d\u4ee4\u884c\u53c2\u6570":13,"pserver\u7ec4\u5408\u4f7f\u7528":13,"py\u6587\u4ef6\u7ed9\u51fa\u4e86\u5b8c\u6574\u4f8b\u5b50":10,"pydataprovider2\u4f1a\u5c3d\u91cf\u4f7f\u7528\u5185\u5b58":20,"pydataprovider2\u6587\u6863":23,"pydataprovider2\u7684\u4f7f\u7528":19,"pydataprovider\u662fpaddlepaddle\u4f7f\u7528python\u63d0\u4f9b\u6570\u636e\u7684\u63a8\u8350\u63a5\u53e3":20,"python\u5305":6,"python\u53ef\u4ee5\u89e3\u9664\u6389\u5185\u90e8\u53d8\u91cf\u7684\u5f15\u7528":20,"python\u7684":6,"python\u7684swig\u63a5\u53e3\u53ef\u4ee5\u65b9\u4fbf\u8fdb\u884c\u9884\u6d4b\u548c\u5b9a\u5236\u5316\u8bad\u7ec3":1,"return":[10,20],"rnn\u914d\u7f6e":11,"server\u8fdb\u7a0b":13,"sh\u662fdocker":0,"shuffle\u8bad\u7ec3\u6570\u636e":20,"softmax\u8f93\u51fa":10,"string\u7684\u683c\u5f0f\u6253\u5370\u51fa\u6765":13,"swig_paddle\u63a5\u53d7\u7684\u539f\u59cb\u6570\u636e\u662fc":23,"tag\u5206\u522b\u4e3a":6,"train\u5373\u4e3apaddlepaddle\u7684\u8bad\u7ec3\u8fdb\u7a0b":13,"train\u5b8c\u6210\u5355\u673a\u591a\u663e\u5361\u591a\u7ebf\u7a0b\u7684\u8bad":13,"train\u7684\u547d\u4ee4\u884c\u53c2\u6570":13,"ubuntu\u7684deb\u5b89\u88c5\u5305\u7b49":5,"v2\u4e4b\u540e\u7684\u4efb\u4f55\u4e00\u4e2acudnn\u7248\u672c\u6765\u7f16\u8bd1\u8fd0\u884c":1,"value\u5373\u4e3asoftmax\u5c42\u7684\u8f93\u51fa":23,"value\u662f\u7279\u5f81\u503c":20,"var":6,"vector\u8868\u793a\u7684\u6bcf\u4e2a\u5355\u8bcd":10,"version\u53ef\u4ee5\u6253\u5370\u51fapaddle\u7684\u7248\u672c\u4fe1\u606f\u548c\u7f16\u8bd1\u7684\u9009\u9879":18,"version\u53ef\u4ee5\u6253\u5370\u51fapaddlepaddle\u7684\u7248\u672c\u548c\u7f16\u8bd1\u65f6\u4fe1\u606f":13,"version\u7684\u547d\u4ee4\u884c\u53c2\u6570":13,"yield\u6587\u672c\u4fe1\u606f\u548c\u7c7b\u522bid":10,__main__:23,__name__:23,abov:20,act:10,act_typ:10,activ:10,adadelta:10,adagrad:10,adam:10,adamoptim:10,afi:20,all:20,allow:10,alreadi:7,also:10,append:20,apt:[6,7],arg:[0,10,20],around:20,arrai:23,assert:23,atla:1,atlas_root:1,avg:10,avgcost:10,avgpool:10,avx:6,bag:10,baidu:[6,7],batch:10,batch_siz:10,binari:10,bla:1,bool:10,both:10,bow:10,build:[0,6],cach:[10,19],cache_pass_in_mem:[10,20],cachetyp:[10,20],calc_batch_s:20,call:10,can:10,can_over_batch_s:20,cat:6,categori:10,check:[7,20],check_fail_continu:20,chines:9,chpasswd:6,classif:10,classification_cost:10,classification_error_evalu:10,close:20,cmake:1,cmd:6,cnn:10,code:[0,20,23],com:[6,7],comment:10,compil:[7,18],conf:23,config:[7,10],config_arg:10,config_pars:23,connect:10,contain:[10,20],context:20,context_len:10,context_start:10,convert:[10,20,23],couldn:7,cpp:[7,10],cpu:[6,7,20],cpuinfo:6,createfromconfigproto:23,cross:10,cuda_so:6,cudastat:7,cudasuccess:7,cudnn:1,cudnn_root:1,cudnnv5:1,current:[10,20],currentcost:10,currentev:10,dalla:20,data:7,data_config:23,data_initialz:10,data_lay:[10,20],dataprovid:10,dataprovider_bow:10,dataprovider_emb:10,dataproviderconvert:23,dataset:10,deb:7,debian:7,decor:20,def:[10,20,23],defin:[10,20],define_py_data_sources2:[10,20],delar:10,demo:[6,10],dense_vector:[20,23],describ:10,descript:22,detail:22,dev:6,devel:6,devic:6,devices:6,dict:[10,20],dict_fil:10,dictionai:10,dictionari:[10,20],dictrionari:10,differ:10,dim:10,dimens:10,dir:10,doc:23,documentari:20,dpkg:7,driver:7,dso_handl:7,dtype:23,dump_config:13,dure:[10,20],dynam:20,each:[10,20],each_pixel_str:20,each_word:20,echo:6,either:10,els:[6,10],emb:10,embed:9,embedding_lay:10,entropi:10,enumer:10,error:[7,10],etc:6,eval:10,exampl:10,expose:6,f0831:7,fail:7,fals:10,fc_layer:10,featur:[10,20],festiv:20,file:[10,20],file_list:20,file_nam:10,filenam:20,fill:10,find:7,first:10,float32:23,fly:10,forwardtest:23,framework:10,from:[6,10,20,23],fulli:10,func:20,gdebi:7,gener:[10,20],get:[6,7,10,20],get_config_arg:10,get_data:10,github:7,give:20,given:10,globe:20,gpu:[6,7],gradient_clipping_threshold:10,gradientmachin:23,grep:6,gru:10,gru_siz:10,help:23,hint:23,hl_cuda_devic:7,hl_dso_load:7,host:6,hot:10,hous:20,http:7,ignor:20,imag:6,imagenet:9,img:20,inarg:23,includ:10,init:10,init_hook:[10,19],init_model_path:10,initi:[10,20],initpaddl:23,input:[10,20],input_typ:[10,19],instal:2,insuffici:7,integ:[10,20],integer_sequ:20,integer_valu:[10,20],integer_value_sequ:10,invok:20,is_predict:10,is_train:20,isinst:23,iterat:20,job:10,kernel:6,kwarg:[10,20],l2regular:10,label:[10,20],label_dim:10,lake:20,later:10,latest:[0,6],layer:10,ld_library_path:7,learning_method:10,learning_r:10,len:[10,20],lib64:[6,7],lib:1,libcuda:6,libnvidia:6,librari:7,list:[10,19,20],load_data_arg:23,loadparamet:23,local:[1,7],log_period:10,logger:20,look:[10,20],loss:10,lowest_dl_speed:0,lstm:10,lstm_size:10,mac:6,main:23,maintainer:6,make:[7,20],make_diagram:13,maxid:10,maxid_lay:10,mean:10,memori:10,merge_model:13,method:20,min_pool_s:20,mkdir:6,mkl:1,mkl_core:1,mkl_root:1,mnist:20,mnist_model:23,mnist_provid:20,mnist_train:20,model_config:23,modul:[10,20],momentum:10,movi:20,must:7,name:[6,10,20],necessari:10,need:10,neg:[10,20],net:6,next:20,no_cache:20,no_sequence:20,noavx:[6,7],none:[10,20,23],normal:6,note:7,nullptr:7,num:10,num_pass:10,nvidia:6,obj:[10,20],object:[10,20],off:[0,1,7,18],omit:10,on_init:20,onli:10,open:[10,20],openbla:1,openblas_root:1,openssh:6,opt:1,other:10,outlin:22,output:10,outsid:20,paddl:[0,6,7,10,13],paddle_gpu:0,paddle_ssh:6,paddle_ssh_machin:6,paddledev:6,paddlepaddl:[6,7,18,23],paramet:10,parse_config:23,pass:[10,20],path:[7,10],period:10,permitrootlogin:6,pixel:20,pixels_float:20,pixels_str:20,place:20,pleas:7,pool_siz:20,pooling_lay:10,pooling_typ:10,posit:[10,20],pred:10,predict_output_dir:10,predict_sampl:23,preprocess:10,print:23,proc:6,process:[10,20],process_pr:10,properli:10,pull:6,put:10,py_paddl:[6,23],pydataprovid:19,pydataprovider2:[10,20,23],pydataproviderwrapp:10,python:10,quick_start:10,rang:10,rank:10,rare:20,read:[10,20],real_process:20,refer:19,reference_cblas_root:1,reffer:1,regular:10,releas:7,repres:10,represent:10,resnet:9,result:[10,20],rmsprop:10,roce:6,root:6,run:6,runtim:[7,20],same:[10,20],sampl:[10,20],save:[10,20],save_dir:10,saw:20,sbin:6,script:0,second:10,sed:6,see:10,sentenc:20,sentiment:20,sentimental_provid:20,separ:10,seq_typ:20,sequel:20,sequence:20,sequence_conv_pool:10,sequencetyp:20,server:6,set:[10,20],setup:10,should_shuffl:20,simple_gru:10,simple_lstm:10,size:[10,20],softmax:10,softmaxactiv:10,sourc:10,spars:10,sparse_binary_vector:[10,20],sparse_float_vector:20,specifi:[7,10],split:[10,20],src_root:23,ssh:6,sshd:6,sshd_config:6,stat:10,stop:6,store:10,string:20,strip:10,structur:10,stun:20,sub_sequence:20,sudo:7,support:6,sure:7,swig_paddl:23,tag:0,take:20,tbd:21,team:6,test:[10,19],test_data:23,test_list:[10,20],text:[10,20],text_conv:10,them:10,thi:[10,20],thing:20,tmp:20,train:7,train_list:[10,20],trainer:[10,20,23],trainer_config:[10,19,20,23],trainer_config_help:[10,20],trainerintern:10,trainermain:7,travel:20,trn:10,tst:10,two:10,txt:[10,20],type:[10,20],unk_idx:10,updat:6,use:[10,22],use_dynamic_ord:20,use_gpu:[10,23],usepam:6,user:10,usr:[1,6,7],valid:7,valu:[10,20,23],version:[6,7],via:7,want:20,what:10,when:20,which:10,whole:20,wilder:20,window:6,with_avx:[1,7,18],with_doc:1,with_doc_cn:1,with_doubl:[7,18],with_double:1,with_dso:1,with_gflag:[7,18],with_gflags:1,with_glog:[1,7,18],with_gpu:[0,1,7,18],with_metric_learn:[7,18],with_predict_sdk:[7,18],with_python:[1,7,18],with_rdma:[1,7,18],with_style_check:1,with_swig_py:1,with_testing:1,with_tim:[7,18],with_timer:1,without:6,wonder:20,word:9,word_dict:10,word_dim:10,word_id:20,word_vector:10,xarg:6,yield:[10,20],you:[7,20],your_host_machine:6},titles:["\u6784\u5efaPaddlePaddle Docker Image","\u8bbe\u7f6ePaddlePaddle\u7684\u7f16\u8bd1\u9009\u9879","\u4f7f\u7528cmake\u7f16\u8bd1PaddlePaddle","\u5b89\u88c5\u7f16\u8bd1PaddlePaddle\u9700\u8981\u7684\u4f9d\u8d56","make\u548cmake install","\u7f16\u8bd1\u4e0e\u5b89\u88c5","\u5b89\u88c5PaddlePaddle\u7684Docker\u955c\u50cf","\u4f7f\u7528deb\u5305\u5728Ubuntu\u4e0a\u5b89\u88c5PaddlePaddle","\u96c6\u7fa4\u8bad\u7ec3","\u4f7f\u7528\u793a\u4f8b","PaddlePaddle\u5feb\u901f\u5165\u95e8\u6559\u7a0b","PaddlePaddle\u6587\u6863","&lt;no title&gt;","\u547d\u4ee4\u884c\u53c2\u6570","&lt;no title&gt;","&lt;no title&gt;","paddle pserver\u7684\u547d\u4ee4\u884c\u53c2\u6570","paddle train\u7684\u547d\u4ee4\u884c\u53c2\u6570","paddle version\u7684\u547d\u4ee4\u884c\u53c2\u6570","PaddlePaddle\u7684\u6570\u636e\u63d0\u4f9b(DataProvider)\u4ecb\u7ecd","PyDataProvider2\u7684\u4f7f\u7528","\u81ea\u5b9a\u4e49\u4e00\u4e2aDataProvider","\u7528\u6237\u63a5\u53e3","PaddlePaddle\u7684Python\u9884\u6d4b\u63a5\u53e3"],titleterms:{"\u4e0b\u8f7d\u548c\u8fd0\u884cdocker\u955c\u50cf":6,"\u4ecb\u7ecd":19,"\u4f18\u5316\u7b97\u6cd5":10,"\u4f7f\u7528\u6307\u5357":11,"\u4f7f\u7528\u6982\u8ff0":10,"\u4f7f\u7528\u793a\u4f8b":9,"\u4f7f\u7528\u811a\u672c\u6784\u5efapaddlepaddl":0,"\u4f7f\u7528cmake\u7f16\u8bd1paddlepaddl":2,"\u4f7f\u7528deb\u5305\u5728ubuntu\u4e0a\u5b89\u88c5paddlepaddl":7,"\u5185\u5b58\u4e0d\u591f\u7528\u7684\u60c5\u51b5":20,"\u5377\u79ef\u6a21\u578b":10,"\u53c2\u8003":20,"\u53ef\u80fd\u7684\u5185\u5b58\u6cc4\u9732\u95ee\u9898":20,"\u53ef\u80fd\u9047\u5230\u7684\u95ee\u9898":7,"\u547d\u4ee4\u884c\u53c2\u6570":[10,13,22],"\u56fe\u50cf":9,"\u5b89\u88c5":[5,10],"\u5b89\u88c5\u7f16\u8bd1paddlepaddle\u9700\u8981\u7684\u4f9d\u8d56":3,"\u5b89\u88c5paddlepaddle\u7684docker\u955c\u50cf":6,"\u5e38\u7528\u6a21\u578b":9,"\u5e8f\u5217\u6a21\u578b\u6570\u636e\u63d0\u4f9b":20,"\u5f00\u53d1\u6307\u5357":11,"\u6027\u80fd\u95ee\u9898":6,"\u603b\u4f53\u6548\u679c\u603b\u7ed3":10,"\u63a8\u8350":9,"\u6570\u636e\u5411\u6a21\u578b\u4f20\u9001":10,"\u6570\u636e\u63d0\u4f9b":22,"\u6570\u636e\u683c\u5f0f\u51c6\u5907":10,"\u65f6\u5e8f\u6a21\u578b":10,"\u6784\u5efapaddlepaddl":0,"\u6ce8\u610f\u4e8b\u9879":[6,20],"\u7528\u6237\u63a5\u53e3":22,"\u7b80\u5355\u7684\u4f7f\u7528\u573a\u666f":20,"\u7b97\u6cd5\u6559\u7a0b":11,"\u7f16\u8bd1":5,"\u7f16\u8bd1\u4e0e\u5b89\u88c5":5,"\u7f51\u7edc\u7ed3\u6784":10,"\u81ea\u5b9a\u4e49\u4e00\u4e2adataprovid":21,"\u81ea\u7136\u8bed\u8a00\u5904\u7406":9,"\u8bad\u7ec3\u6a21\u578b":10,"\u8bbe\u7f6epaddlepaddle\u7684\u7f16\u8bd1\u9009\u9879":1,"\u8bcd\u5411\u91cf\u6a21\u578b":10,"\u8f93\u51fa\u65e5\u5fd7":10,"\u8fdc\u7a0b\u8bbf\u95ee\u95ee\u9898\u548c\u4e8c\u6b21\u5f00\u53d1":6,"\u903b\u8f91\u56de\u5f52\u6a21\u578b":10,"\u914d\u7f6e\u4e2d\u7684\u6570\u636e\u52a0\u8f7d\u5b9a\u4e49":10,"\u9644\u5f55":10,"\u96c6\u7fa4\u8bad\u7ec3":8,"\u9884\u6d4b":[10,22],"blas\u76f8\u5173\u7684\u7f16\u8bd1\u9009\u9879":1,"bool\u578b\u7684\u7f16\u8bd1\u9009\u9879":1,"config\u6587\u4ef6\u627e\u4e0d\u5230":7,"cudnn\u76f8\u5173\u7684\u7f16\u8bd1\u9009\u9879":1,"driver\u627e\u4e0d\u5230":7,"make\u548cmak":4,"paddlepaddle\u5feb\u901f\u5165\u95e8\u6559\u7a0b":10,"paddlepaddle\u63d0\u4f9b\u7684docker\u955c\u50cf\u7248\u672c":6,"paddlepaddle\u6587\u6863":11,"paddlepaddle\u7684\u6570\u636e\u63d0\u4f9b":19,"paddlepaddle\u7684\u7f16\u8bd1\u9009\u9879":1,"paddlepaddle\u7684bool\u578b\u7f16\u8bd1\u9009\u9879":1,"paddlepaddle\u7684cblas\u7f16\u8bd1\u9009\u9879":1,"paddlepaddle\u7684python\u9884\u6d4b\u63a5\u53e3":23,"pserver\u7684\u547d\u4ee4\u884c\u53c2\u6570":16,"pydataprovider2\u7684\u4f7f\u7528":20,"python\u6570\u636e\u52a0\u8f7d\u811a\u672c":10,"so\u627e\u4e0d\u5230":7,"train\u7684\u547d\u4ee4\u884c\u53c2\u6570":17,"version\u7684\u547d\u4ee4\u884c\u53c2\u6570":18,algorithm:10,appendix:10,architectur:10,argument:10,cach:20,command:10,configur:10,convolut:10,cuda:[1,7],data:10,dataprovid:19,docker:0,image:0,init_hook:20,input_typ:20,instal:4,install:10,libcudart:7,libcudnn:7,line:10,log:10,logist:10,model:10,network:10,optimiz:10,overview:10,paddl:[16,17,18],predict:10,prepar:10,provid:[10,20],refer:20,regress:10,script:10,sequenc:10,summari:10,time:10,train:10,transfer:10,vector:10,word:10}})
+Search.setIndex({envversion:49,filenames:["algorithm/rnn/hierarchical-layer","algorithm/rnn/hierarchical-rnn","algorithm/rnn/rnn-tutorial","build/docker/build_docker_image","build_and_install/cmake/compile_options","build_and_install/cmake/index","build_and_install/cmake/install_deps","build_and_install/cmake/make_and_install","build_and_install/index","build_and_install/install/docker_install","build_and_install/install/ubuntu_install","cluster/index","demo/index","demo/quick_start/index","faq/index","index","ui/cmd/dump_config","ui/cmd/index","ui/cmd/make_diagram","ui/cmd/merge_model","ui/cmd/paddle_pserver","ui/cmd/paddle_train","ui/cmd/paddle_version","ui/data_provider/index","ui/data_provider/pydataprovider2","ui/data_provider/write_new_dataprovider","ui/index","ui/predict/swig_py_paddle"],objects:{},objnames:{},objtypes:{},terms:{"0000x":13,"000\u5f20\u7070\u5ea6\u56fe\u7247\u7684\u6570\u5b57\u5206\u7c7b\u6570\u636e\u96c6":24,"00186201e":27,"04\u4e2d\u6b63\u786e":10,"08823112e":27,"0\u5c42\u5e8f\u5217":0,"0b1":10,"100m":14,"10\u4ee5\u4e0a\u7684linux":9,"10\u7ef4\u7684\u6574\u6570\u503c":24,"10gbe":9,"10m":3,"12194102e":27,"12\u7248\u672c\u6d4b\u8bd5\u901a\u8fc7":3,"12\u7248\u672c\u7684\u60c5\u51b5\u4e0b\u5e76\u6ca1\u6709\u6d4b\u8bd5":3,"15501715e":27,"15mb":13,"16mb":13,"1\u7684\u8bdd":24,"252kb":13,"25639710e":27,"25k":13,"27787406e":27,"28\u7684\u50cf\u7d20\u7070\u5ea6\u503c":24,"28\u7684\u7a20\u5bc6\u5411\u91cf\u548c\u4e00\u4e2a":24,"2\u4e09\u7c7b\u7684\u6bd4\u4f8b\u4e3a":14,"2\u4e2a\u5b50\u53e5":1,"2\u53e5\u53cc\u5c42\u5e8f\u5217\u548c5\u53e5\u5355\u5c42\u5e8f\u5217\u7684\u6570\u636e\u5b8c\u5168\u4e00\u6837":1,"2\u8868\u793a\u4e00\u6b21\u8fc72\u53e5\u53cc\u5c42\u5e8f\u5217":1,"2\u8fdb\u884c\u8fdb\u4e00\u6b65\u6f14\u5316":13,"32777140e":27,"36540484e":27,"3\u4e2a\u5b50\u53e5":1,"40gbe":9,"43630644e":27,"48565123e":27,"48684503e":27,"49316648e":27,"500m":14,"50k":3,"51111044e":27,"53018653e":27,"56gbe":9,"5\u5230\u672c\u5730\u73af\u5883\u4e2d":10,"5\u8868\u793a\u4e00\u6b21\u8fc75\u53e5\u5355\u5c42\u5e8f\u5217":1,"70634608e":27,"72296313e":27,"85625684e":27,"93137714e":27,"96644767e":27,"99982715e":27,"9\u7684\u6570\u5b57":24,"\u4e00":1,"\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a0\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u6269\u5c55\u6210\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u6269\u5c55\u6210\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217\u8fdb\u5165":2,"\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217\u6216\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u53d8\u6210\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u53d8\u6210\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217\u8fdb\u5165":2,"\u4e00\u4e2a\u53cc\u5c42rnn\u7531\u591a\u4e2a\u5355\u5c42rnn\u7ec4\u6210":2,"\u4e00\u4e2a\u53ef\u8c03\u7528\u7684\u51fd\u6570":2,"\u4e00\u4e2a\u6570\u636e\u96c6\u5927\u90e8\u5206\u5e8f\u5217\u957f\u5ea6\u662f100":14,"\u4e00\u4e2a\u6587\u4ef6":24,"\u4e00\u4e2a\u72ec\u7acb\u7684\u5143\u7d20":0,"\u4e00\u4e2a\u72ec\u7acb\u7684\u8bcd\u8bed":0,"\u4e00\u4e2a\u91cd\u8981\u7684\u95ee\u9898\u662f\u9009\u62e9\u6b63\u786e\u7684learning_r