Deploy to GitHub Pages: b156bbc8

15e1e23b · Travis CI · 6a094aac · 15e1e23b · 15e1e23b · 15e1e23b
24 changed file
--- a/develop/doc/api/v2/fluid/evaluator.html
+++ b/develop/doc/api/v2/fluid/evaluator.html
@@ -220,6 +220,75 @@
 <h1>Evaluator<a class="headerlink" href="#evaluator" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="id1">
 <h2>Evaluator<a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.evaluator.</code><code class="descname">Evaluator</code><span class="sig-paren">(</span><em>name</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Base Class for all evaluators</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 last simple">
+<li><strong>name</strong> (<em>str</em>) &#8211; The name of evaluator. such as, &#8220;accuracy&#8221;. Used for generate
+temporary variable name.</li>
+<li><strong>main_program</strong> (<em>Program</em><em>, </em><em>optional</em>) &#8211; The evaluator should be added to this
+main_program. Default default_main_program()</li>
+<li><strong>startup_program</strong> (<em>Program</em><em>, </em><em>optional</em>) &#8211; The parameter should be added to this
+startup_program. Default default_startup_program()</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+<dl class="attribute">
+<dt id="paddle.v2.fluid.evaluator.Evaluator.states">
+<code class="descname">states</code><a class="headerlink" href="#paddle.v2.fluid.evaluator.Evaluator.states" title="Permalink to this definition">¶</a></dt>
+<dd><p><em>list</em> &#8211; The list of state variables. states will be reset to zero
+when <cite>reset</cite> is invoked.</p>
+</dd></dl>
+
+<dl class="attribute">
+<dt id="paddle.v2.fluid.evaluator.Evaluator.metrics">
+<code class="descname">metrics</code><a class="headerlink" href="#paddle.v2.fluid.evaluator.Evaluator.metrics" title="Permalink to this definition">¶</a></dt>
+<dd><p><em>list</em> &#8211; The list of metrics variables. They will be calculate
+every mini-batch</p>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">reset</code><span class="sig-paren">(</span><em>executor</em>, <em>reset_program=None</em><span class="sig-paren">)</span></dt>
+<dd><p>reset metric states at the begin of each pass/user specified batch</p>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">eval</code><span class="sig-paren">(</span><em>executor</em>, <em>eval_program=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Evaluate the statistics merged by multiple mini-batches.</p>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">create_state</code><span class="sig-paren">(</span><em>suffix</em>, <em>dtype</em>, <em>shape</em><span class="sig-paren">)</span></dt>
+<dd><p>Create state variable.</p>
+<p>NOTE: It is not a public API.</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 last simple">
+<li><strong>suffix</strong> (<em>str</em>) &#8211; the state suffix.</li>
+<li><strong>dtype</strong> (<em>str|core.DataType</em>) &#8211; the state data type</li>
+<li><strong>shape</strong> (<em>tuple|list</em>) &#8211; the shape of state</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+<p>Returns: State variable</p>
+</dd></dl>
+
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/api/v2/fluid/initializer.html
+++ b/develop/doc/api/v2/fluid/initializer.html
@@ -220,21 +220,90 @@
 <h1>Initializer<a class="headerlink" href="#initializer" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="id1">
 <h2>Initializer<a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">Initializer</code></dt>
+<dd><p>Base class for variable initializers</p>
+<p>Defines the common interface of variable initializers.
+They add operations to the init program that are used
+to initialize variables. Users should not use this class
+directly, but need to use one of its implementations.</p>
+</dd></dl>
+
 </div>
 <div class="section" id="constantinitializer">
 <h2>ConstantInitializer<a class="headerlink" href="#constantinitializer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">ConstantInitializer</code><span class="sig-paren">(</span><em>value=0.0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the constant initializer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="uniforminitializer">
 <h2>UniformInitializer<a class="headerlink" href="#uniforminitializer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">UniformInitializer</code><span class="sig-paren">(</span><em>low=-1.0</em>, <em>high=1.0</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the random uniform distribution initializer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="normalinitializer">
 <h2>NormalInitializer<a class="headerlink" href="#normalinitializer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">NormalInitializer</code><span class="sig-paren">(</span><em>loc=0.0</em>, <em>scale=1.0</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the  random Normal(Gaussian) distribution initializer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="xavierinitializer">
 <h2>XavierInitializer<a class="headerlink" href="#xavierinitializer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">XavierInitializer</code><span class="sig-paren">(</span><em>uniform=True</em>, <em>fan_in=None</em>, <em>fan_out=None</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the Xavier initializer</p>
+<p>This class implements the Xavier weight initializer from the paper
+Understanding the difficulty of training deep feedforward neural
+networks[1] by Xavier Glorot and Yoshua Bengio.</p>
+<p>This initializer is designed to keep the scale of the gradients
+approximately same in all the layers. In case of Uniform distribution,
+the range is [-x, x], where x = sqrt(6 / (fan_in + fan_out)).