":14,"\u4e00\u4e2alabel":1,"\u4e00\u4e2alogging\u5bf9\u8c61":24,"\u4e00\u4e2apass\u8868\u793a\u8fc7\u4e00\u904d\u6240\u6709\u8bad\u7ec3\u6837\u672c":13,"\u4e00\u4eba":1,"\u4e00\u5171\u670910\u4e2a\u6837\u672c":1,"\u4e00\u5171\u67094\u4e2a\u6837\u672c":1,"\u4e00\u53e5\u8bdd\u662f\u7531\u8bcd\u8bed\u6784\u6210\u7684\u5e8f\u5217":2,"\u4e00\u65e9":1,"\u4e00\u662fbatch":14,"\u4e00\u6761":24,"\u4e00\u6b21\u6027\u676f\u5b50":1,"\u4e00\u81f4":1,"\u4e00\u81f4\u7684\u7279\u5f81":24,"\u4e00\u822c\u60c5\u51b5\u4e0b":23,"\u4e00\u822c\u63a8\u8350\u8bbe\u7f6e\u6210true":24,"\u4e00\u884c\u4e3a\u4e00\u4e2a\u6837\u672c":13,"\u4e00\u884c\u5bf9\u5e94\u4e00\u4e2a\u6570\u636e\u6587\u4ef6":23,"\u4e0a\u7684\u6548\u679c":13,"\u4e0a\u7f51":1,"\u4e0a\u8ff0\u4ee3\u7801\u5c06bias\u5168\u90e8\u521d\u59cb\u5316\u4e3a1":14,"\u4e0b\u6587\u4ee5nlp\u4efb\u52a1\u4e3a\u4f8b":2,"\u4e0b\u6b21":1,"\u4e0b\u8f7d\u8fdb\u7a0b\u4f1a\u91cd\u542f":3,"\u4e0b\u8ff0\u5185\u5bb9\u5c06\u5206\u4e3a\u5982\u4e0b\u51e0\u4e2a\u7c7b\u522b\u63cf\u8ff0":9,"\u4e0b\u975e\u5e38\u5c11\u7684\u53d8\u91cf\u5f15\u7528":24,"\u4e0b\u9762\u8fd9\u4e9blayer\u80fd\u591f\u63a5\u53d7\u53cc\u5c42\u5e8f\u5217\u4f5c\u4e3a\u8f93\u5165":0,"\u4e0b\u9762dataprovid":13,"\u4e0d":1,"\u4e0d\u4e00\u5b9a\u548c\u65f6\u95f4\u6709\u5173\u7cfb":24,"\u4e0d\u4f1a\u518d\u4ece":14,"\u4e0d\u4f1a\u6267\u884c\u6d4b\u8bd5\u64cd\u4f5c":23,"\u4e0d\u5305\u542blabel":27,"\u4e0d\u540c\u7684\u4f18\u5316\u7b97\u6cd5\u9700\u8981\u4f7f\u7528\u4e0d\u540c\u5927\u5c0f\u7684\u5185\u5b58":14,"\u4e0d\u540c\u7684\u6570\u636e\u7c7b\u578b\u548c\u5e8f\u5217\u6a21\u5f0f\u8fd4\u56de\u7684\u683c\u5f0f\u4e0d\u540c":24,"\u4e0d\u540c\u8f93\u5165\u542b\u6709\u7684\u5b50\u53e5":2,"\u4e0d\u540c\u8f93\u5165\u5e8f\u5217\u542b\u6709\u7684\u8bcd\u8bed\u6570\u5fc5\u987b\u4e25\u683c\u76f8\u7b49":2,"\u4e0d\u5c11":1,"\u4e0d\u5e94\u8be5\u88ab\u62c6\u89e3":2,"\u4e0d\u6307\u5b9a\u65f6":2,"\u4e0d\u652f\u6301avx\u6307\u4ee4\u96c6\u7684cpu\u4e5f\u53ef\u4ee5\u8fd0\u884c":9,"\u4e0d\u7f13\u5b58\u4efb\u4f55\u6570\u636e":24,"\u4e0d\u8fc7":1,"\u4e0d\u8fdc":1,"\u4e0d\u9519":1,"\u4e0d\u9700\u8981avx\u6307\u4ee4\u96c6\u7684cpu\u4e5f\u53ef\u4ee5\u8fd0\u884c":9,"\u4e0e\u8bad\u7ec3\u7f51\u7edc\u914d\u7f6e\u4e0d\u540c\u7684\u662f":13,"\u4e0e\u8fd9\u4e2a\u8bad\u7ec3\u6570\u636e\u4ea4\u4e92\u7684layer":14,"\u4e14":1,"\u4e14\u5e8f\u5217\u7684\u6bcf\u4e00\u4e2a\u5143\u7d20\u8fd8\u662f\u4e00\u4e2a\u65f6\u95f4\u5e8f\u5217":24,"\u4e24":1,"\u4e24\u4e2a\u5d4c\u5957\u7684":2,"\u4e24\u4e2a\u6587\u6863":9,"\u4e24\u7c7b":13,"\u4e25\u91cd\u7684\u95ee\u9898":24,"\u4e2a":13,"\u4e2a\u5185\u5b58\u6c60\u5b9e\u9645\u4e0a\u51b3\u5b9a\u4e86shuffle\u7684\u7c92\u5ea6":14,"\u4e2ayield":24,"\u4e2d":[13,14],"\u4e2d\u5b9a\u4e49\u4f7f\u7528\u54ea\u79cddataprovider\u53ca\u5176\u53c2\u6570":23,"\u4e2d\u5b9a\u4e49\u548c\u4f7f\u7528":2,"\u4e2d\u5b9a\u4e49\u7684\u987a\u5e8f\u4e00\u81f4":24,"\u4e2d\u5bfb\u627e\u8fd9\u4e9bblas\u7684\u5b9e\u73b0":4,"\u4e2d\u7684":24,"\u4e2d\u7684\u4e8c\u8fdb\u5236\u4f7f\u7528\u4e86":9,"\u4e2d\u7684set":24,"\u4e2d\u914d\u7f6e":24,"\u4e34\u65f6\u53d8\u91cf\u7b49\u7b49":14,"\u4e3a":24,"\u4e3a\u4e86\u63cf\u8ff0\u65b9\u4fbf":2,"\u4e3a\u4e86\u8fd0\u884cpaddlepaddle\u7684docker\u955c\u50cf":9,"\u4e3a\u4f8b\u8fdb\u884c\u9884\u6d4b":13,"\u4e3a\u53c2\u6570\u77e9\u9635\u7684\u5bbd\u5ea6":14,"\u4e3b\u8981\u51fd\u6570\u662fprocess\u51fd\u6570":24,"\u4e3b\u8981\u5206\u4e3a\u4ee5\u4e0b\u51e0\u4e2a\u6b65\u9aa4":27,"\u4e3b\u8981\u5305\u62ec\u4e24\u90e8\u5206":13,"\u4e3b\u8981\u539f\u56e0":1,"\u4e3b\u8981\u662f\u589e\u52a0\u4e86\u521d\u59cb\u5316\u673a\u5236":24,"\u4e3b\u8981\u6b65\u9aa4\u4e3a":27,"\u4e3b\u8981\u7531\u4e8e\u65e7\u7248\u672c":3,"\u4e3b\u8981\u7684\u8f6f\u4ef6\u5305\u4e3apy_paddl":27,"\u4e3e\u4e00\u4e2a\u4f8b\u5b50":14,"\u4e4b\u95f4\u7684\u8fd0\u7b97\u662f\u72ec\u7acb\u7684":2,"\u4e5f":1,"\u4e5f\u4e0d\u5b58\u5728\u4e00\u4e2asubseq\u76f4\u63a5\u751f\u6210\u4e0b\u4e00\u4e2asubseq\u7684\u60c5\u51b5":2,"\u4e5f\u4f1a\u6254\u5230\u8fd9\u6761\u6570\u636e":24,"\u4e5f\u4f1a\u8bfb\u53d6\u76f8\u5173\u8def\u5f84\u53d8\u91cf\u6765\u8fdb\u884c\u641c\u7d22":4,"\u4e5f\u53ef\u4ee5":24,"\u4e5f\u53ef\u4ee5\u4e3ainteg":1,"\u4e5f\u53ef\u4ee5\u4f7f\u7528":24,"\u4e5f\u53ef\u4ee5\u548cpaddl":17,"\u4e5f\u53ef\u4ee5\u662f\u4e00\u4e2a\u8bcd\u8bed":2,"\u4e5f\u53ef\u4ee5\u76f4\u63a5\u6267\u884c":9,"\u4e5f\u53ef\u4ee5\u901a\u8fc7\u5982\u4e0b\u65b9\u5f0f\u9884\u6d4b":13,"\u4e5f\u53ef\u4ee5\u901a\u8fc7save":13,"\u4e5f\u53ef\u4ee5\u9884\u6d4b\u6ca1\u6709label\u7684\u6d4b\u8bd5\u96c6":13,"\u4e5f\u5c31\u662f\u5c06\u8bcd\u5411\u91cf\u6a21\u578b\u989d\u6b65":13,"\u4e5f\u5c31\u662f\u76f4\u63a5\u5199\u5185\u5b58\u7684float\u6570\u7ec4":27,"\u4e5f\u662fdecoder\u5faa\u73af\u5c55\u5f00\u7684\u4f9d\u636e":2,"\u4e5f\u9700\u8981\u4e24\u6b21\u968f\u673a\u9009\u62e9\u5230\u540c\u6837\u7684generator\u7684\u65f6\u5019":24,"\u4e7e":1,"\u4e86":1,"\u4e86\u975e\u5e38\u65b9\u4fbf\u7684\u4e8c\u8fdb\u5236\u5206\u53d1\u624b\u6bb5":9,"\u4e8c\u6b21\u5f00\u53d1\u53ef\u4ee5":9,"\u4e94\u661f\u7ea7":1,"\u4ea4\u901a":1,"\u4ea4\u901a\u4fbf\u5229":1,"\u4eba\u5458\u7b49\u7b49":3,"\u4ec5\u4ec5\u9700\u8981":24,"\u4ecb\u7ecdpaddlepaddle\u4f7f\u7528\u6d41\u7a0b\u548c\u5e38\u7528\u7684\u7f51\u7edc\u57fa\u7840\u5355\u5143\u7684\u914d\u7f6e\u65b9\u6cd5":13,"\u4ece\u4e00\u4e2aword\u751f\u6210\u4e0b\u4e00\u4e2aword":2,"\u4ece\u6587\u4ef6\u4e2d\u8bfb\u53d6\u6bcf\u4e00\u6761\u6570\u636e":24,"\u4ece\u6587\u672c\u6587\u4ef6\u4e2d\u8bfb\u53d6":24,"\u4ece\u800c\u4e0d\u80fd\u5728\u8fd0\u884c\u7f16\u8bd1\u547d\u4ee4\u7684\u65f6\u5019\u63a5\u53d7\u53c2\u6570":3,"\u4ece\u800c\u751f\u6210\u591a\u4e2agener":24,"\u4ece\u800c\u9632\u6b62\u8fc7\u62df\u5408":23,"\u4ece\u8bed\u4e49\u4e0a\u770b":2,"\u4ece\u8f93\u5165\u6570\u636e\u4e0a\u770b":1,"\u4ed6\u4eec\u662f":[9,10,14,17,24],"\u4ed6\u4eec\u7684imag":9,"\u4ed6\u53ef\u4ee5\u5c06\u67d0\u4e00\u4e2a\u51fd\u6570\u6807\u8bb0\u6210\u4e00\u4e2apydataprovid":24,"\u4ee3\u8868\u4e00\u4e2a\u5411\u91cf":1,"\u4ee3\u8868\u4e0d\u540c\u7684\u53cc\u5c42\u5e8f\u5217":1,"\u4ee3\u8868\u5355\u5c42\u5e8f\u5217":1,"\u4ee3\u8868\u53cc\u5c42\u5e8f\u5217":1,"\u4ee4\u884c\u53c2\u6570\u6587\u6863":13,"\u4ee5\u53ca\u53cc\u5c42\u5e8f\u5217":0,"\u4ee5\u53ca\u8ba1\u7b97\u903b\u8f91\u5728\u5e8f\u5217\u4e0a\u7684\u5faa\u73af\u5c55\u5f00":2,"\u4ee5\u592a\u7f51\u5361":9,"\u4ee5\u76f8\u5bf9\u8def\u5f84\u5f15\u7528":23,"\u4ef7\u683c":1,"\u4efb\u610f\u4e00\u79cdcblas\u5b9e\u73b0":4,"\u4f1a\u5171\u4eab\u53c2\u6570":14,"\u4f1a\u5bf9\u6bcf\u4e00\u4e2a\u6fc0\u6d3b\u6682\u5b58\u4e00\u4e9b\u6570\u636e":14,"\u4f1a\u5bf9\u8fd9\u7c7b\u8f93\u5165\u8fdb\u884c\u62c6\u89e3":2,"\u4f1a\u5c06\u6bcf\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51fa\u62fc\u63a5":2,"\u4f1a\u5c06\u7b2c\u4e00\u4e2a":14,"\u4f1a\u6210\u4e3astep\u51fd\u6570\u7684\u8f93\u5165":2,"\u4f1a\u62a5\u5bfb\u627e\u4e0d\u5230\u8fd9\u4e9b\u52a8\u6001\u5e93":10,"\u4f1a\u62a5\u9519":2,"\u4f1a\u6839\u636e\u547d\u4ee4\u884c\u53c2\u6570\u6307\u5b9a\u7684\u6d4b\u8bd5\u65b9\u5f0f":23,"\u4f1a\u6839\u636einput_types\u68c0\u67e5\u6570\u636e\u7684\u5408\u6cd5\u6027":24,"\u4f1a\u751f\u6210\u591a\u4e2agener":24,"\u4f1a\u88ab\u62c6\u89e3\u4e3a\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":2,"\u4f1a\u88ab\u62c6\u89e3\u4e3a\u975e\u5e8f\u5217":2,"\u4f1a\u9884\u5148\u8bfb\u53d6\u5168\u90e8\u6570\u636e\u5230\u5185\u5b58\u4e2d":24,"\u4f20\u5165\u547d\u4ee4\u884c\u53c2\u6570\u521d\u59cb\u5316":27,"\u4f20\u5165\u7684\u662f\u4e00\u4e2a\u51fd\u6570":24,"\u4f20\u5165\u7684\u914d\u7f6e\u53c2\u6570\u5305\u62ec":3,"\u4f20\u5165\u8fd9\u4e2a\u53d8\u91cf\u7684\u65b9\u5f0f\u4e3a":24,"\u4f46\u4ece\u4e0a\u9762\u7684\u6570\u636e\u683c\u5f0f\u53ef\u77e5":1,"\u4f46\u5b50\u53e5\u542b\u6709\u7684\u8bcd\u8bed\u6570\u53ef\u4ee5\u4e0d\u76f8\u7b49":2,"\u4f46\u662f":[3,14,24],"\u4f46\u662f\u5728":24,"\u4f46\u662f\u5982\u679c\u4f7f\u7528\u4e86\u9ad8\u6027\u80fd\u7684\u7f51\u5361":9,"\u4f46\u662f\u65b9\u4fbf\u8c03\u8bd5\u548cbenchmark":4,"\u4f46\u662f\u6709\u65f6\u4e3a\u4e86\u8ba1\u7b97\u5747\u8861\u6027":24,"\u4f46\u662f\u7a81\u7136\u6709\u4e00\u4e2a10000\u957f\u7684\u5e8f\u5217":14,"\u4f46\u662fbatch":14,"\u4f46\u7406\u8bba\u4e0a\u652f\u6301\u5176\u4ed6\u7684":10,"\u4f46\u8fd9\u79cd\u60c5\u51b5\u4e0b":1,"\u4f46\u9700\u8981\u6ce8\u610f\u7684\u662f\u7f16\u8bd1\u548c":4,"\u4f4d\u7f6e":1,"\u4f4e\u4e8edocker":3,"\u4f4f":1,"\u4f53\u53ef\u4ee5\u53c2\u8003":24,"\u4f5c\u4e3a\u4e0b\u4e00\u4e2a\u5b50\u53e5memory\u7684\u521d\u59cb\u72b6\u6001":1,"\u4f5c\u4e3a\u53c2\u6570\u7684id":14,"\u4f5c\u4e3a\u5f53\u524d\u65f6\u523b\u8f93\u5165":2,"\u4f5c\u4e3aboot":1,"\u4f5c\u7528":0,"\u4f7f\u5728python\u73af\u5883\u4e0b\u7684\u9884\u6d4b\u63a5\u53e3\u66f4\u52a0\u7b80\u5355":27,"\u4f7f\u7528":[2,4,9,14,27],"\u4f7f\u7528\u4e00\u4e2a\u8bcd\u524d\u4e24\u4e2a\u8bcd\u548c\u540e\u4e24\u4e2a\u8bcd":14,"\u4f7f\u7528\u4e86\u540c\u6837\u7684parameter\u548cbia":14,"\u4f7f\u7528\u4e86avx\u6307\u4ee4":14,"\u4f7f\u7528\u591a\u5757\u663e\u5361\u8bad\u7ec3":14,"\u4f7f\u7528\u591a\u673a\u8bad\u7ec3\u7684\u65b9\u6cd5\u4e5f\u6bd4\u8f83\u7b80\u5355":14,"\u4f7f\u7528\u591a\u7ebf\u7a0b\u8bad\u7ec3":14,"\u4f7f\u7528\u5982\u4e0b\u811a\u672c\u53ef\u4ee5\u786e\u5b9a\u672c\u673a\u7684cpu\u77e5\u5426\u652f\u6301":9,"\u4f7f\u7528\u65b9\u6cd5\u4e3a":14,"\u4f7f\u7528\u663e\u5361\u8bad\u7ec3":14,"\u4f7f\u7528\u7684\u547d\u4ee4\u4e5f\u662f":4,"\u4f7f\u7528\u8005\u53ea\u9700\u8981\u5173\u6ce8\u4e8e\u8bbe\u8ba1rnn\u5728\u4e00\u4e2a\u65f6\u95f4\u6b65\u4e4b\u5185\u5b8c\u6210\u7684\u8ba1\u7b97":2,"\u4f7f\u7528\u8be5\u63a5\u53e3\u7528\u6237\u53ef\u4ee5\u53ea\u5173\u6ce8\u5982\u4f55":24,"\u4f7f\u7528\u8be5dockerfile\u6784\u5efa\u51fa\u955c\u50cf":9,"\u4f7f\u7528\u8fd9\u4e2a\u5173\u952e\u8bcd":24,"\u4f7f\u7528deb\u5305\u5728ubuntu\u4e0a\u5b89\u88c5paddlepaddl":8,"\u4f7f\u7528dockerfile\u6784\u5efa\u4e00\u4e2a\u5168\u65b0\u7684dock":9,"\u4f7f\u7528mnist\u624b\u5199\u8bc6\u522b\u4f5c\u4e3a\u6837\u4f8b":27,"\u4f7f\u7528ssh\u8bbf\u95eepaddlepaddle\u955c\u50cf":9,"\u4f86":1,"\u4f8b\u5982":[4,13,14,24],"\u4f8b\u5982\u5982\u679c\u4f7f\u7528":14,"\u4f8b\u5982\u6587\u4ef6\u540d\u662f":24,"\u4f8b\u5982rdma\u7f51\u5361":9,"\u4f8b\u5982sigmoid\u53d8\u6362":13,"\u4f9d\u6b21\u8fd4\u56de\u4e86\u6587\u4ef6\u4e2d\u7684\u6bcf\u6761\u6570\u636e":24,"\u4f9d\u7136\u4fdd\u6301\u6bcf\u4e2asubseq\u6700\u540e\u4e00\u4e2a\u5143\u7d20\u7684\u503c\u4e0d\u53d8":1,"\u4fbf\u5229":1,"\u4fbf\u5b9c":1,"\u4fe1\u606f":9,"\u5047\u5982\u6211\u4eec\u662f\u4e09\u5206\u7c7b\u95ee\u9898":14,"\u505a\u6cd5\u53ef\u4ee5\u6709\u4e24\u79cd":14,"\u505c\u7535":1,"\u5143\u7d20":0,"\u5143\u7d20\u4e4b\u95f4\u7684\u987a\u5e8f\u662f\u91cd\u8981\u7684\u8f93\u5165\u4fe1\u606f":0,"\u5168\u5bb6":1,"\u5173\u4e8edataprovider\u4e2dinput":1,"\u5173\u4e8eunbound":2,"\u5173\u95edcontain":9,"\u5176\u4e2d":[3,9,10,14,23,24,27],"\u5176\u4e2d\u6587\u672c\u8f93\u5165\u7c7b\u578b\u5b9a\u4e49\u4e3a\u6574\u6570\u65f6\u5e8f\u7c7b\u578binteg":13,"\u5176\u4e2d\u6bcf\u4e2a\u5143\u7d20\u662f\u53cc\u5c42\u5e8f\u5217\u4e2d\u6bcf\u4e2asubseq\u6700\u540e\u4e00\u4e2a":0,"\u5176\u4e2d\u7b2c\u4e00\u884c\u662f\u5f15\u5165paddlepaddle\u7684pydataprovider2\u5305":24,"\u5176\u4e2d\u7b2ci\u4e2asubseq\u4e2d\u7684\u6240\u6709\u5411\u91cf\u5747\u4e3a\u8f93\u5165\u7684\u5355\u5c42\u5e8f\u5217\u4e2d\u7684\u7b2ci\u4e2a\u5411\u91cf":1,"\u5176\u4ed6\u5185\u5b58\u6742\u9879":14,"\u5176\u4ed6\u5185\u5b58\u6742\u9879\u662f\u6307paddlepaddle\u672c\u8eab\u6240\u7528\u7684\u4e00\u4e9b\u5185\u5b58":14,"\u5176\u4ed6\u53c2\u6570\u5747\u4f7f\u7528kei":24,"\u5176\u4ed6\u53c2\u6570\u8bf7\u53c2\u8003":13,"\u5176\u4ed6\u53c2\u6570\u90fd\u4f7f\u7528kei":24,"\u5176\u4ed6\u7684\u5185\u5b58\u7684\u51cf\u5c11\u65b9\u6cd5\u4f9d\u6b21\u4e3a":14,"\u5176\u4f5c\u7528\u662f\u5c06\u8bad\u7ec3\u6570\u636e\u4f20\u5165\u5185\u5b58\u6216\u8005\u663e\u5b58":23,"\u5176\u5b83\u90e8\u5206\u548c\u903b\u8f91\u56de\u5f52\u7f51\u7edc\u7ed3\u6784\u4e00\u81f4":13,"\u5176\u5b83layer\u7684\u8f93\u51fa":2,"\u5176\u5b9e\u4e5f\u662f\u548c\u6bcf\u4e2amini":14,"\u5176\u6570\u636e\u4f7f\u7528":24,"\u5176\u6b21":1,"\u5176\u7b2c\u4e00\u884c\u8bf4\u660e\u4e86paddle\u7684\u7248\u672c":22,"\u5177":24,"\u5177\u4f53\u4f7f\u7528\u65b9\u6cd5\u4e3a":14,"\u5177\u4f53\u53ef\u4ee5\u8bbe\u7f6e\u6210\u4ec0\u4e48\u5176\u4ed6\u683c":24,"\u5177\u4f53\u53ef\u53c2\u8003\u6587\u6863":2,"\u5177\u4f53\u6709\u54ea\u4e9b\u683c\u5f0f":24,"\u5177\u4f53\u7684\u591a\u673a\u8bad\u7ec3\u65b9\u6cd5\u53c2\u8003":14,"\u5177\u4f53\u8bf7\u53c2\u8003\u6ce8\u610f\u4e8b\u9879\u4e2d\u7684":9,"\u5177\u4f53dataprovider\u8fd8\u5177\u6709\u4ec0\u4e48\u529f\u80fd":24,"\u5177\u6709\u4e24\u4e2a\u53c2\u6570":24,"\u5177\u6709\u548c\u5bbf\u4e3b\u673a\u76f8\u8fd1\u7684\u8fd0\u884c\u6548\u7387":9,"\u5177\u6709\u7684\u5c5e\u6027\u6709":24,"\u5178\u578b\u7684\u8f93\u51fa\u7ed3\u679c\u4e3a":27,"\u5178\u578b\u7684\u9884\u6d4b\u4ee3\u7801\u5982\u4e0b":27,"\u5185\u5b58\u4e0d\u591f\u7528\u7684\u60c5\u51b5":23,"\u5185\u5c42\u662f":1,"\u5185\u5c42inner":1,"\u518d\u6307\u5b9a":4,"\u5199\u5165train":24,"\u5199\u5728train":23,"\u51c6\u5907":1,"\u51c6\u5907\u6570\u636e":27,"\u51cf\u5c0f\u5e8f\u5217\u7684\u957f\u5ea6":14,"\u51cf\u5c0f\u8fd9\u4e2a\u5185\u5b58\u6c60\u5373\u53ef\u51cf\u5c0f\u5185\u5b58\u5360\u7528":14,"\u51cf\u5c0fbatch":14,"\u51fa\u53bb\u73a9":1,"\u51fa\u5dee":1,"\u51fa\u6765":1,"\u51fd\u6570":24,"\u51fd\u6570\u4e2d":24,"\u51fd\u6570\u4e2d\u4f7f\u7528":24,"\u51fd\u6570\u4e2d\u7684":24,"\u51fd\u6570\u53ea\u5173\u6ce8\u4e8ernn\u4e00\u4e2a\u65f6\u95f4\u6b65\u4e4b\u5185\u7684\u8ba1\u7b97":2,"\u51fd\u6570\u5fc5\u987b\u8fd4\u56de\u4e00\u4e2a\u6216\u591a\u4e2alayer\u7684\u8f93\u51fa":2,"\u51fd\u6570\u662f\u4f7f\u7528":24,"\u51fd\u6570\u6765\u4fdd\u8bc1\u517c\u5bb9\u6027":24,"\u51fd\u6570\u67e5\u8be2\u6587\u6863":27,"\u5206\u522b\u4e3a":1,"\u5206\u522b\u4ece\u8bcd\u8bed\u548c\u53e5\u5b50\u7ea7\u522b\u7f16\u7801\u8f93\u5165\u6570\u636e":2,"\u5206\u522b\u5b9a\u4e49\u5b50\u53e5\u7ea7\u522b\u548c\u8bcd\u8bed\u7ea7\u522b\u4e0a\u9700\u8981\u5b8c\u6210\u7684\u8fd0\u7b97":2,"\u5206\u522b\u662f":0,"\u5206\u5e03\u5f0f\u8bad\u7ec3":13,"\u5206\u6790\u5f97\u51fa":1,"\u5206\u7c7b\u6210\u6b63\u9762\u60c5\u7eea\u548c":24,"\u5217\u8868\u5982\u4e0b":24,"\u5219\u53ef\u4ee5\u4f7f\u7528":9,"\u5219\u53ef\u4ee5\u9009\u62e9\u4e0a\u8868\u4e2d\u7684avx\u7248\u672cpaddlepaddl":9,"\u5219\u5b57\u4e0e\u5b57\u4e4b\u95f4\u7528\u7a7a\u683c\u5206\u9694":13,"\u5219\u9700\u8981\u4f7f\u7528":10,"\u5219\u9700\u8981\u4f7f\u7528\u53c2\u6570\u89c4\u6a21\u5927\u7ea65\u500d\u7684\u5185\u5b58":14,"\u5219\u9700\u8981\u5148\u5c06":9,"\u5219\u9700\u8981\u8fdb\u884c\u4e00\u5b9a\u7684\u4e8c\u6b21\u5f00\u53d1":9,"\u521b\u5efa\u4e00\u4e2a":27,"\u521b\u5efagener":24,"\u521d\u59cb\u72b6\u6001":2,"\u5220\u9664contain":9,"\u5229\u7528\u5355\u8bcdid\u67e5\u627e\u5bf9\u5e94\u7684\u8be5\u5355\u8bcd\u7684\u8fde\u7eed\u8868\u793a\u5411\u91cf":13,"\u5229\u7528\u66f4\u591a\u7684\u8ba1\u7b97\u8d44\u6e90\u53ef\u4ee5\u5206\u4e3a\u4e00\u4e0b\u51e0\u4e2a\u65b9\u5f0f\u6765\u8fdb\u884c":14,"\u5229\u7528\u8fd9\u79cd\u7279\u6027":2,"\u5229\u7528\u903b\u8f91\u56de\u5f52\u6a21\u578b\u5bf9\u8be5\u5411\u91cf\u8fdb\u884c\u5206\u7c7b":13,"\u5229\u843d":1,"\u522b":13,"\u5237\u7259":1,"\u524d\u53f0":1,"\u5269\u4e0b\u7684pass\u4f1a\u76f4\u63a5\u4ece\u5185\u5b58\u91cc":24,"\u52a0\u4e86l2\u6b63\u5219\u548c\u68af\u5ea6\u622a\u65ad":13,"\u52a0\u8f7d\u6570\u636e":13,"\u52a0\u901fpaddlepaddle\u8bad\u7ec3\u6709\u5982\u4e0b\u51e0\u4e2a\u65b9\u9762":14,"\u5305":9,"\u5305\u548c":9,"\u5305\u62ec":13,"\u5305\u62ec\u5b57\u7b26\u4e32\u5206\u914d":14,"\u5305\u62ec\u6fc0\u6d3b":14,"\u5305\u62ec\u7b80\u5355\u7684rnn\u6a21\u578b":13,"\u5305\u62ecdocker\u955c\u50cf":8,"\u5305\u62ecpaddle\u7684\u4e8c\u8fdb\u5236":9,"\u5305\u62ecpaddle\u8fd0\u884cdemo\u6240\u9700\u8981\u7684\u4f9d\u8d56":9,"\u5341\u4e00":1,"\u534e\u6da6\u4e07\u5bb6":1,"\u5355\u53cc\u5c42\u5e8f\u5217\u7684\u53e5\u5b50\u662f\u4e00\u6837\u7684":1,"\u5355\u53cc\u5c42\u5e8f\u5217\u7684label\u90fd\u5206\u522b\u662f0\u548c1":1,"\u5355\u5c42":2,"\u5355\u5c42\u5e8f\u5217":[0,1],"\u5355\u5c42\u5e8f\u5217\u7684\u6570\u636e":1,"\u5355\u5c42\u5e8f\u5217\u7684\u6bcf\u4e2a\u5143\u7d20":0,"\u5355\u5c42\u5e8f\u5217\u7684dataprovider\u5982\u4e0b":1,"\u5355\u5c42\u5e8f\u5217\u76f4\u63a5\u53d6\u4e86\u6700\u540e\u4e00\u4e2a\u5143\u7d20":1,"\u5355\u5c42\u5e8f\u5217\u7b2ci\u4e2a\u5143\u7d20":0,"\u5355\u5c42\u5e8f\u5217\u8fc7\u4e86\u4e00\u4e2amix":1,"\u5355\u5c42\u6216\u53cc\u5c42":0,"\u5355\u5c42rnn":[1,2],"\u5355\u5c42rnn\u793a\u4f8b":15,"\u5355\u673acpu\u8bad\u7ec3":14,"\u5355\u673agpu\u8bad\u7ec3":14,"\u5355\u6d4b\u4e2d":1,"\u5355\u8fdb\u5355\u51fa":2,"\u536b\u751f":1,"\u5373":[9,13,14],"\u5373\u4e00\u4e2a\u53e5\u5b50\u4e00\u4e2alabel":1,"\u5373\u4e00\u4e2a\u5b50\u53e5\u4e00\u4e2alabel":1,"\u5373\u4e0d\u5728\u4e4e\u5185\u5b58\u6682\u5b58\u591a\u5c11\u6761\u6570\u636e":24,"\u5373\u4e0d\u662f\u4e00\u6761\u5e8f\u5217":24,"\u5373\u4ece\u5355\u8bcd\u5b57\u7b26\u4e32\u5230\u5355\u8bcdid\u7684\u5b57\u5178":24,"\u5373\u4f1a\u751f\u6210100\u4e2agener":24,"\u5373\u4f7f\u5728check\u4e2d\u6570\u636e\u4e0d\u5408\u6cd5":24,"\u5373\u4f7f\u5728process\u91cc\u9762\u53ea\u4f1a\u6709\u4e00":24,"\u5373\u5185\u5c42memory\u7684boot":1,"\u5373\u521d\u59cb\u72b6\u6001\u4e3a0":2,"\u5373\u5305\u542b\u65f6\u95f4\u6b65\u4fe1\u606f":24,"\u5373\u53cc\u5c42rnn\u7684\u6bcf\u4e2a\u72b6\u6001":2,"\u5373\u53ef":24,"\u5373\u53ef\u4ee5\u4f7f\u7528ssh\u8bbf\u95ee\u5bbf\u4e3b\u673a\u76848022\u7aef\u53e3":9,"\u5373\u53ef\u4ee5\u6781\u5927\u7684\u52a0\u901f\u6570\u636e\u8f7d\u5165\u6d41\u7a0b":14,"\u5373\u53ef\u4ee5\u8bbe\u7f6e\u53c2\u4e0e\u8bad\u7ec3\u7684\u7ebf\u7a0b\u6570\u91cf":14,"\u5373\u53ef\u542f\u52a8\u548c\u8fdb\u5165paddlepaddle\u7684contain":9,"\u5373\u53ef\u5728\u672c\u5730\u7f16\u8bd1\u51fapaddlepaddle\u7684\u955c\u50cf":3,"\u5373\u53ef\u6253\u5370\u51fapaddlepaddle\u7684\u7248\u672c\u548c\u6784\u5efa":9,"\u5373\u5728\u7f51\u7edc\u914d\u7f6e\u4e2d":14,"\u5373\u5927\u90e8\u5206\u503c\u4e3a0":24,"\u5373\u5982\u679ctrain":24,"\u5373\u5bf9\u7b2c3\u6b65\u8fdb\u884c\u66ff\u6362":13,"\u5373\u628a\u5355\u5c42rnn\u751f\u6210\u540e\u7684subseq\u7ed9\u62fc\u63a5\u6210\u4e00\u4e2a\u65b0\u7684\u53cc\u5c42seq":2,"\u5373\u6574\u4e2a\u53cc\u5c42group\u662f\u5c06\u524d\u4e00\u4e2a\u5b50\u53e5\u7684\u6700\u540e\u4e00\u4e2a\u5411\u91cf":1,"\u5373\u6574\u4e2a\u8f93\u5165\u5e8f\u5217":0,"\u5373\u662f\u4e00\u6761\u65f6\u95f4\u5e8f\u5217":24,"\u5373\u8bbe\u7f6e":14,"\u5373\u8d77\u5230\u7684\u4f5c\u7528\u4ec5\u4ec5\u662f\u628a\u53cc\u5c42seq\u62c6\u6210\u5355\u5c42":1,"\u5373input":2,"\u5373train":24,"\u5377\u79ef\u7f51\u7edc\u662f\u4e00\u79cd\u7279\u6b8a\u7684\u4ece\u8bcd\u5411\u91cf\u8868\u793a\u5230\u53e5\u5b50\u8868\u793a\u7684\u65b9\u6cd5":13,"\u53bb\u8fc7":1,"\u53c2\u6570":3,"\u53c2\u6570\u5171\u4eab\u7684\u914d\u7f6e\u793a\u4f8b\u4e3a":14,"\u53c2\u6570\u6570\u91cf":13,"\u53c2\u6570\u6765\u63a7\u5236\u7f13\u5b58\u65b9\u6cd5":14,"\u53c2\u8003":23,"\u53c2\u89c1":[6,7],"\u53c2\u89c1pydataprovider2":1,"\u53c3\u5dee\u7b49\u7b49":14,"\u53c8":1,"\u53c8\u662f\u4e00\u4e2a\u5355\u5c42\u7684\u5e8f\u5217":0,"\u53c8\u8981\u4fdd\u8bc1\u6570\u636e\u662f\u968f\u673a\u7684":14,"\u53cc\u5c42":2,"\u53cc\u5c42\u5e8f\u5217":[0,1],"\u53cc\u5c42\u5e8f\u5217\u5728\u540c\u6837\u7684mix":1,"\u53cc\u5c42\u5e8f\u5217\u662f\u4e00\u4e2a\u5d4c\u5957\u7684\u5e8f\u5217":0,"\u53cc\u5c42\u5e8f\u5217\u662fpaddlepaddle\u652f\u6301\u7684\u4e00\u79cd\u975e\u5e38\u7075\u6d3b\u7684\u6570\u636e\u7ec4\u7ec7\u65b9\u5f0f":2,"\u53cc\u5c42\u5e8f\u5217\u6bcf\u4e2asubseq\u4e2d\u6bcf\u4e2a\u5143\u7d20":0,"\u53cc\u5c42\u5e8f\u5217\u7684\u6570\u636e":1,"\u53cc\u5c42\u5e8f\u5217\u7684dataprovider\u5982\u4e0b":1,"\u53cc\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u53d8\u6210\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"\u53cc\u5c42\u5e8f\u5217\u9996\u5148":1,"\u53cc\u5c42\u6216\u8005\u5355\u5c42":0,"\u53cc\u5c42rnn":[1,2],"\u53cc\u5c42rnn\u793a\u4f8b":15,"\u53cc\u8fdb\u5355\u51fa":2,"\u53d1\u884c\u7248":10,"\u53d6\u4e86\u6bcf\u4e2asubseq\u7684\u5e73\u5747\u503c":1,"\u53d6\u4e86\u6bcf\u4e2asubseq\u7684\u6700\u540e\u4e00\u4e2a\u5143\u7d20":1,"\u53d6\u51b3\u4e8e\u662f\u5426\u5bfb\u627e\u5230gflags":4,"\u53d6\u51b3\u4e8e\u662f\u5426\u5bfb\u627e\u5230glog":4,"\u53d6\u51b3\u4e8e\u662f\u5426\u5bfb\u627e\u5230gtest":4,"\u53d6\u51b3\u4e8e\u662f\u5426\u627e\u5230swig":4,"\u53d8\u4e3a3\u4e2a\u65b0\u7684\u5b50\u6b65\u9aa4":13,"\u53d8\u4f1a\u62a5\u8fd9\u4e2a\u9519\u8bef":10,"\u53d8\u91cf":24,"\u53e3\u5934":1,"\u53e5\u5b50\u8868\u793a\u7684\u8ba1\u7b97\u66f4\u65b0\u4e3a2\u6b65":13,"\u53e6":14,"\u53e6\u4e00\u4e2a\u662f\u6bcf\u6761\u5e8f\u5217":14,"\u53ea\u4f5c\u4e3aread":2,"\u53ea\u5305\u62ecpaddle\u7684\u4e8c\u8fdb\u5236":9,"\u53ea\u662f\u53cc\u5c42\u5e8f\u5217\u5c06\u5176\u53c8\u505a\u4e86\u5b50\u5e8f\u5217\u5212\u5206":1,"\u53ea\u662f\u5c06\u53e5\u5b50\u5229\u7528\u8fde\u7eed\u5411\u91cf\u8868\u793a\u66ff\u6362\u7a00\u758f":13,"\u53ea\u662f\u8bf4\u660e\u6570\u636e\u7684\u987a\u5e8f\u662f\u91cd\u8981\u7684":24,"\u53ea\u6709":1,"\u53ea\u7528\u4e8e\u5728\u5e8f\u5217\u751f\u6210\u4efb\u52a1\u4e2d\u6307\u5b9a\u8f93\u5165\u6570\u636e":2,"\u53ea\u80fd\u591f\u8fd4\u56delist\u6216\u8005tupl":24,"\u53ea\u80fd\u901a\u8fc7":1,"\u53ea\u8bfbmemory\u8f93\u5165":2,"\u53ea\u9488\u5bf9\u5185\u5b58":14,"\u53ea\u9700\u8981\u4f7f\u7528\u4e00\u884c\u4ee3\u7801\u5373\u53ef\u4ee5\u8bbe\u7f6e\u8bad\u7ec3\u5f15\u7528\u8fd9\u4e2adataprovid":24,"\u53ea\u9700\u8981\u5728":24,"\u53ea\u9700\u8981\u77e5\u9053\u8fd9\u53ea\u662f\u4e00\u4e2a\u6807\u8bb0\u5c5e\u6027\u7684\u65b9\u6cd5\u5c31\u53ef\u4ee5\u4e86":24,"\u53ef\u4ee5":1,"\u53ef\u4ee5\u4e3a\u4e00\u4e2a\u6570\u636e\u6587\u4ef6\u8fd4\u56de\u591a\u6761\u8bad\u7ec3\u6837\u672c":24,"\u53ef\u4ee5\u4f20\u516510k":3,"\u53ef\u4ee5\u4f30\u8ba1\u51fa\u5982\u679c\u6a21\u578b\u91c7\u7528\u4e0d\u53d8\u7684\u8f93\u51fa\u6700\u5c0f\u7684cost0\u662f\u591a\u5c11":14,"\u53ef\u4ee5\u4f7f\u7528":[3,14],"\u53ef\u4ee5\u4f7f\u7528\u547d\u4ee4":10,"\u53ef\u4ee5\u4f7f\u7528\u5982\u4e0b\u4ee3\u7801":14,"\u53ef\u4ee5\u4f7f\u7528\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u8bc4\u4f30\u5e26\u6709label\u7684\u9a8c\u8bc1\u96c6":13,"\u53ef\u4ee5\u4f7f\u7528graphviz\u5bf9paddlepaddle\u7684\u7f51\u7edc\u6a21\u578b\u8fdb\u884c\u7ed8\u5236":17,"\u53ef\u4ee5\u4f7f\u7528paddl":17,"\u53ef\u4ee5\u4f7f\u7528python\u7684":27,"\u53ef\u4ee5\u51cf\u5c11\u7f13\u5b58\u6c60\u7684\u5927\u5c0f":14,"\u53ef\u4ee5\u53c2\u8003":13,"\u53ef\u4ee5\u5728\u4e00\u4e2a\u51fd\u6570\u91cc":24,"\u53ef\u4ee5\u5728cmake\u7684\u547d\u4ee4\u884c\u8bbe\u7f6e":4,"\u53ef\u4ee5\u5c06\u4e00\u6761\u6570\u636e\u8bbe\u7f6e\u6210\u591a\u4e2abatch":24,"\u53ef\u4ee5\u5c06memory\u7406\u89e3\u4e3a\u4e00\u4e2a\u65f6\u5ef6\u64cd\u4f5c":2,"\u53ef\u4ee5\u5c06paddlepaddle\u7684\u6a21\u578b\u548c\u914d\u7f6e\u6253\u5305\u6210\u4e00\u4e2a\u6587\u4ef6":17,"\u53ef\u4ee5\u5c06paddlepaddle\u7684\u8bad\u7ec3\u6a21\u578b\u4ee5proto":17,"\u53ef\u4ee5\u65b9\u4fbf\u5d4c\u5165\u5f0f\u5de5\u4f5c":4,"\u53ef\u4ee5\u662f\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":2,"\u53ef\u4ee5\u662f\u4e00\u4e2a\u975e\u5e8f\u5217":2,"\u53ef\u4ee5\u663e\u793a\u5730\u6307\u5b9a\u4e00\u4e2alayer\u7684\u8f93\u51fa\u7528\u4e8e\u521d\u59cb\u5316memori":2,"\u53ef\u4ee5\u6709\u4ee5\u4e0b\u4e24\u79cd":2,"\u53ef\u4ee5\u6839\u636e\u4e0d\u540c\u7684\u6570\u636e\u914d\u7f6e\u4e0d\u540c\u7684\u8f93\u5165\u7c7b\u578b":24,"\u53ef\u4ee5\u770b\u4f5c\u662f\u4e00\u4e2a\u975e\u5e8f\u5217\u8f93\u5165":0,"\u53ef\u4ee5\u8003\u8651\u4f7f\u7528\u4e00\u4e9b\u4f18\u5316\u7b97\u6cd5":14,"\u53ef\u4ee5\u8fd4\u56de\u4e00\u4e2adict":24,"\u53ef\u4ee5\u901a\u8fc7show":13,"\u53ef\u7528\u5728\u6d4b\u8bd5\u6216\u8bad\u7ec3\u65f6\u6307\u5b9a\u521d\u59cb\u5316\u6a21\u578b":13,"\u53ef\u80fd\u7684\u4ee3\u7801\u4e3a":14,"\u53ef\u80fd\u7684\u5185\u5b58\u6cc4\u9732\u95ee\u9898":23,"\u53ef\u80fd\u7684\u8f93\u51fa\u4e3a":10,"\u53ef\u9009":24,"\u5403":1,"\u5403\u996d":1,"\u5404\u65b9\u9762":1,"\u5404\u79cd\u53c2\u6570\u548c\u7ef4\u62a4":3,"\u5408":1,"\u5408\u7406":1,"\u540c\u65f6":[3,4,14,24],"\u540c\u65f6\u4e5f\u53ef\u4ee5\u52a0\u901f\u5f00\u59cb\u8bad\u7ec3\u524d\u6570\u636e\u8f7d\u5165\u7684\u8fc7\u7a0b":14,"\u540c\u65f6\u4e5f\u80fd\u591f\u5f15\u5165\u66f4\u52a0\u590d\u6742\u7684\u8bb0\u5fc6\u673a\u5236":2,"\u540c\u65f6\u4f1a\u8ba1\u7b97\u5206\u7c7b\u51c6\u786e\u7387":13,"\u540c\u65f6\u5728\u5185\u5b58\u91cc\u76f4\u63a5\u968f\u5373\u9009\u53d6\u6570\u636e\u6765\u505ashuffl":14,"\u540c\u65f6\u5c06\u53c2\u6570\u521d\u59cb\u5316\u4e3a":14,"\u540c\u65f6\u6b22\u8fce\u8d21\u732e\u66f4\u591a\u7684\u5b89\u88c5\u5305":8,"\u540c\u65f6\u8bbe\u7f6e\u5185\u5b58\u7f13\u5b58\u529f\u80fd":14,"\u540c\u6837\u53ef\u4ee5\u6269\u5c55\u5230\u53cc\u5c42\u5e8f\u5217\u7684\u5904\u7406\u4e0a":2,"\u540d\u79f0":13,"\u540e":14,"\u540e\u9762\u8ddf\u7740\u4e00\u7cfb\u5217\u7f16\u8bd1\u53c2\u6570":22,"\u5411\u91cf\u8868\u793a":13,"\u5426":4,"\u5426\u5219":23,"\u5426\u5219\u5728\u7b2c0\u4e2a\u65f6\u95f4\u6b65\u65f6":1,"\u5426\u5219\u9700\u8981\u9009\u62e9\u975eavx\u7684paddlepaddl":9,"\u5440":1,"\u5468\u56f4":1,"\u547d\u4ee4":3,"\u547d\u4ee4\u4e3a":9,"\u547d\u4ee4\u5373\u53ef\u5b8c\u6210\u5b89\u88c5":10,"\u547d\u4ee4\u6307\u5b9a\u7684\u53c2\u6570\u4f1a\u4f20\u5165\u7f51\u7edc\u914d\u7f6e\u4e2d":13,"\u547d\u4ee4\u8fd0\u884c\u955c\u50cf":9,"\u547d\u4ee4\u9884\u5148\u4e0b\u8f7d\u955c\u50cf":9,"\u548c\u4e00\u4e2a\u5df2\u7ecf\u5206\u8bcd\u540e\u7684\u53e5\u5b50":1,"\u548c\u4e09\u79cd\u5e8f\u5217\u6a21\u5f0f":24,"\u548c\u4e4b\u524d\u51cf\u5c0f\u901a\u8fc7\u51cf\u5c0f\u7f13\u5b58\u6c60\u6765\u51cf\u5c0f\u5185\u5b58\u5360\u7528\u7684\u539f\u7406\u4e00\u81f4":14,"\u548c\u53cc\u5c42\u5e8f\u5217\u542b\u6709subseq":0,"\u548c\u53cc\u5c42rnn":1,"\u548c\u5dee\u8bc4":13,"\u548c\u5e8f\u5217\u4e2d\u542b\u6709\u5143\u7d20\u7684\u6570\u76ee\u540c":0,"\u548c\u6587\u672c\u4fe1\u606f\u7528tab\u95f4\u9694":13,"\u548c\u6d4b\u8bd5\u6587\u4ef6\u5217\u8868":23,"\u548c\u7528\u6237\u4f20\u5165\u7684\u53c2\u6570":24,"\u548c\u90e8\u5206layer":2,"\u548c\u9884\u5904\u7406\u811a\u672c":13,"\u548cavgpool":0,"\u548ccudnn":10,"\u548cinitalizer\u91cc\u5b9a\u4e49\u987a\u5e8f\u4e00\u81f4":13,"\u54c1\u8d28":1,"\u5546\u52a1":1,"\u554a":1,"\u5668":13,"\u56db\u4e2a\u7248\u672c":10,"\u56db\u79cd\u6570\u636e\u7c7b\u578b\u662f":24,"\u56e0\u6b64":2,"\u56e0\u6b642\u4e2abatch\u5c31\u53ef\u4ee5\u5b8c\u62101\u4e2apass":1,"\u56e0\u6b64\u4e0a\u8ff0\u4e09\u4e2alayer\u7684\u524d\u5411\u4f1a\u62a5\u51fa":1,"\u56e0\u6b64\u4e24\u4e2a\u914d\u7f6e\u5728\u8fd9\u4e24\u5c42\u4e0a\u7684\u8f93\u51fa\u662f\u4e00\u6837\u7684":1,"\u56e0\u6b64\u5355\u53cc\u5c42\u5e8f\u5217\u7684\u8f93\u51fa\u662f\u4e00\u6837\u65f3":1,"\u56e0\u6b64\u53cc\u5c42\u5e8f\u5217\u7684\u914d\u7f6e\u4e2d":1,"\u56e0\u6b64\u53cc\u5c42\u5e8f\u5217\u8fc7\u5b8clstmemory\u7684\u8f93\u51fa\u548c\u5355\u5c42\u7684\u4e00\u6837":1,"\u56e0\u6b64\u5f53\u5916\u5c42\u6709i":1,"\u56fe\u50cf\u5206\u7c7b":12,"\u5728":[0,1,4,10,24],"\u5728\u4e4b\u540e\u7684":14,"\u5728\u4f7f\u7528":14,"\u5728\u53cd\u5411\u4f20\u9012\u7684\u65f6\u5019":14,"\u5728\u58f0\u660edataprovider\u7684\u65f6\u5019\u4f20\u5165\u4e86dictionary\u4f5c\u4e3a\u53c2\u6570":24,"\u5728\u5b8c\u6210\u4e86\u6570\u636e\u548c\u7f51\u7edc\u7ed3\u6784\u642d\u5efa\u4e4b\u540e":13,"\u5728\u5e8f\u5217\u751f\u6210\u4efb\u52a1\u4e2d":2,"\u5728\u672c\u95ee\u9898\u4e2d":13,"\u5728\u6a21\u578b\u914d\u7f6e\u4e2d\u5229\u7528":13,"\u5728\u6b64\u4e3a\u65b9\u4fbf\u5bf9\u6bd4\u4e0d\u540c\u7f51\u7edc\u7ed3\u6784":13,"\u5728\u6bcf\u4e2a\u7ef4\u5ea6\u4e0a\u53d6\u51fa\u5728\u8be5\u53e5\u8bdd\u65b0\u7684\u5411\u91cf\u96c6\u5408\u4e0a\u8be5\u7ef4\u5ea6\u7684\u6700\u5927\u503c\u4f5c\u4e3a\u6700\u540e\u7684\u53e5\u5b50\u8868\u793a\u5411\u91cf":13,"\u5728\u7a0b\u5e8f\u5f00\u59cb\u9636\u6bb5":27,"\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\u4e2d":0,"\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u8fdb\u884c\u6d4b\u8bd5":23,"\u5728\u8bad\u7ec3\u914d\u7f6e\u91cc":24,"\u5728\u8f93\u51fa\u7684\u8fc7\u7a0b\u4e2d":2,"\u5728\u8fd9\u4e2a\u51fd\u6570\u4e2d":24,"\u5728\u8fd9\u79cd\u7ed3\u6784\u4e2d":2,"\u5728\u8fd9\u91cc":2,"\u5728\u914d\u7f6e\u4e2d\u8bfb\u53d6\u4e86\u5b57\u5178":24,"\u5728\u91c7\u7528sgd":14,"\u5728cmake\u914d\u7f6e\u65f6\u53ef\u4ee5\u4f7f\u7528":4,"\u5728paddlepaddle\u4e2d":2,"\u5728pydataprovider\u4e2d":24,"\u5728python\u73af\u5883\u4e0b\u9884\u6d4b\u7ed3\u679c":27,"\u5728step\u51fd\u6570\u4e2d\u5b9a\u4e49":2,"\u5728step\u51fd\u6570\u4e2d\u5b9a\u4e49memori":2,"\u5730\u6bb5":1,"\u5730\u7406\u4f4d\u7f6e":1,"\u5730\u94c1\u7ad9":1,"\u5747\u5300\u5206\u5e03":14,"\u57fa\u4e8e\u53cc\u5c42\u5e8f\u5217\u8f93\u5165":2,"\u57fa\u672c\u4e0a\u4e0d\u80fd\u6574\u4f53\u4fee\u6b63":24,"\u57fa\u672c\u7684\u5904\u7406\u903b\u8f91\u4e5f\u548cmnist\u903b\u8f91\u4e00\u81f4":24,"\u57fa\u672c\u7684pydataprovider\u4f7f\u7528\u4ecb\u7ecd\u5b8c\u6bd5\u4e86":24,"\u5904\u7406\u7684\u8f93\u5165\u5e8f\u5217\u4e3b\u8981\u5206\u4e3a\u4ee5\u4e0b\u4e09\u79cd\u7c7b\u578b":2,"\u5916\u5c42memory\u5fc5\u987b\u6709boot":1,"\u5916\u5c42memory\u662f\u4e00\u4e2a\u5143\u7d20":1,"\u5916\u5c42memory\u662f\u5355\u5c42\u5e8f\u5217":1,"\u5916\u5c42outer":1,"\u591a\u4e2ainput\u4ee5list\u65b9\u5f0f\u8f93\u5165":13,"\u591a\u53e5\u8bdd\u8fdb\u4e00\u6b65\u6784\u6210\u4e86\u6bb5\u843d":2,"\u591a\u673a\u8bad\u7ec3":14,"\u591a\u6b21\u8fd4\u56de\u53d8\u91cf":24,"\u591a\u7ebf\u7a0b\u4e0b\u8f7d\u8fc7\u7a0b\u4e2d":3,"\u591a\u7ebf\u7a0b\u6570\u636e\u8bfb\u53d6":24,"\u591a\u8f6e\u5bf9\u8bdd\u7b49\u66f4\u4e3a\u590d\u6742\u7684\u8bed\u8a00\u6570\u636e":2,"\u5927":24,"\u5929":1,"\u5929\u4e00\u5e7f\u573a":1,"\u5929\u4e00\u9601":1,"\u597d":1,"\u597d\u5403":1,"\u597d\u8bc4":13,"\u5982\u4e0b":1,"\u5982\u679c":[10,24],"\u5982\u679c\u4e0d\u4e86\u89e3":24,"\u5982\u679c\u4e0d\u4f7f\u7528\u5219\u4f1a\u4f7f\u7528\u4e00\u4e2a\u7b80\u5316\u7248\u7684\u547d\u4ee4\u884c\u53c2\u6570\u89e3\u6790":4,"\u5982\u679c\u4e0d\u4f7f\u7528\u5219\u4f1a\u4f7f\u7528\u4e00\u4e2a\u7b80\u5316\u7248\u7684\u65e5\u5fd7\u5b9e\u73b0":4,"\u5982\u679c\u4e0d\u5207\u8bcd":13,"\u5982\u679c\u4e0d\u6536\u655b":14,