+In case of Normal distribution, the mean is 0 and the standard deviation
+is sqrt(2/ (fan_in + fan_out)).</p>
+<p class="rubric">References</p>
+<dl class="docutils">
+<dt>[1] Understanding the difficulty of training deep feedforward neural</dt>
+<dd>networks. International conference on artificial intelligence and
+statistics.
+(<a class="reference external" href="http://proceedings.mlr.press/v9/glorot10a.html">http://proceedings.mlr.press/v9/glorot10a.html</a>)</dd>
+</dl>
+</dd></dl>
+
 </div>
 <div class="section" id="msrainitializer">
 <h2>MSRAInitializer<a class="headerlink" href="#msrainitializer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">MSRAInitializer</code><span class="sig-paren">(</span><em>uniform=True</em>, <em>fan_in=None</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the MSRA initializer a.k.a. Kaiming Initializer</p>
+<p>This class implements the weight initialization from the paper
+Delving Deep into Rectifiers: Surpassing Human-Level Performance on
+ImageNet Classification[1] by Kaiming He, Xiangyu Zhang, Shaoqing Ren
+and Jian Sun. This is a robust initialization method that particularly
+considers the rectifier nonlinearities. In case of Uniform distribution,
+the range is [-x, x], where x = sqrt(6 / fan_in). In case of Normal
+distribution, the mean is 0 and the standard deviation
+is sqrt(2/ fan_in).</p>
+<p class="rubric">References</p>
+<dl class="docutils">
+<dt>[1] Delving Deep into Rectifiers: Surpassing Human-Level Performance</dt>
+<dd>on ImageNet Classification
+(<a class="reference external" href="https://arxiv.org/abs/1502.01852">https://arxiv.org/abs/1502.01852</a>)</dd>
+</dl>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/api/v2/fluid/io.html
+++ b/develop/doc/api/v2/fluid/io.html
@@ -220,6 +220,23 @@
 <h1>IO<a class="headerlink" href="#io" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="is-parameter">
 <h2>is_parameter<a class="headerlink" href="#is-parameter" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.io.</code><code class="descname">is_parameter</code><span class="sig-paren">(</span><em>var</em><span class="sig-paren">)</span></dt>
+<dd><p>Check whether the variable is a Parameter.</p>
+<p>This function checks whether the input variable is a Parameter.</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"><strong>var</strong> &#8211; The input variable.</td>
+</tr>
+<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body">boolean result whether the variable is a Parameter.</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/api/v2/fluid/layers.html
+++ b/develop/doc/api/v2/fluid/layers.html
--- a/develop/doc/api/v2/fluid/nets.html
+++ b/develop/doc/api/v2/fluid/nets.html
@@ -220,18 +220,118 @@
 <h1>Nets<a class="headerlink" href="#nets" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="simple-img-conv-pool">
 <h2>simple_img_conv_pool<a class="headerlink" href="#simple-img-conv-pool" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">simple_img_conv_pool</code><span class="sig-paren">(</span><em>input</em>, <em>num_filters</em>, <em>filter_size</em>, <em>pool_size</em>, <em>pool_stride</em>, <em>act</em>, <em>param_attr=None</em>, <em>pool_type='max'</em>, <em>use_cudnn=True</em><span class="sig-paren">)</span></dt>
+<dd></dd></dl>
+
 </div>
 <div class="section" id="img-conv-group">
 <h2>img_conv_group<a class="headerlink" href="#img-conv-group" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">img_conv_group</code><span class="sig-paren">(</span><em>input</em>, <em>conv_num_filter</em>, <em>pool_size</em>, <em>conv_padding=1</em>, <em>conv_filter_size=3</em>, <em>conv_act=None</em>, <em>param_attr=None</em>, <em>conv_with_batchnorm=False</em>, <em>conv_batchnorm_drop_rate=None</em>, <em>pool_stride=1</em>, <em>pool_type=None</em>, <em>use_cudnn=True</em><span class="sig-paren">)</span></dt>
+<dd><p>Image Convolution Group, Used for vgg net.</p>
+</dd></dl>
+
 </div>
 <div class="section" id="sequence-conv-pool">
 <h2>sequence_conv_pool<a class="headerlink" href="#sequence-conv-pool" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">sequence_conv_pool</code><span class="sig-paren">(</span><em>input</em>, <em>num_filters</em>, <em>filter_size</em>, <em>param_attr=None</em>, <em>act='sigmoid'</em>, <em>pool_type='max'</em><span class="sig-paren">)</span></dt>
+<dd></dd></dl>
+
 </div>
 <div class="section" id="glu">
 <h2>glu<a class="headerlink" href="#glu" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">glu</code><span class="sig-paren">(</span><em>input</em>, <em>dim=-1</em><span class="sig-paren">)</span></dt>