"\u5982\u679c\u4e0d\u8bbe\u7f6e\u7684\u8bdd":24,"\u5982\u679c\u4f7f\u7528gpu\u7248\u672c\u7684paddlepaddl":10,"\u5982\u679c\u5185\u5c42memory\u60f3":1,"\u5982\u679c\u53c2\u6570\u4fdd\u5b58\u4e0b\u6765\u7684":14,"\u5982\u679c\u5728":10,"\u5982\u679c\u5728\u7b2c\u4e00\u6b21cmake\u4e4b\u540e\u60f3\u8981\u91cd\u65b0\u8bbe":4,"\u5982\u679c\u5728\u8bad\u7ec3\u65f6":24,"\u5982\u679c\u5c06\u8fd9\u4e2a\u5185\u5b58\u6c60\u51cf\u5c0f":14,"\u5982\u679c\u5c0f\u4e8e\u8fd9\u4e2a\u4e0b\u8f7d\u901f\u5ea6":3,"\u5982\u679c\u60a8\u4f7f\u7528":9,"\u5982\u679c\u60f3\u8981\u5728\u5916\u90e8\u673a\u5668\u8bbf\u95ee\u8fd9\u4e2acontain":9,"\u5982\u679c\u662ffalse\u7684\u8bdd":24,"\u5982\u679c\u662ftrue\u7684\u8bdd":24,"\u5982\u679c\u6709\u591a\u4e2a\u8f93\u5165":2,"\u5982\u679c\u6709\u591a\u4e2a\u8f93\u5165\u5e8f\u5217":2,"\u5982\u679c\u6709\u66f4\u590d\u6742\u7684\u4f7f\u7528":23,"\u5982\u679c\u6ca1\u6709\u5b9a\u4e49memori":2,"\u5982\u679c\u7528\u6237\u4e0d\u6307\u5b9a\u8fd4\u56de\u6570\u636e\u7684\u5bf9\u5e94\u5173\u7cfb":24,"\u5982\u679c\u7528\u6237\u60f3\u8981\u81ea\u5b9a\u4e49\u521d\u59cb\u5316\u65b9\u5f0f":14,"\u5982\u679c\u8bad\u7ec3\u4e00\u4e2apass":14,"\u5982\u679c\u8bad\u7ec3\u8fc7\u7a0b\u7684\u7684cost\u660e\u663e\u9ad8\u4e8e\u8fd9\u4e2a\u5e38\u6570\u8f93\u51fa\u7684cost":14,"\u5982\u679c\u8bbe\u7f6e\u6210true\u7684\u8bdd":24,"\u5982\u679c\u8f93\u51fa":9,"\u5982\u679c\u8fd0\u884cgpu\u7248\u672c\u7684paddlepaddl":9,"\u5982\u679clearning_rate\u592a\u5927":14,"\u5982\u679clearning_rate\u592a\u5c0f":14,"\u5982\u679ctest":23,"\u5b50":1,"\u5b50\u53e5":2,"\u5b50\u53e5\u7684\u5355\u8bcd\u6570\u548c\u6307\u5b9a\u7684\u4e00\u4e2a\u8f93\u5165\u5e8f\u5217\u4e00\u81f4":2,"\u5b81\u6ce2":1,"\u5b83\u5305\u542b\u7684\u53c2\u6570\u6709":24,"\u5b83\u7684":1,"\u5b83\u7684\u6bcf\u4e00\u4e2a\u5143\u7d20":0,"\u5b89\u6392":1,"\u5b89\u88c5\u5305":14,"\u5b89\u88c5\u5305\u5728ubuntu":10,"\u5b89\u88c5\u5305\u7684\u4e0b\u8f7d\u5730\u5740\u662f":10,"\u5b89\u88c5\u597d\u7684paddlepaddle\u811a\u672c\u5305\u62ec\u591a\u6761\u547d\u4ee4":17,"\u5b89\u88c5\u5b8c\u6210\u540e":10,"\u5b89\u88c5\u5b8c\u6210\u7684paddlepaddle\u4e3b\u4f53\u5305\u62ec\u4e09\u4e2a\u90e8\u5206":9,"\u5b89\u88c5\u5b8c\u6210paddlepaddle\u540e":10,"\u5b89\u88c5\u6559\u7a0b":13,"\u5b89\u88c5\u65b9\u6cd5\u8bf7\u53c2\u8003":9,"\u5b89\u88c5\u7f16\u8bd1\u4f9d\u8d56":6,"\u5b89\u88c5\u7f16\u8bd1paddlepaddle\u9700\u8981\u7684\u4f9d\u8d56":5,"\u5b89\u88c5docker\u9700\u8981\u60a8\u7684\u673a\u5668":9,"\u5b89\u88c5paddlepaddl":13,"\u5b89\u88c5paddlepaddle\u7684docker\u955c\u50cf":8,"\u5b89\u9759":1,"\u5b8c\u6210\u4efb\u610f\u7684\u8fd0\u7b97\u903b\u8f91":2,"\u5b8c\u6210\u591a\u673a\u8bad\u7ec3":17,"\u5b8c\u6210\u76f8\u5e94\u7684\u8ba1\u7b97":0,"\u5b8c\u6574\u4ee3\u7801\u89c1":27,"\u5b9a\u4e49\u4e00\u4e2a\u65f6\u95f4\u6b65\u4e4b\u5185rnn\u5355\u5143\u5b8c\u6210\u7684\u8ba1\u7b97":2,"\u5b9a\u4e49\u4e86\u4e00\u4e2a\u53ea\u8bfb\u7684memori":2,"\u5b9a\u4e49\u5728\u5916\u5c42":2,"\u5b9a\u4e49\u6587\u672c\u4fe1\u606f":13,"\u5b9e\u73b0\u4e86\u6253\u5f00\u6587\u672c\u6587\u4ef6":24,"\u5b9e\u73b0\u8bcd\u8bed\u548c\u53e5\u5b50\u4e24\u4e2a\u7ea7\u522b\u7684\u53cc\u5c42rnn\u7ed3\u6784":2,"\u5b9e\u9645\u4e2d\u5e76\u4e0d\u9700\u8981":1,"\u5ba2\u6237":1,"\u5bb6":1,"\u5bc6\u7801\u4e5f\u662froot":9,"\u5bf9":1,"\u5bf9\u4e8e\u7528\u6237\u6765\u8bf4":24,"\u5bf9\u4e8e\u7ed9\u5b9a\u7684\u4e00\u6761\u6587\u672c":13,"\u5bf9\u4e8ecuda\u7684toolkit\u6709\u65ad\u70b9\u7eed\u4f20\u548c\u4f20\u8f93\u901f\u5ea6\u8fc7\u5c0f\u91cd\u542f\u4e0b\u8f7d\u7684":3,"\u5bf9\u4e8emnist\u800c\u8a00":24,"\u5bf9\u5e94\u4e00\u4e2a\u5b50\u53e5":2,"\u5bf9\u5e94\u4e00\u4e2a\u8bcd":2,"\u5bf9\u8be5\u8868\u793a\u8fdb\u884c\u975e\u7ebf\u6027\u53d8\u6362":13,"\u5bf9\u8c61":[14,24],"\u5bf9\u8c61convert":27,"\u5bf9\u8f93\u51fa\u7684\u5408\u5e76":2,"\u5bf9\u9762":1,"\u5bfc\u81f4\u8bad\u7ec3\u65f6\u95f4\u8fc7\u957f":14,"\u5c06":14,"\u5c06\u4f1a\u6d4b\u8bd5\u914d\u7f6e\u6587\u4ef6\u4e2dtest":13,"\u5c06\u5176\u6269\u5c55\u6210\u4e00\u4e2a\u65b0\u7684\u53cc\u5c42\u5e8f\u5217":1,"\u5c06\u5176\u62fc\u63a5\u6210\u4e00\u4e2a\u65b0\u7684\u5355\u5c42\u5e8f\u5217":1,"\u5c06\u5176\u8bbe\u7f6e\u6210":14,"\u5c06\u542b\u6709\u5b50\u53e5":2,"\u5c06\u542b\u6709\u8bcd\u8bed\u7684\u53e5\u5b50\u5b9a\u4e49\u4e3a\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":2,"\u5c06\u5b57\u5178\u5b58\u5165\u4e86set":24,"\u5c06\u5bbf\u4e3b\u673a\u76848022\u7aef\u53e3\u6620\u5c04\u5230container\u768422\u7aef\u53e3\u4e0a":9,"\u5c06\u6570\u636e\u7ec4\u5408\u6210batch\u8bad\u7ec3":24,"\u5c06\u6587\u4ef6\u7684\u7edd\u5bf9\u8def\u5f84\u6216\u76f8\u5bf9\u8def\u5f84":23,"\u5c06\u8bc4\u8bba\u5206\u4e3a\u597d\u8bc4":13,"\u5c06\u8be5\u53e5\u8bdd\u5305\u542b\u7684\u6240\u6709\u5355\u8bcd\u5411\u91cf\u6c42\u5e73\u5747\u5f97\u5230\u53e5\u5b50\u7684\u8868\u793a":13,"\u5c06ssh\u88c5\u5165\u7cfb\u7edf\u5185\u5e76\u5f00\u542f\u8fdc\u7a0b\u8bbf\u95ee":9,"\u5c1a\u53ef":1,"\u5c31":[1,24],"\u5c31\u50cf\u8fd9\u4e2a\u6837\u4f8b\u4e00\u6837":24,"\u5c31\u5f88\u5bb9\u6613\u5bfc\u81f4\u5185\u5b58\u8d85\u9650":14,"\u5c31\u662f":1,"\u5c31\u662f\u5c06\u8fd9\u4e9b\u52a8\u6001\u5e93\u52a0\u5230\u73af\u5883\u53d8\u91cf\u91cc\u9762":10,"\u5c42\u6b21\u5316\u7684rnn":2,"\u5c45\u7136":1,"\u5c5e\u6027":24,"\u5dee\u8bc4":13,"\u5e2e\u52a9\u6211\u4eec\u5b8c\u6210\u5bf9\u8f93\u5165\u5e8f\u5217\u7684\u62c6\u5206":2,"\u5e2e\u52a9\u6211\u4eec\u66f4\u597d\u5730\u63cf\u8ff0\u6bb5\u843d":2,"\u5e2e\u52a9\u6211\u4eec\u6784\u9020\u4e00\u4e9b\u590d\u6742\u7684\u8f93\u5165\u4fe1\u606f":0,"\u5e38\u89c1\u7684\u8f93\u51fa\u683c\u5f0f\u4e3a":22,"\u5e72\u51c0":1,"\u5e76\u4e14":24,"\u5e76\u4e14\u4f7f\u7528\u5173\u952e\u8bcd":24,"\u5e76\u4e14\u5220\u9664container\u4e2d\u7684\u6570\u636e":9,"\u5e76\u4e14\u53ef\u80fd\u4f1a\u52a0\u901f\u8bad\u7ec3\u8fc7\u7a0b":14,"\u5e76\u4e14\u5728\u5185\u5b58\u8db3\u591f":24,"\u5e76\u4e14\u6807\u8bb0process\u51fd\u6570\u662f\u4e00\u4e2adataprovid":24,"\u5e76\u4e14softmax\u5c42\u7684\u4e24\u4e2a\u8f93\u5165\u4e5f\u4f7f\u7528\u4e86\u540c\u6837\u7684\u53c2\u6570":14,"\u5e76\u4f7f\u7528\u4e86dropout":13,"\u5e76\u572823\u884c\u8fd4\u56de\u7ed9paddlepaddle\u8fdb\u7a0b":24,"\u5e76\u5bf9\u5176\u8be6\u7ec6\u5206\u6790":1,"\u5e76\u5c06\u6bcf\u884c\u8f6c\u6362\u6210\u548c":24,"\u5e76\u63d0\u4f9b":9,"\u5e76\u63d0\u4f9b\u4e86\u7b80\u5355\u7684cache\u529f\u80fd":24,"\u5e76\u8bbe\u7f6e\u597d\u5bf9\u5e94\u7684\u73af\u5883\u53d8\u91cf":10,"\u5e76\u9010\u6e10\u5c55\u793a\u66f4\u52a0\u6df1\u5165\u7684\u529f\u80fd":13,"\u5e8a\u4e0a\u7528\u54c1":1,"\u5e8a\u57ab":1,"\u5e8f\u5217\u4e2d\u542b\u6709\u5143\u7d20\u7684\u6570\u76ee\u540clayer2\u4e00\u81f4":0,"\u5e8f\u5217\u6570\u636e\u548c\u4e0a\u9762\u7684\u5b8c\u5168\u4e00\u6837":1,"\u5e8f\u5217\u6570\u636e\u662f\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\u9762\u5bf9\u7684\u4e00\u79cd\u4e3b\u8981\u8f93\u5165\u6570\u636e\u7c7b\u578b":2,"\u5e8f\u5217\u662f\u4e00\u79cd\u5e38\u89c1\u7684\u6570\u636e\u7c7b\u578b":0,"\u5e8f\u5217\u6a21\u578b\u6570\u636e\u63d0\u4f9b":23,"\u5e8f\u5217\u6a21\u578b\u662f\u6307\u6570\u636e\u7684\u67d0\u4e00\u7ef4\u5ea6\u662f\u4e00\u4e2a\u5e8f\u5217\u5f62\u5f0f":24,"\u5e8f\u5217\u751f\u6210\u4efb\u52a1\u5927\u591a\u9075\u5faaencod":2,"\u5e8f\u5217\u751f\u6210\u4efb\u52a1\u7684\u8f93\u5165":2,"\u5e8f\u5217\u7684\u6bcf\u4e2a\u5143\u7d20\u662f\u539f\u6765\u53cc\u5c42\u5e8f\u5217\u6bcf\u4e2asubseq\u5143\u7d20\u7684\u5e73\u5747\u503c":0,"\u5e93\u7684\u8bdd":10,"\u5e94\u8be5":1,"\u5e94\u8be5\u964d\u4f4e\u5b66\u4e60\u7387":14,"\u5f00\u542fgpu\u8bad\u7ec3":14,"\u5f0f":24,"\u5f15\u7528\u7684dataprovider\u662f":24,"\u5f15\u7528memory\u5f97\u5230\u8fd9layer\u4e0a\u4e00\u65f6\u523b\u8f93\u51fa":2,"\u5f3a\u70c8\u63a8\u8350":1,"\u5f53\u51fd\u6570\u8fd4\u56de\u7684\u65f6\u5019":24,"\u5f53\u524d\u7684\u8f93\u5165y\u548c\u4e0a\u4e00\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51farnn":1,"\u5f53\u524dlog_period\u4e2abatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747\u5206\u7c7b\u9519\u8bef\u7387":13,"\u5f53\u524dlog_period\u4e2abatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747cost":13,"\u5f53\u7136":23,"\u5f53\u8c03":24,"\u5f62\u6210recurr":2,"\u5f62\u6210recurrent\u8fde\u63a5":2,"\u5f88":[1,13],"\u5f88\u591a":1,"\u5f88\u5b89\u9759":1,"\u5f88\u5e72\u51c0":1,"\u5f88\u65b9\u4fbf":1,"\u5f97":1,"\u5f97\u5230\u7ed3\u679c":10,"\u5faa\u73af\u5c55\u5f00\u7684\u6bcf\u4e2a\u65f6\u95f4\u6b65\u603b\u662f\u80fd\u591f\u5f15\u7528\u6240\u6709\u8f93\u5165":2,"\u5fc5\u987b\u5c06\u524d\u4e00\u4e2a\u5b50\u53e5\u7684\u6700\u540e\u4e00\u4e2a\u5143\u7d20":1,"\u5fc5\u987b\u6307\u5411\u4e00\u4e2apaddlepaddle\u5b9a\u4e49\u7684lay":2,"\u5fc5\u987b\u662f\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":2,"\u5fc5\u987b\u662f\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":0,"\u5feb":1,"\u5feb\u901f\u5165\u95e8":15,"\u5ff5\u662f":24,"\u6027\u4ef7\u6bd4":1,"\u603b\u4f53\u6765\u8bf4":1,"\u60a8\u4e5f\u53ef\u4ee5\u91c7\u7528\u522b\u7684\u7ec4\u7ec7\u5f62\u5f0f":1,"\u60a8\u53ef\u4ee5\u4f7f\u7528":9,"\u60a8\u5c31\u53ef\u4ee5\u8fdc\u7a0b\u7684\u4f7f\u7528paddlepaddle\u5566":9,"\u60a8\u9700\u8981\u5728\u673a\u5668\u4e2d\u5b89\u88c5\u597ddocker":9,"\u60a8\u9700\u8981\u8fdb\u5165\u955c\u50cf\u8fd0\u884cpaddlepaddl":9,"\u60c5\u611f\u5206\u6790":12,"\u60f3\u8981\u8fd0\u884cpaddlepaddl":9,"\u611f\u89c9":1,"\u6210\u4e3a\u7ef4\u5ea6\u4e3ahidden":13,"\u6211\u4eec\u4ece\u63d0\u524d\u7ed9\u5b9a\u7684\u7c7b\u522b\u96c6\u5408\u4e2d\u9009\u62e9\u5176\u6240\u5c5e\u7c7b":13,"\u6211\u4eec\u4ee5\u6587\u672c\u5206\u7c7b\u95ee\u9898\u4f5c\u4e3a\u80cc\u666f":13,"\u6211\u4eec\u4f7f\u7528":13,"\u6211\u4eec\u53ef\u4ee5\u6309\u7167\u5982\u4e0b\u5c42\u6b21\u5b9a\u4e49\u975e\u5e8f\u5217":0,"\u6211\u4eec\u53ef\u4ee5\u8bbe\u8ba1\u642d\u5efa\u4e00\u4e2a\u7075\u6d3b\u7684":2,"\u6211\u4eec\u5728":1,"\u6211\u4eec\u5728\u6b64\u603b":13,"\u6211\u4eec\u5c06\u4ee5\u57fa\u672c\u7684\u903b\u8f91\u56de\u5f52\u7f51\u7edc\u4f5c\u4e3a\u8d77\u70b9":13,"\u6211\u4eec\u5c06\u5728\u540e\u9762\u4ecb\u7ecd\u8bad\u7ec3\u548c\u9884\u6d4b\u7684\u6d41\u7a0b\u7684\u811a\u672c":13,"\u6211\u4eec\u5c06\u8bad\u7ec3\u7684\u547d\u4ee4\u884c\u4fdd\u5b58\u5728\u4e86":13,"\u6211\u4eec\u63a8\u8350\u4f7f\u7528docker\u955c\u50cf\u6765\u90e8\u7f72\u73af\u5883":8,"\u6211\u4eec\u63d0\u4f9b\u4e8612\u4e2a":9,"\u6211\u4eec\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5de5\u5177\u7c7bdataproviderconvert":27,"\u6211\u4eec\u770b\u4e00\u4e0b\u5355\u53cc\u5c42\u5e8f\u5217\u7684\u4e0d\u540c\u6570\u636e\u7ec4\u7ec7\u5f62\u5f0f":1,"\u6211\u4eec\u770b\u4e00\u4e0b\u5355\u53cc\u5c42\u5e8f\u5217\u7684\u4e0d\u540c\u6570\u636e\u7ec4\u7ec7\u5f62\u5f0f\u548cdataprovid":1,"\u6211\u4eec\u770b\u4e00\u4e0b\u5355\u53cc\u5c42\u5e8f\u5217\u7684\u4e0d\u540cdataprovid":1,"\u6211\u4eec\u770b\u4e00\u4e0b\u5355\u5c42\u5e8f\u5217\u7684\u914d\u7f6e":1,"\u6211\u4eec\u770b\u4e00\u4e0b\u8bed\u4e49\u76f8\u540c\u7684\u53cc\u5c42\u5e8f\u5217\u914d\u7f6e":1,"\u6211\u4eec\u79f0\u4e4b\u4e3a\u4e00\u4e2a0\u5c42\u7684\u5e8f\u5217":0,"\u6211\u4eec\u8fdb\u5165\u5230\u8bad\u7ec3\u90e8\u5206":13,"\u6211\u4eec\u9009\u53d6\u5355\u53cc\u5c42\u5e8f\u5217\u914d\u7f6e\u4e2d\u7684\u4e0d\u540c\u90e8\u5206":1,"\u6211\u4eec\u91c7\u7528\u5355\u5c42lstm\u6a21\u578b":13,"\u6211\u4eec\u968f\u65f6\u603b\u7ed3\u4e86\u5404\u4e2a\u7f51\u7edc\u7684\u590d\u6742\u5ea6\u548c\u6548\u679c":13,"\u6216":1,"\u6216\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u6216\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u53d8\u6210\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"\u6216\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":0,"\u6216\u4e00\u4e2a\u5411\u91cf":2,"\u6216\u5176\u4ed6":13,"\u6216\u5355\u5c42\u5e8f\u5217":0,"\u6216\u5355\u5c42\u5e8f\u5217\u7ecf\u8fc7\u8fd0\u7b97\u53d8\u6210\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"\u6216\u6700\u5927\u503c":0,"\u6216\u7b2c\u4e00\u4e2a":0,"\u6216\u7b2c\u4e00\u4e2a\u5143\u7d20":0,"\u6216\u8005":[0,9,14],"\u6216\u800510g\u8fd9\u6837\u7684\u5355\u4f4d":3,"\u6216\u8005\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":2,"\u6216\u8005\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":[0,2],"\u6216\u8005\u4f7f\u7528\u4e0b\u9762\u4e00\u6761\u547d\u4ee4\u5b89\u88c5":10,"\u6216\u8005\u5728python":24,"\u6216\u8005\u6570\u636e\u5e93\u8fde\u63a5\u5730\u5740\u7b49\u7b49":23,"\u6216\u8005\u662f\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":0,"\u6216\u8005\u66f4\u65e9":14,"\u6216\u8005\u76f4\u63a5\u6254\u6389\u975e\u5e38\u957f\u7684\u5e8f\u5217":14,"\u6216\u8005\u8bbe\u7f6e\u4e3anone":23,"\u6216\u8005\u9700\u8981\u66f4\u9ad8\u7684\u6548\u7387":23,"\u6216\u8005\u9ad8\u6027\u80fd\u7684":9,"\u623f":1,"\u623f\u95f4":1,"\u6240\u4ee5":[14,24,27],"\u6240\u4ee5\u5728cpu\u7684\u8fd0\u7b97\u6027\u80fd\u4e0a\u5e76\u4e0d\u4f1a\u6709\u4e25\u91cd\u7684\u5f71\u54cd":9,"\u6240\u4ee5\u5982\u679c\u5bf9\u4e8e\u5185\u5b58\u6bd4\u8f83\u5c0f\u7684\u673a\u5668":24,"\u6240\u4ee5\u5982\u679c\u60f3\u8981\u5728\u540e\u53f0\u542f\u7528ssh":9,"\u6240\u4ee5\u5c06":24,"\u6240\u4ee5\u63a8\u8350\u4f7f\u7528\u663e\u5f0f\u6307\u5b9a\u8fd4\u56de\u503c\u548c\u6570\u636e\u5bf9\u5e94\u5173\u7cfb":24,"\u6240\u4ee5\u6700\u4f73\u5b9e\u8df5\u63a8\u8350\u4e0d\u8981\u5c06\u6bcf\u4e00\u4e2a\u6837\u672c\u90fd\u653e\u5165train":24,"\u6240\u4ee5\u7528\u4e8e\u9884\u6d4b\u7684\u914d\u7f6e\u6587\u4ef6\u8981\u505a\u76f8\u5e94\u7684\u4fee\u6539":27,"\u6240\u4ee5\u8f93\u51fa\u7684value\u5305\u542b\u4e24\u4e2a\u5411\u91cf":27,"\u6240\u4ee5gpu\u5728\u8fd0\u7b97\u6027\u80fd\u4e0a\u4e5f\u4e0d\u4f1a\u6709\u4e25\u91cd\u7684\u5f71\u54cd":9,"\u6240\u4ee5init_hook\u5c3d\u91cf\u4f7f\u7528":24,"\u6240\u6709\u5b57\u7b26\u90fd\u5c06\u8f6c\u6362\u4e3a\u8fde\u7eed\u6574\u6570\u8868\u793a\u7684id\u4f20\u7ed9\u6a21\u578b":13,"\u6240\u6709\u6587\u4ef6\u5217\u8868":24,"\u6240\u6709\u7684":4,"\u6240\u6709\u7684\u4e0b\u8f7d\u7ebf\u7a0b\u5173\u95ed\u65f6":3,"\u6240\u6709\u914d\u7f6e\u5728":13,"\u6240\u8c13\u65f6\u95f4\u6b65\u4fe1\u606f":24,"\u624d\u4f1a\u91ca\u653e\u8be5\u6bb5\u5185\u5b58":24,"\u624d\u4f1astop":24,"\u624d\u80fd\u4fdd\u8bc1\u548c\u5355\u5c42\u5e8f\u5217\u7684\u914d\u7f6e\u4e2d":1,"\u6253\u5370\u7684\u65e5\u5fd7\u53d8\u591a":4,"\u6267\u884c":3,"\u6267\u884c\u5982\u4e0b\u547d\u4ee4\u5373\u53ef\u4ee5\u5173\u95ed\u8fd9\u4e2acontain":9,"\u6267\u884c\u65b9\u6cd5\u5982\u4e0b":9,"\u62a5\u9519":10,"\u62c6\u89e3":2,"\u62fc\u63a5\u6210\u4e00\u4e2a\u65b0\u7684\u5411\u91cf\u8868\u793a":13,"\u6307\u4ee4\u96c6":9,"\u6307\u5411\u4e00\u4e2alayer":2,"\u6307\u5b9a":2,"\u6307\u5b9a\u521d\u59cb\u5316\u6a21\u578b\u8def\u5f84":13,"\u6307\u5b9a\u663e\u5361\u6570\u91cf":14,"\u6307\u5b9a\u751f\u6210\u6570\u636e\u7684\u51fd\u6570":13,"\u6307\u5b9a\u7684\u8f93\u5165\u4e0d\u4f1a\u88ab":2,"\u6307\u5b9a\u8bad\u7ec3":13,"\u6307\u5b9abatch":13,"\u6307\u5b9aoutputs\u8f93\u51fa\u6982\u7387\u5c42":13,"\u633a":1,"\u633a\u597d":1,"\u6362":1,"\u6389\u7f16\u8bd1\u76ee\u5f55\u540e":4,"\u6392\u6210\u4e00\u5217\u7684\u591a\u4e2a\u5143\u7d20":0,"\u63a5\u4e0b\u6765\u4f7f\u7528":27,"\u63a5\u53e3\u4f7f\u7528\