+<dd><p>The gated linear unit composed by split, sigmoid activation and elementwise
+multiplication. Specifically, Split the input into two equal sized parts
+<span class="math">\(a\)</span> and <span class="math">\(b\)</span> along the given dimension and then compute as
+following:</p>
+<blockquote>
+<div><div class="math">
+\[{GLU}(a, b)= a \otimes \sigma(b)\]</div>
+</div></blockquote>
+<p>Refer to <a class="reference external" href="https://arxiv.org/pdf/1612.08083.pdf">Language Modeling with Gated Convolutional Networks</a>.</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>input</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+<li><strong>dim</strong> (<em>int</em>) &#8211; The dimension along which to split. If <span class="math">\(dim &lt; 0\)</span>, the
+dimension to split along is <span class="math">\(rank(input) + dim\)</span>.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first">The Tensor variable with half the size of input.</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">Return type:</th><td class="field-body"><p class="first last">Variable</p>
+</td>
+</tr>
+</tbody>
+</table>
+<p class="rubric">Examples</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># x is a Tensor variable with shape [3, 6, 9]</span>
+<span class="n">fluid</span><span class="o">.</span><span class="n">nets</span><span class="o">.</span><span class="n">glu</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">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>  <span class="c1"># shape of output: [3, 3, 9]</span>
+</pre></div>
+</div>
+</dd></dl>
+
 </div>
 <div class="section" id="dot-product-attention">
 <h2>dot_product_attention<a class="headerlink" href="#dot-product-attention" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">dot_product_attention</code><span class="sig-paren">(</span><em>querys</em>, <em>keys</em>, <em>values</em><span class="sig-paren">)</span></dt>
+<dd><p>The dot-product attention.</p>
+<p>Attention mechanism can be seen as mapping a query and a set of key-value
+pairs to an output. The output is computed as a weighted sum of the values,
+where the weight assigned to each value is computed by a compatibility
+function (dot-product here) of the query with the corresponding key.</p>
+<p>The dot-product attention can be implemented through (batch) matrix
+multipication as follows:</p>
+<blockquote>
+<div><div class="math">
+\[Attention(Q, K, V)= softmax(QK^\mathrm{T})V\]</div>
+</div></blockquote>
+<p>Refer to <a class="reference external" href="https://arxiv.org/pdf/1706.03762.pdf">Attention Is All You Need</a>.</p>
+<p>Note that batch data containing sequences with different lengths is not
+supported by this because of the (batch) matrix multipication.</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>query</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+<li><strong>key</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+<li><strong>value</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first">The Tensor variables representing the output and attention scores.</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">Return type:</th><td class="field-body"><p class="first last">tuple</p>
+</td>
+</tr>
+</tbody>
+</table>
+<p class="rubric">Examples</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># Suppose q, k, v are tensor variables with the following shape:</span>
+<span class="c1"># q: [3, 5, 9], k: [3, 6, 9], v: [3, 6, 10]</span>
+<span class="n">out</span><span class="p">,</span> <span class="n">attn_scores</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">nets</span><span class="o">.</span><span class="n">dot_product_attention</span><span class="p">(</span><span class="n">q</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span>
+<span class="n">out</span><span class="o">.</span><span class="n">shape</span>  <span class="c1"># [3, 5, 10]</span>
+<span class="n">attn_scores</span><span class="o">.</span><span class="n">shape</span>  <span class="c1"># [3, 5, 6]</span>
+</pre></div>
+</div>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/api/v2/fluid/optimizer.html
+++ b/develop/doc/api/v2/fluid/optimizer.html
@@ -220,24 +220,105 @@
 <h1>Optimizer<a class="headerlink" href="#optimizer" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="id1">
 <h2>Optimizer<a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">Optimizer</code><span class="sig-paren">(</span><em>global_step=None</em>, <em>regularization=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Optimizer Base class.</p>
+<p>Define the common interface of an optimizer.