u591a\u7ebf\u7a0b\u8bfb\u53d6\u6570\u636e":24,"\u63a5\u53e3\u6709\u4e00\u4e2a":14,"\u63a5\u7740":1,"\u63a8\u8350":1,"\u63a8\u8350\u4f7f\u7528\u5c06\u672c\u5730\u7f51\u5361":9,"\u63a8\u8350\u4f7f\u7528\u6700\u65b0\u7248\u672c\u7684cudnn":4,"\u63a8\u8350\u6e05\u7406":4,"\u63a8\u8350\u76f4\u63a5\u653e\u7f6e\u5230\u8bad\u7ec3\u76ee\u5f55":23,"\u63a8\u8350\u8bbe\u7f6e":24,"\u63cf\u8ff0":4,"\u63cf\u8ff0\u4e86docker":3,"\u63d0\u4f9b\u6269\u5c55\u7684\u957f\u5ea6\u4fe1\u606f":0,"\u653e\u5fc3":1,"\u6548\u679c\u4e00\u81f4":24,"\u6548\u679c\u603b\u7ed3":13,"\u6559\u7a0b":13,"\u6570":2,"\u6570\u5fc5\u987b\u4e25\u683c\u76f8\u7b49":2,"\u6570\u636e":24,"\u6570\u636e\u4e2d0":14,"\u6570\u636e\u4f20\u8f93\u65e0\u9700label\u6570\u636e":13,"\u6570\u636e\u5904\u7406python\u6587\u4ef6\u540d":13,"\u6570\u636e\u5982\u4f55\u5b58\u50a8\u7b49\u7b49":24,"\u6570\u636e\u63d0\u4f9b":23,"\u6570\u636e\u6587\u4ef6\u5b58\u653e\u5728\u672c\u5730\u78c1\u76d8\u4e2d":23,"\u6570\u636e\u662f\u7ed9\u4e00\u6bb5\u82f1\u6587\u6587\u672c":24,"\u6570\u636e\u683c\u5f0f\u548c\u8be6\u7ec6\u6587\u6863\u8bf7\u53c2\u8003":13,"\u6570\u636e\u8f93\u5165":2,"\u6574\u4f53":1,"\u6574\u6d01":1,"\u6587\u4ef6":24,"\u6587\u4ef6\u4e2d":13,"\u6587\u4ef6\u4e3a":14,"\u6587\u672c\u4e2d\u7684\u5355\u8bcd\u7528\u7a7a\u683c\u5206\u9694":13,"\u6587\u672c\u4fe1\u606f\u5c31\u662f\u4e00\u4e2a\u5e8f\u5217":24,"\u6587\u672c\u5206\u7c7b\u95ee\u9898":13,"\u6587\u672c\u5377\u79ef\u5206\u4e3a\u4e09\u4e2a\u6b65\u9aa4":13,"\u6587\u672c\u751f\u6210":12,"\u6587\u6863":14,"\u65b0":1,"\u65b0\u5199layer":15,"\u65b9\u4fbf":1,"\u65b9\u4fbf\u8c03\u8bd5\u4f7f\u7528":17,"\u65b9\u4fbf\u90e8\u7f72\u5206\u53d1":17,"\u65c1\u8fb9":1,"\u65e0":1,"\u65e0\u6cd5\u76f4\u63a5\u4f7f\u7528":1,"\u65e0\u9700label\u76f8\u5173\u7684\u5c42":13,"\u65e9\u9910":1,"\u65f6":[0,14],"\u65f6\u5019":1,"\u65f6\u5e8f\u6a21\u578b\u5373\u4e3arnn\u6a21\u578b":13,"\u65f6\u5e8f\u6a21\u578b\u5747\u4f7f\u7528\u8be5\u811a\u672c":13,"\u662f":[1,4],"\u662f\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u662f\u4e00\u4e2a\u53cc\u5c42\u7684\u5e8f\u5217":0,"\u662f\u4e00\u4e2abatch":24,"\u662f\u4e00\u4e2apython\u7684":24,"\u662f\u4e00\u4e2aswig\u5c01\u88c5\u7684paddlepaddle\u5305":9,"\u662f\u4e00\u4e2aunbound":2,"\u662f\u4e00\u79cd\u4efb\u610f\u590d\u6742\u7684rnn\u5355\u5143":2,"\u662f\u4e0d\u662f\u5f88\u7b80\u5355\u5462":24,"\u662f\u4e2adataprovider\u662f\u4e0d\u662f\u8981\u505ashuffl":24,"\u662f\u4ec0\u4e48\u4e5f\u6ca1\u5173\u7cfb":24,"\u662f\u4ece\u8bad\u7ec3\u914d\u7f6e\u4f20\u5165\u7684dict\u5bf9\u8c61":24,"\u662f\u51e0\u4e4e\u4e0d\u5360\u5185\u5b58\u7684":24,"\u662f\u521d\u59cb\u5316\u65f6\u8c03\u7528\u7684\u51fd\u6570":24,"\u662f\u540c\u4e00\u4e2a\u5bf9\u8c61":24,"\u662f\u5426\u4ee5\u9006\u5e8f\u5904\u7406\u8f93\u5165\u5e8f\u5217":2,"\u662f\u5426\u4f7f\u7528\u53cc\u7cbe\u5ea6\u6d6e\u70b9\u6570":4,"\u662f\u5426\u4f7f\u7528\u8fd0\u884c\u65f6\u52a8\u6001\u52a0\u8f7dcuda\u52a8\u6001\u5e93":4,"\u662f\u5426\u4f7f\u7528gflags":4,"\u662f\u5426\u4f7f\u7528glog":4,"\u662f\u5426\u5185\u5d4cpython\u89e3\u91ca\u5668":4,"\u662f\u5426\u5bfb\u627e\u5230cuda\u5de5\u5177\u94fe":4,"\u662f\u5426\u5f00\u542f\u5355\u5143\u6d4b\u8bd5":4,"\u662f\u5426\u5f00\u542f\u8ba1\u65f6\u529f\u80fd\u5f00\u542f\u8ba1\u65f6\u529f\u80fd\u4f1a\u5bfc\u81f4\u8fd0\u884c\u7565\u6162":4,"\u662f\u5426\u5f00\u542fgpu\u529f\u80fd":3,"\u662f\u5426\u5f00\u542frdma\u652f\u6301":4,"\u662f\u5426\u7f16\u8bd1\u4e2d\u6587\u6587\u6863":4,"\u662f\u5426\u7f16\u8bd1\u542b\u6709avx\u6307\u4ee4\u96c6\u7684paddlepaddle\u4e8c\u8fdb\u5236":4,"\u662f\u5426\u7f16\u8bd1\u65f6\u8fdb\u884c\u4ee3\u7801\u98ce\u683c\u68c0\u67e5":4,"\u662f\u5426\u7f16\u8bd1\u82f1\u6587\u6587\u6863":4,"\u662f\u5426\u7f16\u8bd1gpu\u652f\u6301":4,"\u662f\u5426\u7f16\u8bd1python\u7684swig\u63a5\u53e3":4,"\u662f\u5728\u8fd0\u884c\u65f6\u6267\u884c\u7684":24,"\u662f\u5f85\u6269\u5c55\u7684\u6570\u636e":0,"\u662f\u60f3\u8981\u5171\u4eab\u7684\u53c2\u6570\u4f7f\u7528\u540c\u6837\u7684":14,"\u662f\u6570\u636e\u7f13\u5b58\u7684\u7b56\u7565":24,"\u662f\u6570\u636e\u8f93\u5165\u683c\u5f0f":24,"\u662f\u8bbe\u7f6e\u8fd9\u4e2adataprovider\u8fd4\u56de\u4ec0\u4e48\u6837\u7684\u6570\u636e":24,"\u662f\u8bbe\u7f6edataprovider\u5728\u5185\u5b58\u4e2d\u6682\u5b58\u7684\u6570\u636e\u6761\u6570":24,"\u662f\u8bbe\u7f6edataprovider\u5728\u5185\u5b58\u4e2d\u6700\u5c0f\u6682\u5b58\u7684\u6570\u636e\u6761\u6570":24,"\u662fdecoder\u7684\u6570\u636e\u8f93\u5165":2,"\u662fpaddlepaddle\u652f\u6301\u7684\u4e00\u79cd\u4efb\u610f\u590d\u6742\u7684rnn\u5355\u5143":2,"\u662fpaddlepaddle\u8d1f\u8d23\u63d0\u4f9b\u6570\u636e\u7684\u6a21\u5757":23,"\u662fpython\u7684\u4e00\u4e2a\u5173\u952e\u8bcd":24,"\u663e":13,"\u665a":1,"\u666e\u901a\u7528\u6237\u8bf7\u8d70\u5b89\u88c5\u6d41\u7a0b":8,"\u66f4\u591a\u5173\u4e8esparse\u8bad\u7ec3\u7684\u5185\u5bb9\u8bf7\u53c2\u8003":14,"\u66f4\u597d\u5730\u5b8c\u6210\u4e00\u4e9b\u590d\u6742\u7684\u8bed\u8a00\u7406\u89e3\u4efb\u52a1":2,"\u66f4\u65b0\u6a21\u5f0f":14,"\u66f4\u65b9\u4fbf\u7684\u8bbe\u7f6e\u65b9\u5f0f":14,"\u66f4\u8be6\u7ec6\u7528\u4f8b\u8bf7\u53c2\u8003\u6587\u6863":13,"\u66f4\u8be6\u7ec6\u7684\u4ecb\u7ecd\u8bf7\u53c2\u8003\u5404\u4e2a\u547d\u4ee4\u7684\u547d\u4ee4\u884c\u53c2\u6570\u6587\u6863":17,"\u66f4\u8be6\u7ec6\u7684\u7f51\u7edc\u914d\u7f6e":13,"\u66f4\u8fdb\u4e00\u6b65":2,"\u66ff\u6211\u4eec\u5b8c\u6210\u4e86\u539f\u59cb\u8f93\u5165\u6570\u636e\u7684\u62c6\u5206":2,"\u6700":1,"\u6700\u4f4e\u7ebf\u7a0b\u7684\u4e0b\u8f7d\u901f\u5ea6":3,"\u6700\u540e":1,"\u6700\u540e\u4e00\u4e2a":0,"\u6700\u540e\u4f7f\u7528":27,"\u6700\u7ec8\u5b9e\u73b0\u4e00\u4e2a\u5c42\u6b21\u5316\u7684\u590d\u6742rnn":2,"\u6700\u7ec8\u7684\u8f93\u51fa\u7ed3\u679c":2,"\u6708\u6e56":1,"\u6709":1,"\u6709100\u4e2a\u8bad\u7ec3\u6587\u4ef6":24,"\u6709\u4e24\u53e5":1,"\u6709\u503c\u7684\u4f4d\u7f6e\u53ea\u80fd\u53d61":24,"\u6709\u503c\u7684\u90e8\u5206\u53ef\u4ee5\u662f\u4efb\u4f55\u6d6e\u70b9\u6570":24,"\u6709\u90e8\u5206\u53c2\u6570\u662fpaddle\u81ea\u52a8\u751f\u6210\u7684":24,"\u670d\u52a1":1,"\u670d\u52a1\u5458":1,"\u672c\u6765":1,"\u672c\u8282\u6211\u4eec\u5c06\u4e13\u6ce8\u4e8e\u7f51\u7edc\u7ed3\u6784\u7684\u4ecb\u7ecd":13,"\u6765":1,"\u6765\u521d\u59cb\u5316\u53c2\u6570":14,"\u6765\u5b89\u88c5":10,"\u6765\u5bf9\u6bd4\u5206\u6790\u4e24\u8005\u8bed\u4e49\u76f8\u540c\u7684\u539f\u56e0":1,"\u6765\u5f15\u7528\u8fd9\u4e2aimag":9,"\u6765\u6307\u5b9a\u6bcf\u4e2apserver\u7684ip\u5730\u5740":14,"\u6765\u63a5\u53d7\u4e0d\u4f7f\u7528\u7684":24,"\u6765\u786e\u5b9a\u5bf9\u5e94\u5173\u7cfb":24,"\u6765\u81ea\u5b9a\u4e49\u4f20\u6570\u636e\u7684\u8fc7\u7a0b":23,"\u6765\u8bbe\u7f6e":14,"\u6765\u8bf4\u660e\u7b80\u5355\u7684pydataprovider\u5982\u4f55\u4f7f\u7528":24,"\u6765\u8fdb\u884c\u8bad\u7ec3":9,"\u6765\u914d\u7f6ecudnn\u7684\u5b89\u88c5\u8def\u5f84":4,"\u6765\u9884\u6d4b\u8fd9\u4e2a\u4e2d\u95f4\u7684\u8bcd":14,"\u676f\u5b50":1,"\u6784\u6210\u4e86\u8f93\u51fa\u53cc\u5c42\u5e8f\u5217\u7684\u7b2ci\u4e2asubseq":0,"\u6784\u9020gradientmachin":27,"\u6790\u597d\u7684\u914d\u7f6e\u521b\u5efa\u795e\u7ecf\u7f51\u7edc":27,"\u67e5\u770b\u5b89\u88c5\u540e\u7684paddl":10,"\u6807\u51c6\u5dee\u4e3a":14,"\u6807\u7b7e\u662f0":24,"\u6837\u4f8b\u6570\u636e\u4e3a":24,"\u6837\u4f8b\u6570\u636e\u5982\u4e0b":24,"\u6837\u672c":24,"\u6837\u672c\u95f4\u7528\u7a7a\u884c\u5206\u5f00":1,"\u6839\u636e\u4e0a\u4e00\u6b65\u89e3":27,"\u6839\u636e\u6a21\u578b\u914d\u7f6e\u6587\u4ef6\u4e2d":24,"\u683c\u5f0f\u5982\u4e0b":13,"\u68d2":13,"\u697c\u5c42":1,"\u6a21\u578b\u5b58\u50a8\u8def\u5f84":13,"\u6a21\u578b\u8bad\u7ec3\u4f1a\u770b\u5230\u8fd9\u6837\u7684\u65e5\u5fd7":13,"\u6a21\u578b\u914d\u7f6e":15,"\u6a2a\u5411\u5305\u62ec\u4e09\u4e2a\u7248\u672c":9,"\u6b21":1,"\u6b63\u5e38\u7684docker":9,"\u6b63\u6837\u672c":13,"\u6b64\u5904\u90fd\u4e3a2":1,"\u6bb5\u843d\u53ef\u4ee5\u770b\u4f5c\u662f\u4e00\u4e2a\u5d4c\u5957\u7684\u53cc\u5c42\u7684\u5e8f\u5217":2,"\u6bcf\u4e00\u4e2a\u4efb\u52a1\u6d41\u7a0b\u90fd\u53ef\u4ee5\u5206\u4e3a\u5982\u4e0b5\u4e2a\u57fa\u7840\u90e8\u5206":13,"\u6bcf\u4e00\u4e2a\u65f6\u95f4\u6b65":1,"\u6bcf\u4e00\u4e2a\u65f6\u95f4\u6b65\u90fd\u7528\u4e86\u4e0a\u4e00\u4e2a\u65f6\u95f4\u6b65\u7684\u8f93\u51fa\u7ed3\u679c":1,"\u6bcf\u4e00\u6761\u8bad\u7ec3\u6570\u636e\u90fd\u662f\u4e00\u4e2a\u6587\u4ef6":24,"\u6bcf\u4e00\u884c":24,"\u6bcf\u4e2a\u5143\u7d20\u662f\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"\u6bcf\u4e2a\u5143\u7d20\u662f\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u6bcf\u4e2a\u5355\u5c42rnn":2,"\u6bcf\u4e2a\u5c42\u90fd\u6709\u4e00\u4e2a\u6216\u591a\u4e2ainput":13,"\u6bcf\u4e2a\u65f6\u95f4\u6b65\u4e4b\u5185\u7684\u8fd0\u7b97\u662f\u72ec\u7acb\u7684":2,"\u6bcf\u4e2a\u6837\u672c\u7531\u4e24\u90e8\u5206\u7ec4\u6210":1,"\u6bcf\u4e2a\u6837\u672c\u7684\u5b50\u53e5\u6570\u5206\u522b\u4e3a2":1,"\u6bcf\u4e2a\u72b6\u6001":2,"\u6bcf\u4e2agenerator\u5728\u6ca1\u6709\u8c03\u7528\u7684\u65f6\u5019":24,"\u6bcf\u4e2apass\u7684\u7b2c0\u4e2abatch\u5230\u5f53\u524dbatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747\u5206\u7c7b\u9519\u8bef\u7387":13,"\u6bcf\u4e2apass\u7684\u7b2c0\u4e2abatch\u5230\u5f53\u524dbatch\u6240\u6709\u6837\u672c\u7684\u5e73\u5747cost":13,"\u6bcf\u4e2asubseq\u7684\u6700\u540e\u4e00\u4e2a\u5143\u7d20\u5c31\u7b49\u4e8e\u5355\u5c42\u5e8f\u5217\u7684\u6700\u540e\u4e00\u4e2a\u5143\u7d20":1,"\u6bcf\u6b21\u90fd\u4f1a\u4ecepython\u7aef\u8bfb\u53d6\u6570\u636e":24,"\u6bcf\u884c\u4fdd\u5b58\u4e00\u6761\u6837\u672c":13,"\u6bcf\u9694\u591a\u5c11batch\u6253\u5370\u4e00\u6b21\u65e5\u5fd7":13,"\u6bd4\u5982":14,"\u6bd4\u5982\u8bbe\u7f6e\u4e00\u4e2a\u5168\u8fde\u63a5\u5c42\u7684\u53c2\u6570\u521d\u59cb\u5316\u65b9\u5f0f\u548cbias\u521d\u59cb\u5316\u65b9\u5f0f":14,"\u6bd4\u5982\u901a\u8fc7\u7528\u6237\u5bf9\u7535\u5b50\u5546\u52a1\u7f51\u7ad9\u8bc4\u8bba":13,"\u6bd4\u8f83\u53ef\u80fd\u7684\u547d\u4ee4\u5982\u4e0b":10,"\u6c34\u6e29":1,"\u6c49\u5ead":1,"\u6ca1":1,"\u6ca1\u6709\u4f5c\u7528":24,"\u6ca1\u6709\u5b89\u88c5":10,"\u6ca1\u6709\u8bbe\u7f6e":10,"\u6ce8\u610f":[1,3,24],"\u6cf3\u6c60":1,"\u6d41":1,"\u6d41\u7a0b\u5982\u4e0b":13,"\u6d44":1,"\u6d4b\u8bd5\u6570\u636e":13,"\u6d4b\u8bd5\u7684\u65f6\u5019\u9ed8\u8ba4\u4e0dshuffl":24,"\u6d4b\u8bd5\u811a\u672c\u5982\u4e0b":13,"\u6e29\u99a8":1,"\u6e90\u7801":13,"\u6e90\u7801\u6839\u76ee\u5f55":3,"\u6fc0\u6d3b\u51fd\u6570\u7c7b\u578b":13,"\u70ed\u60c5":1,"\u7136\u540e\u4ea4\u7ed9step\u51fd\u6570":2,"\u7136\u540e\u6267\u884c\u5982\u4e0b":10,"\u7136\u540e\u8fd0\u884c\u8fd9\u4e2acontainer\u5373\u53ef":9,"\u7248\u672c":10,"\u7279\u522b\u662f\u5728lstm\u7b49rnn\u4e2d":14,"\u751f\u6210\u5404\u4e2a\u5e73\u53f0\u7684makefil":4,"\u751f\u6210\u7684\u6570\u636e\u7f13\u5b58\u5728\u5185\u5b58\u91cc":14,"\u75280\u548c1\u8868\u793a":24,"\u7528\u4e86\u4e24\u4e2a\u6708\u4e4b\u540e\u8fd9\u4e2a\u663e\u793a\u5668\u5c4f\u5e55\u788e\u4e86":13,"\u7528\u4e8e\u4e0d\u652f\u6301avx\u6307\u4ee4\u96c6\u7684cpu":10,"\u7528\u6237\u4e5f\u53ef\u4ee5\u5728c":23,"\u7528\u6237\u4e5f\u53ef\u4ee5\u663e\u5f0f\u6307\u5b9a\u8fd4\u56de\u7684\u6570\u636e\u5bf9\u5e94\u5173\u7cfb":24,"\u7528\u6237\u53ea\u9700\u5b9a\u4e49rnn\u5728\u4e00\u4e2a\u65f6\u95f4\u6b65\u5185\u5b8c\u6210\u7684\u8ba1\u7b97":2,"\u7528\u6237\u53ef\u4ee5\u4f7f\u7528python\u7684":23,"\u7528\u6237\u53ef\u4ee5\u6839\u636e\u8bad\u7ec3log\u9009\u62e9test\u7ed3\u679c\u6700\u597d\u7684\u6a21\u578b\u6765\u9884\u6d4b":13,"\u7528\u6237\u53ef\u4ee5\u9009\u62e9\u5bf9\u5e94\u7248\u672c\u7684docker":9,"\u7528\u6237\u540d\u4e3a":9,"\u7528\u6237\u5728dataprovider\u4e2d\u9700\u8981\u5b9e\u73b0\u5982\u4f55\u8bbf\u95ee\u5176\u4e2d\u6bcf\u4e00\u4e2a\u6587\u4ef6":23,"\u7528\u6237\u5b9a\u4e49\u7684\u53c2\u6570\u4f7f\u7528args\u5728\u8bad\u7ec3\u914d\u7f6e\u4e2d\u8bbe\u7f6e":24,"\u7528\u6237\u63a5\u53e3":15,"\u7528\u6237\u9700\u8981\u5148\u5c06paddlepaddle\u5b89\u88c5\u5305\u4e0b\u8f7d\u5230\u672c\u5730":10,"\u7528\u6765\u505a\u9884\u6d4b\u548c\u7b80\u5355\u7684\u5b9a\u5236\u5316":9,"\u7528\u6b64\u65b9\u6cd5\u90e8\u5206\u865a\u62df\u673a\u53ef\u80fd\u68c0\u6d4b\u5230\u652f\u6301avx\u6307\u4ee4\u4f46\u662f\u5b9e\u9645\u8fd0\u884c\u4f1a\u6302\u6389":14,"\u7528\u8fc7\u4e00\u6b21\u7684\u65f6\u5019":24,"\u7531":2,"\u7531\u4e8e\u5916\u5c42\u6bcf\u4e2a\u65f6\u95f4\u6b65\u8fd4\u56de\u7684\u662f\u4e00\u4e2a\u5b50\u53e5":1,"\u7531\u4e8e\u5916\u5c42memory\u6ca1\u6709\u4efb\u4f55seq\u4fe1\u606f":1,"\u7531\u4e8e\u6570\u636e\u662f\u4e24\u6761":27,"\u7531\u4e8e\u8fd9\u4e2a\u5916\u5c42group\u91cc\u9762\u6ca1\u6709memori":1,"\u7531\u4e8edocker\u662f\u57fa\u4e8e\u5bb9\u5668\u7684\u8f7b\u91cf\u5316\u865a\u62df\u65b9\u6848":9,"\u7531\u4e8epaddlepaddle\u7684docker\u955c\u50cf\u5e76\u4e0d\u5305\u542b\u4efb\u4f55\u9884\u5b9a\u4e49\u7684\u8fd0\u884c\u547d\u4ee4":9,"\u7531\u4e8estep":2,"\u7531\u6613\u5230\u96be\u5c55\u793a4\u79cd\u4e0d\u540c\u7684\u7f51\u7edc\u914d\u7f6e":13,"\u7531\u8bcd\u8bed\u6784\u6210\u7684\u53e5\u5b50":0,"\u7535\u8111":1,"\u7684":[1,13],"\u7684\u4e00\u4e2a\u7b80\u5355\u8c03\u7528\u5982\u4e0b":2,"\u7684\u4efb\u4e00\u4e00\u79cd":14,"\u7684\u5185\u5b58":14,"\u7684\u540d\u5b57":24,"\u7684\u5747\u5300\u5206\u5e03":14,"\u7684\u5b89\u88c5\u6587\u6863":9,"\u7684\u5e73\u5747\u503c":0,"\u7684\u60c5\u51b5\u4e0b\u8d8a\u5927\u8d8a\u597d":24,"\u7684\u6570\u76ee\u4e00\u81f4":0,"\u7684\u6587\u6863":24,"\u7684\u65f6\u5019\u5982\u679c\u62a5\u4e00\u4e9b\u4f9d\u8d56\u672a\u627e\u5230\u7684\u9519\u8bef\u662f\u6b63\u5e38\u7684":10,"\u7684\u65f6\u95f4\u6b65\u4fe1\u606f\u6210\u6b63\u6bd4":14,"\u7684\u662f":24,"\u7684\u673a\u5668\u4e0a\u8fdb\u884c":3,"\u7684\u6838\u5fc3\u662f\u8bbe\u8ba1step\u51fd\u6570\u7684\u8ba1\u7b97\u903b\u8f91":2,"\u7684\u6bb5\u843d\u5b9a\u4e49\u4e3a\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":2,"\u7684\u72b6\u6001":2,"\u7684\u7f51\u6865\u6765\u8fdb\u884c\u7f51\u7edc\u901a\u4fe1":9,"\u7684\u8bdd":14,"\u7684\u8f93\u5165":2,"\u7684\u9519\u8bef":1,"\u7684demo\u5b66\u4e60\u5982\u4f55\u8fdb\u884c\u591a\u673a\u8bad\u7ec3":13,"\u7684docker\u53ef\u80fd\u7f3a\u4e4f":3,"\u7684matrix":27,"\u7684python\u5305\u662fpaddlepaddle\u7684\u8bad\u7ec3\u4e3b\u8981\u7a0b\u5e8f":9,"\u7684python\u5305\u6765\u505a\u914d\u7f6e\u6587\u4ef6\u89e3\u6790\u7b49\u5de5\u4f5c":9,"\u7684python\u9884\u6d4b\u8fc7\u7a0b":13,"\u76ee\u524d":2,"\u76ee\u524d\u652f\u6301\u4e24\u79cd":0,"\u76ee\u524d\u8fd8\u672a\u652f\u6301":2,"\u76ee\u5f55":13,"\u76ee\u5f55\u4e0b":3,"\u76f4\u5230\u8bad\u7ec3\u6536\u655b\u4e3a\u6b62":14,"\u76f4\u63a5\u52a0\u4e86\u4e00\u5c42group":1,"\u76f4\u63a5\u63d0\u53d6\u51fa\u795e\u7ecf\u7f51\u7edcoutput\u5c42\u7684\u8f93\u51fa\u7ed3\u679c":27,"\u76f8\u5173\u547d\u4ee4\u4e3a":9,"\u76f8\u5173\u7684\u6982":24,"\u76f8\u540c\u540d\u5b57\u7684\u53c2\u6570":14,"\u76f8\u5bf9":1,"\u76f8\u5bf9\u4e8epaddlepaddle\u7a0b\u5e8f\u8fd0\u884c\u65f6\u7684\u8def\u5f84":23,"\u76f8\u5f53":1,"\u770b\u4e0b\u9762\u7684\u89e3\u51b3\u65b9\u6848":14,"\u77e5\u9053\u5982\u4f55\u4ece":24,"\u793a":13,"\u795e\u7ecf\u7f51\u7edc\u5728\u8bad\u7ec3\u7684\u65f6\u5019":14,"\u795e\u7ecf\u7f51\u7edc\u7684\u8bad\u7ec3\u672c\u8eab\u662f\u4e00\u4e2a\u975e\u5e38\u6d88\u8017\u5185\u5b58\u548c\u663e\u5b58\u7684\u5de5\u4f5c":14,"\u79bb":1,"\u79f0\u4e4b\u4e3a\u53cc\u5c42\u5e8f\u5217\u7684\u4e00\u4e2a\u5b50\u5e8f\u5217":0,"\u7a0b\u5e8f\u6216\u8005\u81ea\u5b9a\u4e49\u4e00\u4e2a\u542b\u6709\u542f\u52a8\u811a\u672c\u7684imag":9,"\u7a97\u6237":1,"\u7aef\u81ea\u5b9a\u4e49\u4e00\u4e2a":23,"\u7aef\u8bfb\u53d6\u6570\u636e":14,"\u7b2c":1,"\u7b2c\u4e00\u4e2a\u53c2\u6570\u662f":24,"\u7b2c\u4e00\u4e2alast":1,"\u7b2c\u4e00\u4e2apass\u4f1a\u4ecepython\u7aef\u8bfb\u53d6\u6570\u636e":24,"\u7b2c\u4e00\u5929":1,"\u7b2c\u4e00\u6bb5\u6570\u636e\u4e3a\u8fd9\u5f20\u56fe\u7247\u7684label":24,"\u7b2c\u4e8c\u4e2a\u53c2\u6570\u662ffilenam":24,"\u7b2c\u4e8c\u6bb5\u6570\u636e\u4e3a\u8fd9\u4e2a\u56fe\u7247\u7684\u50cf\u7d20\u503c":24,"\u7b80\u5355\u4f18\u5316":3,"\u7b80\u5355\u7684\u4f7f\u7528":23,"\u7b80\u5355\u7684\u4f7f\u7528\u573a\u666f":23,"\u7b80\u5355\u7684\u4f7f\u7528\u6837\u4f8b\u4e3a":3,"\u7b80\u5355\u7684\u5168\u8fde\u63a5\u7f51\u7edc":14,"\u7b80\u5355\u7684\u542b\u6709ssh\u7684dockerfile\u5982\u4e0b":9,"\u7b80\u5355\u7684pydataprovider\u6837\u4f8b\u5c31\u8bf4\u660e\u5b8c\u6bd5\u4e86":24,"\u7b80\u76f4":1,"\u7b97\u6cd5":14,"\u7c7b\u522bid":13,"\u7c7b\u522bid\u7684\u6570\u636e\u7c7b\u578b":13,"\u7c7b\u578b\u53ef\u4ee5\u662fpaddlepaddle\u652f\u6301\u7684\u4efb\u610f\u8f93\u5165\u6570\u636e\u7c7b\u578b":0,"\u7c7b\u578b\u6765\u8bbe\u7f6e":24,"\u7eb5\u5411\u5305\u62ec\u56db\u4e2a\u7248\u672c":9,"\u7ec3":17,"\u7ecf\u5e38\u4f1a\u6d88\u8017\u6570\u5341g\u7684\u5185\u5b58\u548c\u6570g\u7684\u663e\u5b58":14,"\u7ed3\u4e0a\u8ff0\u7f51\u7edc\u7ed3\u6784\u5728amazon":13,"\u7ed9":1,"\u7ed9\u5b9aencoder\u8f93\u51fa\u548c\u5f53\u524d\u8bcd":2,"\u7ee7\u7eed\u8bad\u7ec3":24,"\u7ef4\u5ea6\u4e3aword":13,"\u7ef4\u5ea6\u662f\u7c7b\u522b\u4e2a\u6570":13,"\u7ef4\u5ea6\u662f\u8bcd\u5178\u5927\u5c0f":13,"\u7f13\u5b58\u6c60\u7684\u51cf\u5c0f":14,"\u7f13\u5b58\u8bad\u7ec3\u6570\u636e\u5230\u5185\u5b58":24,"\u7f16\u8bd1\u53c2\u6570\u9009\u9879\u6587\u4ef6":22,"\u7f16\u8bd1\u73af\u5883\u548c\u6e90\u4ee3\u7801":9,"\u7f16\u8bd1\u9009\u9879":4,"\u7f16\u8bd1\u9009\u9879\u4e3b\u8981\u63a8\u8350\u9ad8\u7ea7\u7528\u6237\u67e5\u770b":8,"\u7f16\u8bd1\u9009\u9879\u5217\u8868\u5982\u4e0b":4,"\u7f16\u8bd1paddlepaddle\u7684gpu\u7248\u672c\u5e76\u4e0d\u9700\u8981\u4e00\u5b9a\u5728\u5177\u6709gpu":3,"\u7f51\u7edc\u540d\u79f0":13,"\u7f51\u7edc\u914d\u7f6e":13,"\u7f6e\u8fd9\u4e9b\u53d8\u91cf":4,"\u800c":9,"\u800c\u4e09\u79cd\u5e8f\u5217\u6a21\u5f0f\u4e3a":24,"\u800c\u4e0d\u4f7f\u7528docker":9,"\u800c\u4e0d\u7528\u5173\u5fc3\u6570\u636e\u5982\u4f55\u4f20\u8f93\u7ed9paddlepaddl":24,"\u800c\u4e14\u9884\u6d4b\u7f51\u7edc\u901a\u5e38\u76f4\u63a5\u8f93\u51fa\u6700\u540e\u4e00\u5c42\u7684\u7ed3\u679c\u800c\u4e0d\u662f\u50cf\u8bad\u7ec3\u65f6\u4e00\u6837\u4ee5cost":27,"\u800c\u5728":[4,24],"\u800c\u5982\u679c\u6309\u987a\u5e8f\u8c03\u7528\u8fd9\u4e9bgenerator\u5c31\u4e0d\u4f1a\u51fa\u73b0\u8fd9\u4e2a\u95ee\u9898":24,"\u800c\u662f\u5c06\u6837\u672c\u7684\u5730\u5740\u653e\u5165\u53e6\u4e00\u4e2a\u6587\u672c":24,"\u800c\u662f\u76f4\u63a5\u4ece\u5185\u5b58\u7684\u7f13\u5b58\u91cc\u8bfb\u53d6\u6570\u636e":14,"\u800c\u6ca1\u6709\u6d4b\u8bd5\u6570\u636e":24,"\u800c\u7279\u5f81\u5373\u4e3a":24,"\u800c\u8fd9\u4e2a\u4e00\u822c\u8bf4\u660epaddlepaddle\u5df2\u7ecf\u5b89\u88c5\u5b8c\u6bd5\u4e86":10,"\u800c\u8fd9\u4e2a\u53d8\u91cf\u63a8\u8350\u5927\u4e8e\u8bad\u7ec3\u7684batch":24,"\u800c\u8fd9\u4e2acontext\u53ef\u80fd\u4f1a\u975e\u5e38":24,"\u800c\u975e\u9759\u6001\u52a0\u8f7dcuda\u52a8\u6001\u5e93":4,"\u800cexpand":1,"\u800cgpu\u7684\u9a71\u52a8\u548c\u8bbe\u5907\u5168\u90e8\u6620\u5c04\u5230\u4e86\u5bb9\u5668\u5185":9,"\u800cpaddlepaddle\u8fdb\u7a0b\u5e2e\u52a9\u7528\u6237\u505a\u4e86":24,"\u800crnn\u662f\u6700\u6d41\u884c\u7684\u9009\u62e9":2,"\u80fd\u591f\u5904\u7406\u53cc\u5c42\u5e8f\u5217":2,"\u80fd\u591f\u5bf9\u53cc\u5411\u5e8f\u5217\u8fdb\u884c\u5904\u7406\u7684\u6709":2,"\u80fd\u591f\u8bb0\u5f55\u4e0a\u4e00\u4e2asubseq":2,"\u811a\u672c":9,"\u811a\u672c\u53ef\u4ee5\u542f\u52a8paddlepaddle\u7684\u8bad\u7ec3\u8fdb\u7a0b\u548cpserv":9,"\u811a\u672c\u548c":9,"\u811a\u672c\u7c7b\u4f3c\u4e8e":9,"\u81ea\u52a8\u5b8c\u6210\u8fd9\u4e00\u8fc7\u7a0b":2,"\u81ea\u5b9a\u4e49\u4e00\u4e2adataprovid":23,"\u81f3\u5c11\u5177\u67093":9,"\u81f3\u6b64":[9,24],"\u8212\u9002":1,"\u82e5\u5e72\u4e2a\u53e5\u5b50\u6784\u6210\u4e00\u4e2a\u6bb5\u843d":0,"\u82e5\u8f93\u51fa\u662f\u5355\u5c42\u5e8f\u5217":0,"\u82e5\u8f93\u51fa\u662f\u53cc\u5c42\u5e8f\u5217":0,"\u83b7\u53d6\u5229\u7528one":13,"\u83b7\u53d6\u6bcf\u4e2a\u5355\u8bcd\u5de6\u53f3\u5404k\u4e2a\u8fd1\u90bb":13,"\u83b7\u53d6\u8be5\u6761\u6837\u672c\u7c7b\u522bid":13,"\u8868\u793a\u5c06\u5916\u5c42\u7684outer":1,"\u8868\u793a\u6574\u6570\u6807\u7b7e":24,"\u8868\u793a\u662f\u5426\u5141\u8bb8paddle\u6682\u5b58\u7565\u5fae\u591a\u4f59pool_size\u7684\u6570\u636e":24,"\u8868\u793a\u7a00\u758f\u7684\u5411\u91cf":24,"\u8868\u793a\u7a00\u758f\u7684\u96f6\u4e00\u5411\u91cf":24,"\u8868\u793a\u7a20\u5bc6\u7684\u6d6e\u70b9\u6570\u5411\u91cf":24,"\u8868\u793a\u8fc7\u4e8620\u4e2abatch":13,"\u8868\u793a\u8fc7\u4e862560\u4e2a\u6837\u672c":13,"\u8868\u793a\u8fd9\u4e2adataprovider\u662f\u8bad\u7ec3\u7528\u7684dataprovider\u6216\u8005\u6d4b\u8bd5\u7528\u7684":24,"\u8868\u793asubseq\u95f4\u4e0d\u5b58\u5728\u8054\u7cfb":1,"\u88ab\u6269\u5c55\u4e3a\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"\u8981\u6c42\u5355\u5c42\u5e8f\u5217\u542b\u6709\u5143\u7d20\u7684\u6570\u76ee":0,"\u8981\u751f\u6210\u7684\u76ee\u6807\u5e8f\u5217":2,"\u89c1":1,"\u89e3\u51b3\u529e\u6cd5\u662f":14,"\u89e3\u51b3\u529e\u6cd5\u662f\u5c06cuda":10,"\u89e3\u51b3\u65b9\u6cd5\u5f88\u7b80\u5355":10,"\u89e3\u6790\u8bad\u7ec3\u65f6\u7684\u914d\u7f6e\u6587\u4ef6":27,"\u89e3\u91ca":13,"\u8ba9\u795e\u7ecf\u7f51\u7edc\u53ef\u4ee5\u8fdb\u884c\u8bad\u7ec3":23,"\u8bad\u7ec3":9,"\u8bad\u7ec3\u6570\u636e\u975e\u5e38\u591a\u7684\u60c5\u51b5\u4e0b":24,"\u8bad\u7ec3\u6587\u4ef6\u5217\u8868":23,"\u8bad\u7ec3\u65f6\u6240\u9700\u8bbe\u7f6e\u7684\u4e3b\u8981\u53c2\u6570\u5982\u4e0b":13,"\u8bad\u7ec3\u7684\u65f6\u5019\u9ed8\u8ba4shuffl":24,"\u8bad\u7ec3\u811a\u672c":13,"\u8bad\u7ec3\u811a\u672c\u5728":13,"\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b\u8ddd\u79bb":14,"\u8bad\u7ec3\u8f6e\u6b21":13,"\u8bb2\u6570\u636e\u53d1\u9001\u7ed9paddlepaddl":24,"\u8bb2\u89e3\u5982\u4f55\u4f7f\u7528\u53cc\u5c42rnn":1,"\u8bbe\u7f6e\u4e0b\u5217\u7f16\u8bd1\u9009\u9879\u65f6":4,"\u8bbe\u7f6e\u53c2\u6570\u7684\u540d\u5b57":14,"\u8bbe\u7f6e\u547d\u4ee4\u884c\u53c2\u6570":14,"\u8bbe\u7f6e\u6210":[14,24],"\u8bbe\u7f6e\u6210\u4e00\u4e2a\u5c0f\u4e00\u4e9b\u7684\u503c":14,"\u8bbe\u7f6e\u6210\u4e86\u5e8f\u5217":24,"\u8bbe\u7f6e\u6210true\u7684\u8bdd":24,"\u8bbe\u7f6e\u8f93\u5165\u7c7b\u578b":24,"\u8bc4\u4f30\u4ea7\u54c1\u7684\u8d28\u91cf":13,"\u8bcd\u6027\u6807\u6ce8":12,"\u8be5\u5c42\u795e\u7ecf\u5143\u4e2a\u6570":13,"\u8be5\u6570\u636e":24,"\u8be5\u6784\u5efa\u811a\u672c\u5145\u5206\u8003\u8651\u4e86\u7f51\u7edc\u4e0d\u7a33\u5b9a\u7684\u60c5\u51b5":3,"\u8be5\u6a21\u578b\u4f9d\u7136\u662f\u4f7f\u7528\u903b\u8f91\u56de\u5f52\u5206\u7c7b\u7f51\u7edc\u7684\u6846\u67b6":13,"\u8be5\u76ee\u5f55\u4e0b\u6709\u4e24\u4e2a\u6587\u4ef6":3,"\u8be5\u811a\u672c\u7684\u4f7f\u7528\u65b9\u6cd5\u662f":3,"\u8be5image\u57fa\u4e8eubuntu":3,"\u8be5image\u7684\u6784\u5efa\u5728dock":3,"\u8be6\u60c5\u8bf7\u53c2\u8003":27,"\u8be6\u7ec6\u6587\u6863\u53c2\u8003":14,"\u8be6\u7ec6\u7684\u53c2\u6570\u89e3\u91ca\u5982\u4e0b\u9762\u8868\u683c":13,"\u8be6\u7ec6\u7684\u547d\u4ee4\u884c\u53c2\u6570\u8bf7\u53c2\u8003":27,"\u8be6\u7ec6\u7684cmake\u4f7f\u7528\u65b9\u6cd5\u53ef\u4ee5\u53c2\u8003":4,"\u8be6\u7ec6\u89c1":0,"\u8bed\u4e49\u5b8c\u5168\u76f8\u540c":1,"\u8bf4\u660e":4,"\u8bf4\u660e\u547d\u4ee4\u884c\u53c2\u6570":10,"\u8bf7\u53c2\u8003":[9,24],"\u8bf7\u53c2\u8003\u4e0b\u8282refer":24,"\u8bf7\u53c2\u8003\u4e0b\u8ff0\u6587\u7ae0":23,"\u8bf7\u5b89\u88c5cuda":10,"\u8bf7\u5f53\u6210\u662f\u4e0d\u652f\u6301":14,"\u8bfb\u5165\u89e3\u6790\u8bad\u7ec3\u914d\u7f6e":27,"\u8bfb\u53d6\u6570\u636e":24,"\u8c03\u7528":4,"\u8c03\u7528\u4e00\u6b21":24,"\u8c03\u7528\u7b2c\u4e8c\u6b21\u7684\u65f6\u5019":24,"\u8d1f\u6837\u672c":13,"\u8d1f\u8d23\u591a\u673a\u8bad\u7ec3\u4e2d\u7684\u53c2\u6570\u805a\u5408\u5de5\u4f5c":17,"\u8d1f\u9762\u60c5\u7eea\u4e24\u7c7b":24,"\u8d77":1,"\u8def\u5f84\u53d8\u91cf\u4e3a":4,"\u8f83":1,"\u8f93\u5165":0,"\u8f93\u5165\u548c\u8f93\u51fa\u90fd\u662f\u5355\u5c42\u5e8f\u5217":2,"\u8f93\u5165\u548c\u8f93\u51fa\u90fd\u662f\u53cc\u5c42\u5e8f\u5217":2,"\u8f93\u5165n\u4e2a\u5355\u8bcd":13,"\u8f93\u51fa":0,"\u8f93\u51fa\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":2,"\u8f93\u51fa\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":2,"\u8f93\u51fa\u4e3an\u4e2aword":13,"\u8f93\u51fa\u5e8f\u5217\u7684\u7c7b\u578b":0,"\u8f93\u51fa\u5e8f\u5217\u7684\u8bcd\u8bed\u6570\u548c\u8f93\u5165\u5e8f\u5217\u4e00\u81f4":2,"\u8fc7\u4e86\u4e00\u4e2a\u5f88\u7b80\u5355\u7684recurr":1,"\u8fc7\u5b8c\u6240\u6709\u8bad\u7ec3\u6570\u636e\u5373\u4e3a\u4e00\u4e2apass":14,"\u8fd0\u884c":[9,10],"\u8fd0\u884c\u4f7f\u7528\u7684cudnn\u5c3d\u91cf\u662f\u540c\u4e00\u4e2a\u7248\u672c":4,"\u8fd0\u884c\u8fd9\u4e2acontain":9,"\u8fd0\u884cpaddlepaddle\u7684gpu\u7248\u672c\u4e00\u5b9a\u8981\u5728\u5177\u6709cuda\u7684\u673a\u5668\u4e0a\u8fd0\u884c":3,"\u8fd1":1,"\u8fd4\u56de0":24,"\u8fd4\u56de\u4e00\u4e2alist\u6216\u8005tupl":24,"\u8fd4\u56de\u6570\u636e\u5728paddlepaddle\u4e2d\u662f\u4ec5\u4ec5\u8fd4\u56de\u4e00\u6761\u5b8c\u6574\u7684\u8bad\u7ec3\u6837\u672c":24,"\u8fd4\u56de\u7684\u987a\u5e8f\u9700\u8981\u548c":24,"\u8fd4\u56debatch_size\u7684\u5927\u5c0f":24,"\u8fd8\u4f1a":1,"\u8fd8\u662f":1,"\u8fd8\u6709":1,"\u8fd9":[1,13,14],"\u8fd93\u4e2a\u5b50\u6b65\u9aa4\u53ef\u914d\u7f6e\u4e3a":13,"\u8fd9\u4e00\u8fc7\u7a0b\u5bf9\u7528\u6237\u662f\u5b8c\u5168\u900f\u660e\u7684":2,"\u8fd9\u4e09\u4e2alayer\u5c06\u5b83\u5148\u53d8\u6210\u4e00\u4e2a\u5143\u7d20":1,"\u8fd9\u4e24\u5c42\u4ec5\u662f\u4e3a\u4e86\u5c55\u793a\u5b83\u4eec\u7684\u7528\u6cd5":1,"\u8fd9\u4e2a":1,"\u8fd9\u4e2a\u4e5f\u662fpaddlepaddle\u6240\u80fd\u591f\u4fdd\u8bc1\u7684shuffle\u7c92\u5ea6":24,"\u8fd9\u4e2a\u4efb\u52a1\u7684\u914d\u7f6e\u4e3a":14,"\u8fd9\u4e2a\u4efb\u52a1\u7684dataprovider\u4e3a":14,"\u8fd9\u4e2a\u51fd\u6570\u4ee5\u4e00\u6761\u6570\u636e\u4e3a\u53c2\u6570":24,"\u8fd9\u4e2a\u51fd\u6570\u4f1a\u5728":24,"\u8fd9\u4e2a\u51fd\u6570\u5728\u521d\u59cb\u5316\u7684\u65f6\u5019\u4f1a\u88ab\u8c03\u7528":24,"\u8fd9\u4e2a\u51fd\u6570\u7684\u53c2\u6570\u662f":24,"\u8fd9\u4e2a\u521d\u59cb\u5316\u51fd\u6570\u5177\u6709\u5982\u4e0b\u53c2\u6570":24,"\u8fd9\u4e2a\u53c2\u6570\u5728\u8fd9\u4e2a\u6837\u4f8b\u91cc\u6ca1\u6709\u4f7f\u7528":24,"\u8fd9\u4e2a\u53c2\u6570\u88abpaddlepaddle\u8fdb\u7a0b\u4f20\u5165":24,"\u8fd9\u4e2a\u548c\u5728":24,"\u8fd9\u4e2a\u58f0\u660e\u57fa\u672c\u4e0a\u548cmnist\u7684\u6837\u4f8b\u4e00\u81f4":24,"\u8fd9\u4e2a\u5916\u5c42memori":1,"\u8fd9\u4e2a\u5b57\u5178\u53ef\u4ee5\u5728":24,"\u8fd9\u4e2a\u5bf9\u5e94\u5173\u7cfb\u53ef\u80fd\u4e0d\u6b63\u786e":24,"\u8fd9\u4e2a\u5bf9\u8c61\u548cprocess\u7684\u7b2c\u4e00\u4e2a\u53c2\u6570\u4e00\u81f4":24,"\u8fd9\u4e2a\u5de5\u5177\u7c7b\u63a5\u6536\u548cpydataprovider2\u4e00\u6837\u7684\u8f93\u5165\u6570\u636e":27,"\u8fd9\u4e2a\u5e8f\u5217\u6a21\u578b\u6bd4\u8f83\u590d\u6742":24,"\u8fd9\u4e2a\u5e8f\u5217\u7684\u6bcf\u4e2a\u5143\u7d20\u53c8\u662f\u4e00\u4e2a\u5e8f\u5217":2,"\u8fd9\u4e2a\u63a5\u53e3\u5e76\u4e0d\u7528\u6237\u53cb\u597d":27,"\u8fd9\u4e2a\u663e\u793a\u5668\u5f88\u68d2":13,"\u8fd9\u4e2a\u672c\u8eab\u4e0d\u662f\u4e00\u4e2a\u5f88":24,"\u8fd9\u4e2a\u6a21\u5757\u4e2d\u7684":24,"\u8fd9\u4e2a\u8bbe\u7f6e\u4e3a":24,"\u8fd9\u4e2a\u8f6f\u4ef6\u5305\u6587\u6863\u76f8\u5bf9\u5b8c\u5584":27,"\u8fd9\u4e2a\u8fc7\u7a0b\u5bf9\u7528\u6237\u4e5f\u662f\u900f\u660e\u7684":2,"\u8fd9\u4e2a\u95ee\u9898\u662fpydataprovider\u8bfb\u6570\u636e\u65f6\u5019\u7684\u903b\u8f91\u95ee\u9898":24,"\u8fd9\u4e2alayer\u7684\u8f93\u51fa\u4f1a\u4f5c\u4e3a\u6574\u4e2a":2,"\u8fd9\u4e5f\u4f1a\u6781\u5927\u51cf\u5c11\u6570\u636e\u8bfb\u5165\u7684\u8017\u65f6":14,"\u8fd9\u4e9b\u5185\u5b58\u5c31\u4e0d\u8003\u8651\u5982\u4f55\u7f29\u51cf\u4e86":14,"\u8fd9\u4e9b\u53c2\u6570\u5305\u62ecpaddle\u5b9a\u4e49\u7684\u53c2\u6570":24,"\u8fd9\u4e9b\u53d8":4,"\u8fd9\u4e9b\u53d8\u91cf\u53ea\u5728\u7b2c\u4e00\u6b21cmake\u7684\u65f6\u5019\u6709\u6548":4,"\u8fd9\u4e9b\u53d8\u91cf\u5747\u53ef\u4ee5\u4f7f\u7528":4,"\u8fd9\u4e9b\u5b50\u53e5\u7684\u957f\u5ea6\u5f80\u5f80\u4e0d\u7b49\u957f":1,"\u8fd9\u4e9b\u6570\u636e\u4f1a\u88ab\u7528\u6765\u66f4\u65b0\u53c2\u6570":14,"\u8fd9\u4e9b\u6570\u636e\u4f7f\u7528\u7684\u5185\u5b58\u4e3b\u8981\u548c\u4e24\u4e2a\u53c2\u6570\u6709\u5173\u7cfb":14,"\u8fd9\u4e9b\u6d41\u7a0b\u4e2d\u7684\u6570\u636e\u4e0b\u8f7d":13,"\u8fd9\u5176\u4e2d":14,"\u8fd9\u662f\u4e00\u79cd\u975e\u5e38\u7075\u6d3b\u7684\u6570\u636e\u7ec4\u7ec7\u65b9\u5f0f":0,"\u8fd9\u6837\u505a\u53ef\u4ee5\u6781\u5927\u7684\u51cf\u5c11\u5185\u5b58\u5360\u7528":14,"\u8fd9\u6837\u505a\u53ef\u4ee5\u907f\u514d\u5f88\u591a\u6b7b\u9501\u95ee\u9898":24,"\u8fd9\u79cd\u521d\u59cb\u5316\u65b9\u5f0f\u5728\u4e00\u822c\u60c5\u51b5\u4e0b\u4e0d\u4f1a\u4ea7\u751f\u5f88\u5dee\u7684\u7ed3\u679c":14,"\u8fd9\u79cd\u7c7b\u578b\u7684\u8f93\u5165\u5fc5\u987b\u901a\u8fc7":2,"\u8fd9\u884c\u7684\u4f5c\u7528\u662f\u8bbe\u7f6edataprovider\u7684\u4e00\u4e9b\u5c5e\u6027":24,"\u8fd9\u91cc":[14,24],"\u8fd9\u91cc\u4e3e\u4f8b\u7684\u6570\u636e\u662f\u82f1\u6587\u60c5\u611f\u5206\u7c7b\u7684\u6570\u636e":24,"\u8fd9\u91cc\u4ee5":13,"\u8fd9\u91cc\u4ee5mnist\u624b\u5199\u8bc6\u522b\u4e3a\u4f8b":24,"\u8fd9\u91cc\u4f7f\u7528\u7b80\u5355\u7684":14,"\u8fd9\u91cc\u53ef\u4ee5\u53c2\u8003paddle\u7684":22,"\u8fd9\u91cc\u6211\u4eec\u4f7f\u7528\u7b80\u5355\u7684\u6587\u672c\u6587\u4ef6\u8868\u793amnist\u56fe\u7247":24,"\u8fd9\u91cc\u6307\u5b9a\u8bcd\u5178":13,"\u8fd9\u91cc\u6ca1\u6709\u4ecb\u7ecd\u591a\u673a\u5206\u5e03\u5f0f\u8bad\u7ec3":13,"\u8fd9\u91cc\u7684":24,"\u8fd9\u91cc\u7684\u8f93\u5165\u7279\u5f81\u662f\u8bcdid\u7684\u5e8f\u5217":24,"\u8fd9\u91cc\u8981\u6ce8\u610f\u9884\u6d4b\u6570\u636e\u901a\u5e38":27,"\u8fd9\u91cc\u8bbe\u7f6e\u7684\u662f\u8fd4\u56de\u4e00\u4e2a":24,"\u8fd9\u91cc\u8bf4\u660e\u4e86\u8bad\u7ec3\u6570\u636e\u662f":24,"\u8fd9\u91cc\u91c7\u7528adam\u4f18\u5316\u65b9\u6cd5":13,"\u8fdb\u5165\u8be5\u6e90\u7801\u76ee\u5f55":3,"\u8fdb\u5165docker":9,"\u8fdc\u7a0b\u8bbf\u95ee":9,"\u8fde\u63a5":2,"\u8fde\u63a5\u8bf7\u53c2\u8003":13,"\u9002\u4e2d":1,"\u9009":1,"\u9009\u62e9":1,"\u9009\u62e9\u666e\u901acpu\u7248\u672c\u7684devel\u7248\u672c\u7684imag":9,"\u9009\u9879":4,"\u901a\u5e38\u505a\u6cd5\u662f\u4ece\u4e00\u4e2a\u6bd4\u8f83\u5927\u7684learning_rate\u5f00\u59cb\u8bd5":14,"\u901a\u5e38\u6765\u8bf4\u6267\u884c\u4e0bgrep":14,"\u901a\u5e38\u6839\u636e\u4efb\u52a1\u9700\u6c42\u8fdb\u884c\u4e0d\u540c\u8bbe\u7f6e":1,"\u901a\u77e5":1,"\u901a\u8fc7\u4e24\u4e2a\u5d4c\u5957\u7684":2,"\u901a\u8fc7\u591a\u7ec4\u8bed\u4e49\u76f8\