+User should not use this class directly,
+but need to use one of it&#8217;s implementation.</p>
+<dl class="method">
+<dt>
+<code class="descname">create_optimization_pass</code><span class="sig-paren">(</span><em>parameters_and_grads</em>, <em>loss</em>, <em>startup_program=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Add optimization operators to update gradients to variables.</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>loss</strong> &#8211; the target that this optimization is for.</li>
+<li><strong>parameters_and_grads</strong> &#8211; a list of (variable, gradient) pair to update.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first">a list of operators that will complete one step of
+optimization. This will include parameter update ops, global step
+update ops and any other custom ops required by subclasses to manage
+their internal state.
+:param startup_program:</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">Return type:</th><td class="field-body"><p class="first last">return_op_list</p>
+</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">minimize</code><span class="sig-paren">(</span><em>loss</em>, <em>startup_program=None</em>, <em>parameter_list=None</em>, <em>no_grad_set=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Add operations to minimize <cite>loss</cite> by updating <cite>parameter_list</cite>.</p>
+<p>This method combines interface <cite>append_backward()</cite> and
+<cite>create_optimization_pass()</cite> into one.</p>
+</dd></dl>
+
+</dd></dl>
+
 </div>
 <div class="section" id="sgdoptimizer">
 <h2>SGDOptimizer<a class="headerlink" href="#sgdoptimizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">SGDOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple SGD optimizer without any state.</p>
+</dd></dl>
+
 </div>
 <div class="section" id="momentumoptimizer">
 <h2>MomentumOptimizer<a class="headerlink" href="#momentumoptimizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">MomentumOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>momentum</em>, <em>use_nesterov=False</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple Momentum optimizer with velocity state</p>
+</dd></dl>
+
 </div>
 <div class="section" id="adagradoptimizer">
 <h2>AdagradOptimizer<a class="headerlink" href="#adagradoptimizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">AdagradOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>epsilon=1e-06</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple Adagrad optimizer with moment state</p>
+</dd></dl>
+
 </div>
 <div class="section" id="adamoptimizer">
 <h2>AdamOptimizer<a class="headerlink" href="#adamoptimizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">AdamOptimizer</code><span class="sig-paren">(</span><em>learning_rate=0.001</em>, <em>beta1=0.9</em>, <em>beta2=0.999</em>, <em>epsilon=1e-08</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the Adam Optimizer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="adamaxoptimizer">
 <h2>AdamaxOptimizer<a class="headerlink" href="#adamaxoptimizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">AdamaxOptimizer</code><span class="sig-paren">(</span><em>learning_rate=0.001</em>, <em>beta1=0.9</em>, <em>beta2=0.999</em>, <em>epsilon=1e-08</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the Adamax Optimizer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="decayedadagradoptimizer">
 <h2>DecayedAdagradOptimizer<a class="headerlink" href="#decayedadagradoptimizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">DecayedAdagradOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>decay=0.95</em>, <em>epsilon=1e-06</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple Decayed Adagrad optimizer with moment state</p>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/api/v2/fluid/profiler.html
+++ b/develop/doc/api/v2/fluid/profiler.html
@@ -220,6 +220,35 @@
 <h1>Profiler<a class="headerlink" href="#profiler" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="id1">
 <h2>Profiler<a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.profiler.</code><code class="descname">cuda_profiler</code><span class="sig-paren">(</span><em>*args</em>, <em>**kwds</em><span class="sig-paren">)</span></dt>
+<dd><p>The CUDA profiler.
+This fuctions is used to profile CUDA program by CUDA runtime application
+programming interface. The profiling result will be written into
+<cite>output_file</cite> with Key-Value pair format or Comma separated values format.