u540c\u7684\u5355\u53cc\u5c42rnn\u914d\u7f6e":1,"\u901a\u8fc7\u5f15\u7528memory\u5f97\u5230\u8fd9\u4e2alayer\u4e0a\u4e00\u4e2a\u65f6\u523b\u7684\u8f93\u51fa":2,"\u901a\u8fc7\u5f15\u7528memory\u5f97\u5230\u8fd9\u4e2alayer\u4e0a\u4e00\u4e2a\u65f6\u523b\u8f93\u51fa":2,"\u901a\u8fc7\u7f16\u8bd1\u65f6\u6307\u5b9a\u8def\u5f84\u6765\u5b9e\u73b0\u5f15\u7528\u5404\u79cdbla":4,"\u901a\u8fc7data":2,"\u903b\u8f91\u56de\u5f52":13,"\u9053\u6b49":1,"\u9069":1,"\u90a3\u4e48":[2,24],"\u90a3\u4e480\u5c42\u5e8f\u5217\u5373\u4e3a\u4e00\u4e2a\u8bcd\u8bed":2,"\u90a3\u4e48\u53ef\u4ee5\u8ba4\u4e3a\u8bad\u7ec3\u4e0d\u6536\u655b":14,"\u90a3\u4e48\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d":23,"\u90a3\u4e48\u5982\u4f55\u5224\u65ad\u8bad\u7ec3\u4e0d\u6536\u655b\u5462":14,"\u90a3\u4e48\u5bf9\u5e94\u7684dataprovider\u65e2\u4e3a":24,"\u90a3\u4e48\u5e38\u6570\u8f93\u51fa\u6240\u80fd\u8fbe\u5230\u7684\u6700\u5c0fcost\u662f":14,"\u90a3\u4e48\u6211\u4eec\u53ef\u4ee5\u5224\u65ad\u4e3a\u8bad\u7ec3\u4e0d\u6536\u655b":14,"\u90a3\u4e48\u6536\u655b\u53ef\u80fd\u5f88\u6162":14,"\u90a3\u4e48\u6700\u597d\u5c06\u6570\u636e\u6587\u4ef6\u5728\u6bcf\u6b21\u8bfb\u53d6\u4e4b\u524d\u505a\u4e00\u6b21shuffl":14,"\u90a3\u4e48\u8bad\u7ec3\u6709\u53ef\u80fd\u4e0d\u6536\u655b":14,"\u90a3\u4e48\u8be5\u4f18\u5316\u7b97\u6cd5\u81f3\u5c11\u9700\u8981":14,"\u90a3\u4e48\u8fd9\u4e2a\u4e0b\u8f7d\u7ebf\u7a0b\u5c06\u4f1a\u5173\u95ed":3,"\u90a3\u4e48paddlepaddle\u4f1a\u7c97\u7565\u7684\u6839\u636elayer\u7684\u58f0\u660e\u987a\u5e8f":24,"\u90a3\u51cf\u5c11\u5b66\u4e60\u738710\u500d\u7ee7\u7eed\u8bd5\u9a8c":14,"\u90e8\u5206\u8001\u7684cpu\u578b\u53f7\u65e0\u6cd5\u652f\u6301\u8fd9\u6837\u7684\u6307\u4ee4":14,"\u90fd":1,"\u90fd\u4f20\u9012\u7ed9process\u51fd\u6570":24,"\u90fd\u662f\u5bf9layer1\u5143\u7d20\u7684\u62f7\u8d1d":0,"\u914d\u7f6e":1,"\u914d\u7f6e\u4e86":24,"\u914d\u7f6e\u53c2\u6570\u914d\u7f6e\u7ed9dataprovider\u7684":24,"\u914d\u7f6e\u6587\u4ef6":13,"\u914d\u7f6eapi":0,"\u9152\u5e97":1,"\u91c7\u7528multi":14,"\u91cc\u4f1a\u7ee7\u7eed\u5b89\u88c5":10,"\u91cc\u63d0\u4f9b\u4e86\u6570\u636e\u4e0b\u8f7d\u811a\u672c":13,"\u91cc\u9762\u8bfb\u53d6":24,"\u91cd\u65b0\u7f16\u8bd1paddlepaddl":14,"\u91cf\u4e5f\u53ef\u4ee5\u901a\u8fc7\u8c03\u7528cmake\u547d\u4ee4\u524d\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u6307\u5b9a":4,"\u9488\u5bf9\u5185\u5b58\u548c\u663e\u5b58":14,"\u9488\u5bf9\u672c\u95ee\u9898":13,"\u94fe\u63a5\u4f55\u79cdblas\u7b49\u7b49":4,"\u9519\u8bef\u7387":13,"\u957f\u5ea6":14,"\u95f4\u63a5\u4f7f\u7528":1,"\u95f4\u9694":24,"\u9664\u4e86":24,"\u9664\u4e86boot":1,"\u9664\u8fc7data\u5c42":13,"\u9700\u8981\u5148\u5728\u6bcf\u4e2a\u8282\u70b9\u542f\u52a8":14,"\u9700\u8981\u53c2\u8003":9,"\u9700\u8981\u5c06\u5176parameter\u8bbe\u7f6e\u6210":14,"\u9700\u8981\u652f\u6301avx\u6307\u4ee4\u96c6\u7684cpu":9,"\u9700\u8981\u6ce8\u610f":24,"\u9700\u8981\u6ce8\u610f\u7684\u662f":[4,10],"\u9700\u8981\u9075\u5faa\u4ee5\u4e0b\u7ea6\u5b9a":2,"\u9884\u6d4b\u6570\u636e\u6307\u5b9atest":13,"\u9884\u6d4b\u7ed3\u679c\u4ee5\u6587\u672c\u7684\u5f62\u5f0f\u4fdd\u5b58\u5728":13,"\u9884\u6d4b\u811a\u672c":13,"\u9884\u6d4bid":13,"\u989d\u5916\u7684\u53c2\u6570":13,"\u9996\u5148":1,"\u9996\u5148\u5217\u4e3e\u903b\u8f91\u56de\u5f52\u7f51\u7edc":13,"\u9996\u5148\u6211\u4eec\u5c06\u8fd9\u4e2a\u6570\u636e\u6587\u4ef6":24,"\u9996\u5148\u8bf7\u53c2\u8003":13,"\u9aa43":13,"\u9ad8\u65af\u5206\u5e03":14,"\u9ed8\u8ba4\u4e00\u4e2apass\u4fdd\u5b58\u4e00\u6b21\u6a21\u578b":13,"\u9ed8\u8ba4\u4e0d\u8bbe\u7f6e":2,"\u9ed8\u8ba4\u4e3a\u7b2c\u4e00\u4e2a\u8f93\u5165":2,"\u9ed8\u8ba4\u503c":[0,4],"\u9ed8\u8ba4\u521d\u59cb\u72b6\u4e3a0":2,"\u9ed8\u8ba4\u5355\u4f4d\u662fbyte":3,"\u9ed8\u8ba4\u60c5\u51b5\u4e0b":14,"\u9ed8\u8ba4\u60c5\u51b5\u4e0b\u4e00\u6761\u6570\u636e":24,"adamax\u7b49":13,"amazon\u7535\u5b50\u4ea7\u54c1\u8bc4\u8bba\u6570\u636e":13,"api\u9884\u6d4b":13,"argument\u4f20\u5165":24,"argument\u5f62\u5f0f\u4f20\u5165":24,"async_sgd\u8fdb\u884c\u8bad\u7ec3\u65f6":14,"atlas\u5e93\u7684\u8def\u5f84":4,"batch\u4e2d\u5305\u542b":14,"batches\u8bbe\u7f6e\u6bcf\u9694\u591a\u5c11batch\u4fdd\u5b58\u4e00\u6b21\u6a21\u578b":13,"bool\u53c2\u6570":24,"case":[13,26],"cd\u5230\u542b\u6709dockerfile\u7684\u8def\u5f84\u4e2d":9,"check\u662ffalse\u7684\u8bdd":24,"class":14,"cmake\u53ef\u4ee5\u5c06cmake\u9879\u76ee\u6587\u4ef6":4,"cmake\u662f\u4e00\u4e2a\u8de8\u5e73\u53f0\u7684\u7f16\u8bd1\u811a\u672c":4,"cmake\u7684\u5b98\u65b9\u6587\u6863":4,"cmake\u7f16\u8bd1\u65f6\u4f1a\u9996\u5148\u5728\u7cfb\u7edf\u8def\u5f84":4,"container\u540e":9,"cost\u8fd8\u5927\u4e8e\u8fd9\u4e2a\u6570":14,"cpu\u7248\u672c":9,"cpuinfo\u770b\u770b\u662f\u5426\u6709\u8f93\u51fa\u5373\u53ef\u77e5\u9053\u662f\u5426\u652f\u6301":14,"cuda\u76f8\u5173\u7684driver\u548c\u8bbe\u5907\u6620\u5c04\u8fdbcontainer\u4e2d":9,"d\u547d\u4ee4\u5373\u53ef":4,"d\u547d\u4ee4\u6307\u5b9a":4,"dataprovider\u521b\u5efa\u7684\u65f6\u5019\u6267\u884c":24,"dataprovider\u53ef\u4ee5\u662f":24,"dataprovider\u63d0\u4f9b\u4e86\u4e24\u79cd\u7b80\u5355\u7684cache\u7b56\u7565":24,"dataprovider\u7684\u5177\u4f53\u7528\u6cd5\u548c\u5982\u4f55\u5b9e\u73b0\u4e00\u4e2a\u65b0\u7684dataprovid":23,"dataprovider\u7f13\u51b2\u6c60\u5185\u5b58":14,"decoder\u5faa\u73af\u5c55\u5f00\u7684\u6bcf\u4e00\u4e2a\u65f6\u95f4\u6b65\u4f1a\u5f15\u7528\u5168\u90e8\u7ed3\u679c":2,"decoder\u63a5\u53d7\u4e24\u4e2a\u8f93\u5165":2,"decoder\u6bcf\u6b21\u9884\u6d4b\u4ea7\u751f\u4e0b\u4e00\u4e2a\u6700\u53ef\u80fd\u7684\u8bcd\u8bed":2,"decoer\u67b6\u6784":2,"devel\u548cdemo":9,"dim\u7684\u65b0\u7684\u5411\u91cf":13,"dim\u7ef4\u5ea6\u5411\u91cf":13,"docker\u662f\u4e00\u4e2a\u57fa\u4e8e\u5bb9\u5668\u7684\u8f7b\u91cf\u7ea7\u865a\u62df\u73af\u5883":9,"docker\u7684\u5b98\u65b9\u6587\u6863":9,"dockerfile\u548cbuild":3,"dockerfile\u662fdock":3,"dockerfile\u7684\u6587\u6863":9,"dockerfile\u7684\u6700\u4f73\u5b9e\u8df5":9,"driver\u6dfb\u52a0\u5230ld_library_path\u4e2d":10,"elec\u6d4b\u8bd5\u96c6":13,"embedding\u6a21\u578b\u9700\u8981\u7a0d\u5fae\u6539\u53d8\u6570\u636e\u63d0\u4f9b\u7684\u811a\u672c":13,"encoder\u548cdecoder\u53ef\u4ee5\u662f\u80fd\u591f\u5904\u7406\u5e8f\u5217\u7684\u4efb\u610f\u795e\u7ecf\u7f51\u7edc\u5355\u5143":2,"encoder\u8f93\u51fa":2,"entropy\u4f5c\u4e3acost":14,"export":[4,9,10],"f\u4ee3\u8868\u4e00\u4e2a\u6d6e\u70b9\u6570":24,"float":24,"generator\u4fbf\u4f1a\u5b58\u4e0b\u5f53\u524d\u7684\u4e0a\u4e0b\u6587":24,"generator\u7684\u4e0a\u4e0b\u6587\u4e2d\u5c3d\u91cf\u7559":24,"generator\u81f3\u5c11\u8c03\u7528\u4e24\u6b21\u624d\u4f1a\u77e5\u9053\u662f\u5426\u505c\u6b62":24,"gpu\u53cc\u7f13\u5b58":24,"gpu\u7248\u672c":9,"gpu\u7248\u672c\u4e8c\u8fdb\u5236":4,"group\u548c\u5355\u5c42\u5e8f\u5217\u7684\u51e0\u4e4e\u4e00\u6837":1,"group\u5916":1,"gru\u6a21\u578b":13,"gru\u6a21\u578b\u914d\u7f6e":13,"i\u4ee3\u8868\u4e00\u4e2a\u6574\u6570":24,"id\u4e3a0\u7684\u6982\u7387":13,"id\u4e3a1\u7684\u6982\u7387":13,"image\u6784\u5efa\u6e90\u7801\u653e\u7f6e\u5728":3,"image\u7684\u4e3b\u8981\u63cf\u8ff0\u6587\u4ef6":3,"image\u7684\u4e3b\u8981\u6784\u5efa\u6b65\u9aa4":3,"image\u7684\u6784\u5efa\u6b65\u9aa4":3,"import":[13,14,24,27],"include\u4e0b\u9700\u8981\u5305\u542bcbla":4,"include\u4e0b\u9700\u8981\u5305\u542bmkl":4,"init_hook\u53ef\u4ee5\u4f20\u5165\u4e00\u4e2a\u51fd\u6570":24,"int":[1,13,24],"key\u662fdata_layer\u7684\u540d\u5b57":24,"label\u662finteg":1,"layer1\u5fc5\u987b\u662f\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"layer1\u5fc5\u987b\u662f\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"layer2\u4e00\u81f4":0,"layer2\u53ef\u4ee5\u662f\u4e00\u4e2a\u5355\u5c42\u5e8f\u5217":0,"layer2\u5fc5\u987b\u662f\u4e00\u4e2a\u53cc\u5c42\u5e8f\u5217":0,"layer\u4e0d\u5173\u5fc3\u6570\u636e\u662f\u5426\u662f\u5e8f\u5217\u683c\u5f0f":1,"layer\u4e0d\u80fd\u94fe\u63a5\u5916\u5c42\u7684\u8fd9\u4e2amemori":1,"layer\u4f20\u7ed9\u4e0b\u4e00\u4e2a\u5b50\u53e5\u7684memori":1,"layer\u4f5c\u4e3a\u8f93\u51fa":27,"layer\u540e":1,"layer\u548caverag":1,"layer\u548cembed":1,"layer\u548clstmemori":1,"layer\u5c42":1,"layer\u62ff\u5230\u7684\u7528\u6237\u8f93\u5165":2,"layer\u6587\u6863":13,"layer\u7684\u4f7f\u7528\u793a\u4f8b\u5982\u4e0b":0,"ld_library_path\u7b49\u7b49":10,"ld_library_path\u91cc\u9762\u627e\u4e0d\u5230\u8fd9\u4e9b\u52a8\u6001":10,"lib\u4e0b\u9700\u8981\u5305\u542bcblas\u548catlas\u4e24\u4e2a\u5e93":4,"lib\u4e0b\u9700\u8981\u5305\u542bcblas\u5e93":4,"lib\u4e0b\u9700\u8981\u5305\u542bopenblas\u5e93":4,"lib\u76ee\u5f55\u4e0b\u9700\u8981\u5305\u542b":4,"list\u4e0d\u8bbe\u7f6e":23,"list\u4e2d":[23,24],"list\u4e2d\u7684\u4e00\u884c":24,"list\u4e2d\u7684\u6bcf\u4e00\u884c":24,"list\u4e3a\u7eaf\u6587\u672c\u6587\u4ef6":23,"list\u4e5f\u53ef\u4ee5\u653e\u7f6ehdfs\u6587\u4ef6\u8def\u5f84":23,"list\u5199\u5165\u90a3\u4e2a\u6587\u672c\u6587\u4ef6\u7684\u5730\u5740":24,"list\u5373\u4e3a":24,"list\u548ctest":23,"list\u5747\u4e3a\u672c\u5730\u7684\u4e24\u4e2a\u6587\u4ef6":23,"list\u6307\u5b9a\u7684\u6570\u636e":13,"list\u7684\u4f4d\u7f6e":13,"list\u82e5\u5e72\u6570\u636e\u6587\u4ef6\u8def\u5f84\u7684\u67d0\u4e00\u4e2a\u8def\u5f84":24,"lstm\u6a21\u578b\u7b49":13,"lstm\u6a21\u578b\u914d\u7f6e":13,"make\u548cmak":5,"mem\u4f5c\u4e3a\u5185\u5c42memory\u7684\u521d\u59cb\u72b6\u6001":1,"mem\u662f\u4e00\u4e2a\u5b50\u53e5\u7684\u6700\u540e\u4e00\u4e2a\u5411\u91cf":1,"memory\u4e0d\u80fd\u72ec\u7acb\u5b58\u5728":2,"memory\u53ea\u80fd\u5728":2,"memory\u6307\u5411\u4e00\u4e2alay":2,"memory\u7684\u521d\u59cb\u72b6\u6001":2,"memory\u7684\u66f4\u591a\u8ba8\u8bba\u8bf7\u53c2\u8003\u8bba\u6587":2,"memory\u7684i":2,"memory\u9ed8\u8ba4\u521d\u59cb\u5316\u4e3a0":2,"mkl\u7684\u8def\u5f84":4,"mkl_sequential\u548cmkl_intel_lp64\u4e09\u4e2a\u5e93":4,"mnist\u662f\u4e00\u4e2a\u5305\u542b\u6709":24,"movielens\u6570\u636e\u96c6":12,"movielens\u8bc4\u5206\u56de\u5f52":12,"name\u90fd\u662f":9,"no_avx\u7684":14,"osx\u6216\u8005\u662fwindows\u673a\u5668":9,"osx\u7684\u5b89\u88c5\u6587\u6863":9,"paddle\u5728\u8fdb\u884c\u8ba1\u7b97\u7684\u65f6\u5019\u4e3a\u4e86\u63d0\u5347\u8ba1\u7b97\u6027\u80fd":14,"paddle\u5b9a\u4e49\u7684\u53c2\u6570\u5305\u62ec":24,"paddle\u7684":10,"paddlepaddle\u4e2d":[0,2],"paddlepaddle\u4f7f\u7528\u5747\u503c0":14,"paddlepaddle\u4f7f\u7528\u8fd0\u884c\u65f6\u52a8\u6001\u8fde\u63a5cuda\u7684so":10,"paddlepaddle\u4fdd\u7559\u6dfb\u52a0\u53c2\u6570\u7684\u6743\u529b":24,"paddlepaddle\u53ef\u4ee5\u4f7f\u7528":4,"paddlepaddle\u53ef\u4ee5\u8bfb\u53d6python\u5199\u7684\u4f20\u8f93\u6570\u636e\u811a\u672c":13,"paddlepaddle\u5728\u8fd0\u884c\u65f6\u627e\u4e0d\u5230\u5bf9\u5e94\u7684config\u6587\u4ef6":10,"paddlepaddle\u5c06train":24,"paddlepaddle\u63a8\u8350\u4f7f\u7528docker\u8fdb\u884cpaddlepaddle\u7684\u90e8\u7f72\u548c":9,"paddlepaddle\u63d0\u4f9b\u4e86docker\u7684\u4f7f\u7528\u955c\u50cf":9,"paddlepaddle\u63d0\u4f9b\u6570\u4e2a\u9884\u7f16\u8bd1\u7684\u4e8c\u8fdb\u5236\u6765\u8fdb\u884c\u5b89\u88c5":8,"paddlepaddle\u63d0\u4f9b\u7684\u955c\u50cf\u5e76\u4e0d\u5305\u542b\u4efb\u4f55\u547d\u4ee4\u8fd0\u884c":9,"paddlepaddle\u652f\u6301\u975e\u5e38\u591a\u7684\u4f18\u5316\u7b97\u6cd5":14,"paddlepaddle\u652f\u6301sparse\u7684\u8bad\u7ec3":14,"paddlepaddle\u662f\u795e\u7ecf\u7f51\u7edc\u8bad\u7ec3\u5e73\u53f0":14,"paddlepaddle\u7684\u5185\u5b58\u5360\u7528\u4e3b\u8981\u5206\u4e3a\u5982\u4e0b\u51e0\u4e2a\u65b9\u9762":14,"paddlepaddle\u7684\u53c2\u6570\u4f7f\u7528\u540d\u5b57":14,"paddlepaddle\u7684\u6570\u636e\u5305\u62ec\u56db\u79cd\u4e3b\u8981\u7c7b\u578b":24,"paddlepaddle\u7684\u7f16\u8bd1\u9009\u9879\u53ef\u4ee5\u5728\u8c03\u7528cmake\u7684\u65f6\u5019\u8bbe\u7f6e":4,"paddlepaddle\u7684\u7f16\u8bd1\u9009\u9879\u662f\u53ef\u4ee5\u63a7\u5236paddlepaddle\u751f\u6210cpu":4,"paddlepaddle\u7684dock":3,"paddlepaddle\u7684python\u9884\u6d4b\u63a5\u53e3":26,"paddlepaddle\u7684ubuntu\u5b89\u88c5\u5305\u5206\u4e3a\u56db\u4e2a\u7248\u672c":10,"paddlepaddle\u76ee\u524d\u4f7f\u7528swig\u5bf9\u5176\u5e38\u7528\u7684\u9884\u6d4b\u63a5\u53e3\u8fdb\u884c\u4e86\u5c01\u88c5":27,"paddlepaddle\u76ee\u524d\u63d0\u4f9b\u4e24\u79cd\u53c2\u6570\u521d\u59cb\u5316\u7684\u65b9\u5f0f":14,"paddlepaddle\u76ee\u524d\u652f\u6301\u4f7f\u7528deb\u5305\u5b89\u88c5":10,"paddlepaddle\u8d1f\u8d23\u5b8c\u6210\u4fe1\u606f\u548c\u68af\u5ea6\u5728\u65f6\u95f4\u5e8f\u5217\u4e0a\u7684\u4f20\u64ad":2,"paddlepaddle\u8d1f\u8d23\u5b8c\u6210\u4fe1\u606f\u548c\u8bef\u5dee\u5728\u65f6\u95f4\u5e8f\u5217\u4e0a\u7684\u4f20\u64ad":2,"paddlepaddle\u8fd0\u884c\u65f6\u5982\u679c\u6ca1\u6709\u5bfb\u627e\u5230cuda\u7684driv":10,"paddlepaddle\u9700\u8981\u7528\u6237\u5728\u7f51\u7edc\u914d\u7f6e":23,"period\u8bbe\u7f6e\u6253\u5370\u53c2\u6570\u4fe1\u606f\u7b49":13,"process\u51fd\u6570":24,"process\u51fd\u6570\u662f\u5b9e\u73b0\u6570\u636e\u8f93\u5165\u7684\u4e3b\u51fd\u6570":24,"process\u51fd\u6570\u8c03\u7528\u591a\u6b21":24,"pserver\u4e3apaddlepaddle\u7684paramet":17,"pserver\u7684\u547d\u4ee4\u884c\u53c2\u6570":17,"pserver\u7ec4\u5408\u4f7f\u7528":17,"py\u6587\u4ef6\u7ed9\u51fa\u4e86\u5b8c\u6574\u4f8b\u5b50":13,"pydataprovider2\u4f1a\u5c3d\u91cf\u4f7f\u7528\u5185\u5b58":24,"pydataprovider2\u6587\u6863":27,"pydataprovider2\u7684\u4f7f\u7528":23,"pydataprovider\u4f7f\u7528\u7684\u662f\u5f02\u6b65\u52a0\u8f7d":14,"pydataprovider\u662fpaddlepaddle\u4f7f\u7528python\u63d0\u4f9b\u6570\u636e\u7684\u63a8\u8350\u63a5\u53e3":24,"python\u5305":9,"python\u53ef\u4ee5\u89e3\u9664\u6389\u5185\u90e8\u53d8\u91cf\u7684\u5f15\u7528":24,"python\u7684":9,"python\u7684swig\u63a5\u53e3\u53ef\u4ee5\u65b9\u4fbf\u8fdb\u884c\u9884\u6d4b\u548c\u5b9a\u5236\u5316\u8bad\u7ec3":4,"return":[1,13,24],"rnn\u603b\u662f\u5f15\u7528\u4e0a\u4e00\u65f6\u523b\u9884\u6d4b\u51fa\u7684\u8bcd\u7684\u8bcd\u5411\u91cf":2,"search\u7684\u751f\u6210":1,"seq\u53c2\u6570\u5fc5\u987b\u4e3afals":2,"seq\u540e":1,"seq\u5c42":1,"seq\u7684\u4f7f\u7528\u793a\u4f8b\u5982\u4e0b":0,"seq\u7c7b\u4f3c":0,"sequence\u7c7b\u578b":1,"server\u8fdb\u7a0b":17,"sh\u662fdocker":3,"shuffle\u8bad\u7ec3\u6570\u636e":24,"size\u53ef\u80fd\u4f1a\u5bf9\u8bad\u7ec3\u7ed3\u679c\u4ea7\u751f\u5f71\u54cd":14,"size\u672c\u8eab\u662f\u795e\u7ecf\u7f51\u7edc\u7684\u8d85\u53c2\u6570":14,"slot\u662finteg":1,"softmax\u8f93\u51fa":13,"sparse\u8bad\u7ec3\u7684\u6587\u6863":14,"sparse\u8bad\u7ec3\u9700\u8981\u8bad\u7ec3\u7279\u5f81\u662f":14,"state\u505a\u4e86\u4e00\u4e2a\u5168\u94fe\u63a5":1,"step\u4e2d":1,"step\u51fd\u6570\u4e2d\u7684memori":2,"step\u51fd\u6570\u5185\u90e8\u53ef\u4ee5\u81ea\u7531\u7ec4\u5408paddlepaddle\u652f\u6301\u7684\u5404\u79cdlay":2,"step\u7684recurr":1,"string\u7684\u683c\u5f0f\u6253\u5370\u51fa\u6765":17,"subseq\u7684\u6bcf\u4e2a\u5143\u7d20\u662f\u4e00\u4e2a0\u5c42\u5e8f\u5217":0,"swig_paddle\u63a5\u53d7\u7684\u539f\u59cb\u6570\u636e\u662fc":27,"tag\u5206\u522b\u4e3a":9,"train\u5373\u4e3apaddlepaddle\u7684\u8bad\u7ec3\u8fdb\u7a0b":17,"train\u5b8c\u6210\u5355\u673a\u591a\u663e\u5361\u591a\u7ebf\u7a0b\u7684\u8bad":17,"train\u7684\u547d\u4ee4\u884c\u53c2\u6570":17,"true":[1,14],"true\u7684memory\u65f6":1,"types\u7684\u8be6\u7ec6\u7528\u6