+The user can set the output mode by <cite>output_mode</cite> argument and set the
+counters/options for profiling by <cite>config</cite> argument. The default config
+is [&#8216;gpustarttimestamp&#8217;, &#8216;gpustarttimestamp&#8217;, &#8216;gridsize3d&#8217;,
+&#8216;threadblocksize&#8217;, &#8216;streamid&#8217;, &#8216;enableonstart 0&#8217;, &#8216;conckerneltrace&#8217;].</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 last simple">
+<li><strong>output_file</strong> (<em>string</em>) &#8211; The output file name, the result will be
+written into this file.</li>
+<li><strong>output_mode</strong> (<em>string</em>) &#8211; The output mode has Key-Value pair format and
+Comma separated values format. It should be &#8216;kvp&#8217; or &#8216;csv&#8217;.</li>
+<li><strong>config</strong> (<em>list of string</em>) &#8211; The profiler options and counters can refer
+to &#8220;Compute Command Line Profiler User Guide&#8221;.</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/api/v2/fluid/regularizer.html
+++ b/develop/doc/api/v2/fluid/regularizer.html
@@ -220,12 +220,36 @@
 <h1>Regularizer<a class="headerlink" href="#regularizer" title="Permalink to this headline">¶</a></h1>
 <div class="section" id="weightdecayregularizer">
 <h2>WeightDecayRegularizer<a class="headerlink" href="#weightdecayregularizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.regularizer.</code><code class="descname">WeightDecayRegularizer</code></dt>
+<dd><p>Base class for weight decay regularizers</p>
+<p>Defines the common interface of weight-decay regularizers.
+Weight-decay regularizers are added only during the backward
+pass for faster regularization. They add operations to the network
+that correspond to gradient of the regularization function.
+Users should not use this class directly, but need to use one
+of its implementations</p>
+</dd></dl>
+
 </div>
 <div class="section" id="l2decayregularizer">
 <h2>L2DecayRegularizer<a class="headerlink" href="#l2decayregularizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.regularizer.</code><code class="descname">L2DecayRegularizer</code><span class="sig-paren">(</span><em>regularization_coeff=0.0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the L2 Weight Decay Regularization</p>
+</dd></dl>
+
 </div>
-<div class="section" id="l1decayregularizer">
-<h2>L1DecayRegularizer<a class="headerlink" href="#l1decayregularizer" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="module-paddle.v2.fluid.regularizer">
+<span id="l1decayregularizer"></span><h2>L1DecayRegularizer<a class="headerlink" href="#module-paddle.v2.fluid.regularizer" title="Permalink to this headline">¶</a></h2>
+<dl class="class">
+<dt id="paddle.v2.fluid.regularizer.L1DecayRegularizer">
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.regularizer.</code><code class="descname">L1DecayRegularizer</code><span class="sig-paren">(</span><em>regularization_coeff=0.0</em><span class="sig-paren">)</span><a class="headerlink" href="#paddle.v2.fluid.regularizer.L1DecayRegularizer" title="Permalink to this definition">¶</a></dt>
+<dd><p>Implements the L1 Weight Decay Regularization</p>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc/genindex.html
+++ b/develop/doc/genindex.html
@@ -217,6 +217,7 @@
 <a href="#B"><strong>B</strong></a>
 | <a href="#C"><strong>C</strong></a>
 | <a href="#L"><strong>L</strong></a>
+ | <a href="#M"><strong>M</strong></a>
 | <a href="#P"><strong>P</strong></a>
 | <a href="#R"><strong>R</strong></a>
 | <a href="#S"><strong>S</strong></a>
@@ -242,12 +243,14 @@
 <h2 id="L">L</h2>
 <table style="width: 100%" class="indextable genindextable"><tr>
  <td style="width: 33%; vertical-align: top;"><ul>
-      <li><a href="api/v2/data/image.html#paddle.v2.image.left_right_flip">left_right_flip() (in module paddle.v2.image)</a>
+      <li><a href="api/v2/fluid/regularizer.html#paddle.v2.fluid.regularizer.L1DecayRegularizer">L1DecayRegularizer (class in paddle.v2.fluid.regularizer)</a>
 </li>
-      <li><a href="api/v2/data/image.html#paddle.v2.image.load_and_transform">load_and_transform() (in module paddle.v2.image)</a>
+      <li><a href="api/v2/data/image.html#paddle.v2.image.left_right_flip">left_right_flip() (in module paddle.v2.image)</a>
 </li>
  </ul></td>