cd5":1,"ubuntu\u7684deb\u5b89\u88c5\u5305\u7b49":8,"v2\u4e4b\u540e\u7684\u4efb\u4f55\u4e00\u4e2acudnn\u7248\u672c\u6765\u7f16\u8bd1\u8fd0\u884c":4,"value\u5373\u4e3asoftmax\u5c42\u7684\u8f93\u51fa":27,"value\u662f\u7279\u5f81\u503c":24,"value\u7c7b\u578b":1,"var":9,"vector\u8868\u793a\u7684\u6bcf\u4e2a\u5355\u8bcd":13,"version\u53ef\u4ee5\u6253\u5370\u51fapaddle\u7684\u7248\u672c\u4fe1\u606f\u548c\u7f16\u8bd1\u7684\u9009\u9879":22,"version\u53ef\u4ee5\u6253\u5370\u51fapaddlepaddle\u7684\u7248\u672c\u548c\u7f16\u8bd1\u65f6\u4fe1\u606f":17,"version\u7684\u547d\u4ee4\u884c\u53c2\u6570":17,"yield\u6587\u672c\u4fe1\u606f\u548c\u7c7b\u522bid":13,__main__:27,__name__:27,abov:24,act:[1,13,14],act_typ:13,activ:13,adadelta:[13,14],adagrad:13,adam:13,adamoptim:13,afi:24,agg_level:[0,1],aggregatelevel:[0,1],all:[2,24],allow:13,alreadi:10,also:13,append:[1,24],apt:[9,10],arg:[3,13,24],around:24,arrai:27,assert:27,atla:4,atlas_root:4,averag:1,avg:13,avgcost:13,avgpool:[0,1,13],avx:[9,14],bag:13,baidu:[9,10],batch:13,batch_siz:[1,13,14],batchsiz:1,beam:1,beam_search:2,befor:14,bias_attr:[1,14],bias_param:14,binari:13,bla:4,bool:13,boot:2,boot_lay:1,both:13,bow:13,build:[3,9],cach:[13,14,23],cache_pass_in_mem:[13,14,24],cachetyp:[13,14,24],calc_batch_s:24,call:13,can:13,can_over_batch_s:24,cat:9,categori:13,check:[1,10,24],check_fail_continu:24,chines:12,chpasswd:9,classif:13,classification_cost:[1,13,14],classification_error_evalu:13,close:24,cmake:4,cmd:9,cnn:13,code:[3,14,24,27],com:[9,10],comment:[1,13],compil:[10,22],conf:[1,27],config:[10,13],config_arg:13,config_pars:27,connect:13,contain:[13,24],context:24,context_len:13,context_start:13,convert:[13,24,27],couldn:10,cpp:[10,13],cpu:[9,10,24],cpuinfo:9,createfromconfigproto:27,cross:[13,14],cuda_so:9,cudastat:10,cudasuccess:10,cudnn:4,cudnn_root:4,cudnnv5:4,current:[13,24],currentcost:13,currentev:13,dalla:24,data:[1,10],data_config:27,data_initialz:13,data_lay:[1,13,14,24],dataprovid:[13,14],dataprovider_bow:13,dataprovider_emb:13,dataproviderconvert:27,dataset:13,deb:10,debian:10,decod:2,decor:24,def:[1,13,14,24,27],defin:[13,14,24],define_py_data_sources2:[13,24],delar:13,demo:[9,13],dense_vector:[24,27],describ:13,descript:26,detail:26,dev:9,devel:9,devic:9,devices:9,dict:[13,24],dict_dim:[1,14],dict_fil:[1,13],dictionai:13,dictionari:[13,14,24],dictrionari:13,differ:13,dim:13,dimens:[13,14],dir:13,doc:27,documentari:24,dpkg:10,driver:10,dso_handl:10,dtype:27,dump_config:17,dure:[13,24],dwith_avx:14,dynam:24,each:[13,24],each_pixel_str:24,each_sequence:[0,1],each_word:24,echo:9,either:13,els:[1,9,13],emb:[1,13,14],emb_group:1,emb_sum:14,embed:12,embedding_lay:[1,13,14],entropi:13,enumer:13,equal:1,error:[10,13],error_clipping_threshold:1,etc:9,eval:13,exampl:13,expand:[0,1],expand_a:[0,1],expand_level:[0,1],expandlevel:[0,1],expose:9,extralayerattribut:1,f0831:10,fail:[1,10],fals:[13,14],fc_layer:[1,13,14],fc_param:14,fdata:1,featur:[13,24],feature_a:14,feature_b:14,festiv:24,file:[13,24],file_list:24,file_nam:[1,13],filenam:[14,24],fill:13,find:10,first:[0,13],float32:27,fly:13,forwardtest:27,framework:13,from:[2,9,13,14,24,27],from_sequence:[0,1],from_timestep:0,full_matrix_project:1,fulli:13,func:24,gate_act:1,gdebi:10,gener:[13,24],generatedinput:2,get:[9,10,13,24],get_config_arg:13,get_data:13,get_sample_from_lin:14,github:10,give:24,given:13,globe:24,gpu:[9,10],gradient_clipping_threshold:13,gradientmachin:27,grep:9,group:1,group_input:1,gru:13,gru_siz:13,gserver:1,hassubseq:1,help:27,hidden:14,hidden_a:14,hidden_b:14,hidden_dim:1,hierach:2,hint:27,hl_cuda_devic:10,hl_dso_load:10,hook2:1,hook:1,host:9,hot:13,hous:24,howardjohnson:1,http:10,ignor:24,imag:[9,14],imagenet:12,img:24,inarg:27,includ:13,init:13,init_hook:[1,13,23],init_model_path:13,initi:[13,24],initial_max:14,initial_mean:14,initial_min:14,initial_std:14,initpaddl:27,inner:14,inner_mem:1,inner_rnn_output:1,inner_rnn_st:1,inner_step:1,input:[0,1,2,13,14,24],input_typ:[1,13,14,23],instal:5,insuffici:10,integ:[13,24],integer_sequ:[14,24],integer_valu:[1,13,14,24],integer_value_sequ:[1,13],integer_value_sub_sequ:1,invok:24,ipt:14,is_predict:13,is_train:24,isinst:27,iterat:24,job:13,join:1,kernel:9,kwarg:[1,13,24],l2regular:13,label:[1,13,14,24],label_dim:[1,13],label_list:1,lake:24,last:[0,1],later:13,latest:[3,9],layer1:0,layer2:0,layer:[0,1,2,13],ld_library_path:10,learning_method:13,learning_r:[13,14],len:[1,13,24],level:2,lib64:[9,10],lib:4,libcuda:9,libnvidia:9,librari:10,line:1,link:2,list:[13,23,24],load_data_arg:27,loadparamet:27,local:[4,10],log_period:13,logger:24,look:[13,24],loss:13,lowest_dl_speed:3,lstm:[1,13],lstm_averag:1,lstm_expand:1,lstm_group:1,lstm_group_input:1,lstm_input:1,lstm_last:1,lstm_layer_attr:1,lstm_nest_group:1,lstm_output:1,lstm_size:13,lstmemori:1,lstmemory_group:1,mac:9,machin:2,main:27,maintainer:9,make:[10,24],make_diagram:17,maxid:13,maxid_lay:13,maxpool:0,mean:[13,14],mem:1,memori:1,merge_model:17,method:24,min_pool_s:[14,24],mixed_lay:1,mkdir:9,mkl:4,mkl_core:4,mkl_root:4,mnist:24,mnist_model:27,mnist_provid:24,mnist_train:24,model_config:27,modul:[13,24],momentum:[13,14],movi:24,must:10,name:[1,9,13,14,24],necessari:13,need:13,neg:[13,24],nest:1,net:9,neural:2,next:24,no_cache:24,no_sequence:24,noavx:[9,10],none:[13,24,27],normal:9,note:10,now:2,nullptr:10,num:13,num_pass:13,nvidia:9,obj:[13,24],object:[13,24],off:[3,4,10,14,22],omit:[13,14],on_init:24,onli:[2,13],open:[1,13,14,24],openbla:4,openblas_root:4,openssh:9,opt:4,other:13,out:[1,2],outer:1,outer_mem:1,outer_rnn_st:1,outer_step:1,outlin:26,output:[1,13,14],outsid:24,paddl:[1,3,9,10,13,14,17],paddle_gpu:3,paddle_ssh:9,paddle_ssh_machin:9,paddledev:9,paddlepaddl:[9,10,22,27],param_attr:14,paramattr:14,paramet:13,parse_config:27,pass:[13,14,24],path:[10,13],period:13,permitrootlogin:9,pixel:24,pixels_float:24,pixels_str:24,place:24,pleas:10,pool:0,pool_siz:24,pooling_typ:[0,1,13,14],posit:[13,24],pred:13,predict_output_dir:13,predict_sampl:27,preprocess:13,print:27,proc:[9,14],process2:1,process:[1,13,14,24],process_pr:13,process_seq:1,process_subseq:1,properli:13,provid:1,pserver:14,pull:9,put:13,py_paddl:[9,27],pydataprovid:[14,23],pydataprovider2:[13,24,27],pydataproviderwrapp:13,python:[13,14],quick_start:13,rang:13,rank:13,rare:24,read:[13,24],read_next_from_fil:14,real_process:24,recurrent_group:[1,2],refer:23,reference_cblas_root:4,reffer:4,regular:13,releas:10,repres:13,represent:13,resnet:12,result:[13,24],revers:2,rmsprop:13,rnn:2,rnn_data_provid:1,rnn_state:1,roce:9,root:9,run:9,runtim:[10,24],same:[13,24],sampl:[13,24],save:[13,24],save_dir:13,saw:24,sbin:9,script:3,second:13,sed:9,see:13,sentanc:14,sentenc:24,sentiment:24,sentimental_provid:24,separ:13,seq:[0,1],seq_pool:0,seq_typ:24,seqlastin:1,sequel:24,sequenc:[1,2],sequence:24,sequence_conv_pool:13,sequence_layer_group:1,sequence_nest_layer_group:1,sequence_nest_rnn:1,sequence_nest_rnn_readonly_memori:1,sequence_rnn:1,sequencegen:1,sequencestartposit:1,sequencetyp:24,server:9,set:[1,13,14,24],setup:13,should:2,should_shuffl:24,shuf:14,shuffl:14,sigmoidactiv:1,simple_gru:13,simple_lstm:13,size:[1,13,14,24],softmax:[13,14],softmax_param:14,softmaxactiv:[1,13,14],sourc:13,spars:[13,14],sparse_binary_vector:[13,14,24],sparse_float_vector:24,sparse_upd:14,sparse_vector:14,specifi:[10,13],split:[1,13,24],src_root:27,ssh:9,sshd:9,sshd_config:9,stat:13,state:2,state_act:1,staticinput:2,step:[1,2],stop:9,store:13,string:24,strip:[1,13],structur:13,stun:24,sub:1,sub_sequence:24,subseq:[0,2],subsequenceinput:1,sudo:10,sumpool:14,support:9,sure:10,swig_paddl:27,system:14,tag:3,take:24,tanhactiv:1,tbd:[1,25],team:9,test:[1,13,23],test_data:27,test_list:[13,24],test_recurrentgradientmachin:1,text:[13,24],text_conv:13,them:13,thi:[13,24],thing:24,timestep:0,tmp:24,tour_train_wdseg:1,train:10,train_list:[13,24],trainer:[13,24,27],trainer_config:[13,23,24,27],trainer_config_help:[13,14,24],trainer_count:14,trainerintern:13,trainermain:10,travel:24,trn:13,tst:13,turn:2,two:13,txt:[13,24],type:[13,24],unk_idx:13,updat:9,use:[13,26],use_dynamic_ord:24,use_gpu:[13,14,27],usepam:9,user:13,usr:[4,9,10],valid:10,valu:[1,13,24,27],version:[9,10],via:10,want:24,what:13,when:24,which:13,whole:24,wilder:24,window:9,with_avx:[4,10,22],with_doc:4,with_doc_cn:4,with_doubl:[10,22],with_double:4,with_dso:4,with_gflag:[10,22],with_gflags:4,with_glog:[4,10,22],with_gpu:[3,4,10,22],with_metric_learn:[10,22],with_predict_sdk:[10,22],with_python:[4,10,22],with_rdma:[4,10,22],with_style_check:4,with_swig_py:4,with_testing:4,with_tim:[10,22],with_timer:4,without:9,wonder:24,word2vec:14,word:[1,2,12],word_dict:[1,13],word_dim:[1,13],word_id:[14,24],word_slot:1,word_slot_list:1,word_vector:13,xarg:9,yield:[1,13,14,24],you:[10,24],your_host_machine:9,your_param_name:14},titles:["\u652f\u6301\u53cc\u5c42\u5e8f\u5217\u4f5c\u4e3a\u8f93\u5165\u7684Layer","\u53cc\u5c42RNN\u914d\u7f6e\u4e0e\u793a\u4f8b","Recurrent Group\u6559\u7a0b","\u6784\u5efaPaddlePaddle Docker Image","\u8bbe\u7f6ePaddlePaddle\u7684\u7f16\u8bd1\u9009\u9879","\u4f7f\u7528cmake\u7f16\u8bd1PaddlePaddle","\u5b89\u88c5\u7f16\u8bd1PaddlePaddle\u9700\u8981\u7684\u4f9d\u8d56","make\u548cmake install","\u7f16\u8bd1\u4e0e\u5b89\u88c5","\u5b89\u88c5PaddlePaddle\u7684Docker\u955c\u50cf","\u4f7f\u7528deb\u5305\u5728Ubuntu\u4e0a\u5b89\u88c5PaddlePaddle","\u96c6\u7fa4\u8bad\u7ec3","\u4f7f\u7528\u793a\u4f8b","PaddlePaddle\u5feb\u901f\u5165\u95e8\u6559\u7a0b","PaddlePaddle\u5e38\u89c1\u95ee\u9898","PaddlePaddle\u6587\u6863","&lt;no title&gt;","\u547d\u4ee4\u884c\u53c2\u6570","&lt;no title&gt;","&lt;no title&gt;","paddle pserver\u7684\u547d\u4ee4\u884c\u53c2\u6570","paddle train\u7684\u547d\u4ee4\u884c\u53c2\u6570","paddle version\u7684\u547d\u4ee4\u884c\u53c2\u6570","PaddlePaddle\u7684\u6570\u636e\u63d0\u4f9b(DataProvider)\u4ecb\u7ecd","PyDataProvider2\u7684\u4f7f\u7528","\u81ea\u5b9a\u4e49\u4e00\u4e2aDataProvider","\u7528\u6237\u63a5\u53e3","PaddlePaddle\u7684Python\u9884\u6d4b\u63a5\u53e3"],titleterms:{"\u4e0b\u8f7d\u548c\u8fd0\u884cdocker\u955c\u50cf":9,"\u4ecb\u7ecd":23,"\u4f18\u5316\u7b97\u6cd5":13,"\u4f7f\u7528\u6307\u5357":15,"\u4f7f\u7528\u6982\u8ff0":13,"\u4f7f\u7528\u793a\u4f8b":12,"\u4f7f\u7528\u811a\u672c\u6784\u5efapaddlepaddl":3,"\u4f7f\u7528cmake\u7f16\u8bd1paddlepaddl":5,"\u4f7f\u7528deb\u5305\u5728ubuntu\u4e0a\u5b89\u88c5paddlepaddl":10,"\u5185\u5b58\u4e0d\u591f\u7528\u7684\u60c5\u51b5":24,"\u51cf\u5c11\u6570\u636e\u8f7d\u5165\u7684\u8017\u65f6":14,"\u51cf\u5c11dataprovider\u7f13\u51b2\u6c60\u5185\u5b58":14,"\u5229\u7528\u66f4\u591a\u7684\u8ba1\u7b97\u8d44\u6e90":14,"\u52a0\u901f\u8bad\u7ec3\u901f\u5ea6":14,"\u5377\u79ef\u6a21\u578b":13,"\u53c2\u6570\u5185\u5b58":14,"\u53c2\u8003":24,"\u53cc\u5c42rnn\u4ecb\u7ecd":2,"\u53cc\u5c42rnn\u7684\u4f7f\u7528":2,"\u53cc\u5c42rnn\u914d\u7f6e\u4e0e\u793a\u4f8b":1,"\u53cc\u8fdb\u53cc\u51fa":1,"\u53ef\u80fd\u7684\u5185\u5b58\u6cc4\u9732\u95ee\u9898":24,"\u53ef\u80fd\u9047\u5230\u7684\u95ee\u9898":10,"\u547d\u4ee4\u884c\u53c2\u6570":[13,17,26],"\u548c":0,"\u56fe\u50cf":12,"\u57fa\u672c\u539f\u7406":2,"\u5982\u4f55\u5171\u4eab\u53c2\u6570":14,"\u5982\u4f55\u51cf\u5c11paddlepaddle\u7684\u5185\u5b58\u5360\u7528":14,"\u5982\u4f55\u521d\u59cb\u5316\u53c2\u6570":14,"\u5982\u4f55\u52a0\u901fpaddlepaddle\u7684\u8bad\u7ec3\u901f\u5ea6":14,"\u5982\u4f55\u9009\u62e9sgd\u7b97\u6cd5\u7684\u5b66\u4e60\u7387":14,"\u5b89\u88c5":[8,13],"\u5b89\u88c5\u7f16\u8bd1paddlepaddle\u9700\u8981\u7684\u4f9d\u8d56":6,"\u5b89\u88c5paddlepaddle\u7684docker\u955c\u50cf":9,"\u5e38\u7528\u6a21\u578b":12,"\u5e38\u89c1\u95ee\u9898":15,"\u5e8f\u5217\u6a21\u578b\u6570\u636e\u63d0\u4f9b":24,"\u5f00\u53d1\u6307\u5357":15,"\u6027\u80fd\u95ee\u9898":9,"\u603b\u4f53\u6548\u679c\u603b\u7ed3":13,"\u6216\u8005\u662f":14,"\u63a8\u8350":12,"\u652f\u6301\u53cc\u5c42\u5e8f\u5217\u4f5c\u4e3a\u8f93\u5165\u7684layer":0,"\u6570\u636e\u5411\u6a21\u578b\u4f20\u9001":13,"\u6570\u636e\u63d0\u4f9b":26,"\u6570\u636e\u683c\u5f0f\u51c6\u5907":13,"\u65f6\u5e8f\u6a21\u578b":13,"\u6784\u5efapaddlepaddl":3,"\u6982\u8ff0":[0,2],"\u6a21\u578b\u4e2d\u7684\u914d\u7f6e":1,"\u6ce8\u610f\u4e8b\u9879":[9,24],"\u751f\u6210\u6d41\u7a0b\u7684\u4f7f\u7528\u65b9\u6cd5":2,"\u7528\u6237\u63a5\u53e3":26,"\u76f8\u5173\u6982\u5ff5":2,"\u793a\u4f8b1":1,"\u793a\u4f8b2":1,"\u793a\u4f8b3":1,"\u793a\u4f8b4":1,"\u795e\u7ecf\u5143\u6fc0\u6d3b\u5185\u5b58":14,"\u7b80\u5355\u7684\u4f7f\u7528\u573a\u666f":24,"\u7b97\u6cd5\u6559\u7a0b":15,"\u7f16\u8bd1":8,"\u7f16\u8bd1\u4e0e\u5b89\u88c5":8,"\u7f51\u7edc\u7ed3\u6784":13,"\u81ea\u5b9a\u4e49\u4e00\u4e2adataprovid":25,"\u81ea\u7136\u8bed\u8a00\u5904\u7406":12,"\u8bad\u7ec3\u6a21\u578b":13,"\u8bad\u7ec3\u6d41\u7a0b\u7684\u4f7f\u7528\u65b9\u6cd5":2,"\u8bbe\u7f6epaddlepaddle\u7684\u7f16\u8bd1\u9009\u9879":4,"\u8bcd\u5411\u91cf\u6a21\u578b":13,"\u8bfb\u53d6\u53cc\u5c42\u5e8f\u5217\u7684\u65b9\u6cd5":1,"\u8f93\u5165":2,"\u8f93\u5165\u4e0d\u7b49\u957f":1,"\u8f93\u5165\u793a\u4f8b":2,"\u8f93\u51fa":2,"\u8f93\u51fa\u65e5\u5fd7":13,"\u8fdc\u7a0b\u8bbf\u95ee\u95ee\u9898\u548c\u4e8c\u6b21\u5f00\u53d1":9,"\u903b\u8f91\u56de\u5f52\u6a21\u578b":13,"\u9047\u5230":14,"\u914d\u7f6e\u4e2d\u7684\u6570\u636e\u52a0\u8f7d\u5b9a\u4e49":13,"\u9644\u5f55":13,"\u96c6\u7fa4\u8bad\u7ec3":11,"\u975e\u6cd5\u6307\u4ee4":14,"\u9884\u6d4b":[13,26],"beam_search\u7684\u751f\u6210":1,"blas\u76f8\u5173\u7684\u7f16\u8bd1\u9009\u9879":4,"bool\u578b\u7684\u7f16\u8bd1\u9009\u9879":4,"config\u6587\u4ef6\u627e\u4e0d\u5230":10,"cudnn\u76f8\u5173\u7684\u7f16\u8bd1\u9009\u9879":4,"driver\u627e\u4e0d\u5230":10,"group\u6559\u7a0b":2,"make\u548cmak":7,"paddlepaddle\u5e38\u89c1\u95ee\u9898":14,"paddlepaddle\u5feb\u901f\u5165\u95e8\u6559\u7a0b":13,"paddlepaddle\u63d0\u4f9b\u7684docker\u955c\u50cf\u7248\u672c":9,"paddlepaddle\u6587\u6863":15,"paddlepaddle\u7684\u6570\u636e\u63d0\u4f9b":23,"paddlepaddle\u7684\u7f16\u8bd1\u9009\u9879":4,"paddlepaddle\u7684bool\u578b\u7f16\u8bd1\u9009\u9879":4,"paddlepaddle\u7684cblas\u7f16\u8bd1\u9009\u9879":4,"paddlepaddle\u7684python\u9884\u6d4b\u63a5\u53e3":27,"pserver\u7684\u547d\u4ee4\u884c\u53c2\u6570":20,"pydataprovider2\u7684\u4f7f\u7528":24,"python\u6570\u636e\u52a0\u8f7d\u811a\u672c":13,"so\u627e\u4e0d\u5230":10,"subseq\u95f4\u65e0memori":1,"subseq\u95f4\u6709memori":1,"train\u7684\u547d\u4ee4\u884c\u53c2\u6570":21,"version\u7684\u547d\u4ee4\u884c\u53c2\u6570":22,algorithm:13,appendix:13,architectur:13,argument:13,cach:24,command:13,configur:13,content:14,convolut:13,cuda:[4,10],data:13,dataprovid:23,docker:3,expand_lay:0,first_seq:0,illeg:14,image:3,init_hook:24,input_typ:24,instal:7,install:13,instruct:14,last_seq:0,libcudart:10,libcudnn:10,line:13,log:13,logist:13,memori:2,model:13,network:13,optimiz:13,overview:13,paddl:[20,21,22],pooling_lay:0,predict:13,prepar:13,provid:[13,24],recurr:2,refer:24,regress:13,script:13,sequenc:13,summari:13,time:13,train:13,transfer:13,vector:13,word:13}})
\ No newline at end of file