  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/data/image.html#paddle.v2.image.load_and_transform">load_and_transform() (in module paddle.v2.image)</a>
+</li>
      <li><a href="api/v2/data/image.html#paddle.v2.image.load_image">load_image() (in module paddle.v2.image)</a>
 </li>
      <li><a href="api/v2/data/image.html#paddle.v2.image.load_image_bytes">load_image_bytes() (in module paddle.v2.image)</a>
@@ -255,13 +258,23 @@
  </ul></td>
 </tr></table>

+<h2 id="M">M</h2>
+<table style="width: 100%" class="indextable genindextable"><tr>
+  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/fluid/evaluator.html#paddle.v2.fluid.evaluator.Evaluator.metrics">metrics (paddle.v2.fluid.evaluator.Evaluator attribute)</a>
+</li>
+  </ul></td>
+</tr></table>
+
 <h2 id="P">P</h2>
 <table style="width: 100%" class="indextable genindextable"><tr>
  <td style="width: 33%; vertical-align: top;"><ul>
-      <li><a href="api/v2/data/image.html#module-paddle.v2.image">paddle.v2.image (module)</a>
+      <li><a href="api/v2/fluid/regularizer.html#module-paddle.v2.fluid.regularizer">paddle.v2.fluid.regularizer (module)</a>
 </li>
  </ul></td>
  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/data/image.html#module-paddle.v2.image">paddle.v2.image (module)</a>
+</li>
      <li><a href="api/v1/data_provider/pydataprovider2_en.html#paddle.trainer.PyDataProvider2.provider">provider() (in module paddle.trainer.PyDataProvider2)</a>
 </li>
  </ul></td>
@@ -283,6 +296,10 @@
 <table style="width: 100%" class="indextable genindextable"><tr>
  <td style="width: 33%; vertical-align: top;"><ul>
      <li><a href="api/v2/data/image.html#paddle.v2.image.simple_transform">simple_transform() (in module paddle.v2.image)</a>
+</li>
+  </ul></td>
+  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/fluid/evaluator.html#paddle.v2.fluid.evaluator.Evaluator.states">states (paddle.v2.fluid.evaluator.Evaluator attribute)</a>
 </li>
  </ul></td>
 </tr></table>

--- a/develop/doc/objects.inv
+++ b/develop/doc/objects.inv
--- a/develop/doc/py-modindex.html
+++ b/develop/doc/py-modindex.html
@@ -229,6 +229,11 @@
       <td>
       <code class="xref">paddle</code></td><td>
       <em></em></td></tr>
+     <tr class="cg-1">
+       <td></td>
+       <td>&#160;&#160;&#160;
+       <a href="api/v2/fluid/regularizer.html#module-paddle.v2.fluid.regularizer"><code class="xref">paddle.v2.fluid.regularizer</code></a></td><td>
+       <em></em></td></tr>
     <tr class="cg-1">
       <td></td>
       <td>&#160;&#160;&#160;

--- a/develop/doc/searchindex.js
+++ b/develop/doc/searchindex.js
--- a/develop/doc_cn/api/v2/fluid/evaluator.html
+++ b/develop/doc_cn/api/v2/fluid/evaluator.html
@@ -239,6 +239,75 @@
 <h1>Evaluator<a class="headerlink" href="#evaluator" title="永久链接至标题">¶</a></h1>
 <div class="section" id="id1">
 <h2>Evaluator<a class="headerlink" href="#id1" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.evaluator.</code><code class="descname">Evaluator</code><span class="sig-paren">(</span><em>name</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Base Class for all evaluators</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">参数:</th><td class="field-body"><ul class="first last simple">
+<li><strong>name</strong> (<em>str</em>) &#8211; The name of evaluator. such as, &#8220;accuracy&#8221;. Used for generate
+temporary variable name.</li>
+<li><strong>main_program</strong> (<em>Program</em><em>, </em><em>optional</em>) &#8211; The evaluator should be added to this
+main_program. Default default_main_program()</li>
+<li><strong>startup_program</strong> (<em>Program</em><em>, </em><em>optional</em>) &#8211; The parameter should be added to this
+startup_program. Default default_startup_program()</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+<dl class="attribute">
+<dt id="paddle.v2.fluid.evaluator.Evaluator.states">
+<code class="descname">states</code><a class="headerlink" href="#paddle.v2.fluid.evaluator.Evaluator.states" title="永久链接至目标">¶</a></dt>
+<dd><p><em>list</em> &#8211; The list of state variables. states will be reset to zero
+when <cite>reset</cite> is invoked.</p>
+</dd></dl>
+
+<dl class="attribute">
+<dt id="paddle.v2.fluid.evaluator.Evaluator.metrics">
+<code class="descname">metrics</code><a class="headerlink" href="#paddle.v2.fluid.evaluator.Evaluator.metrics" title="永久链接至目标">¶</a></dt>
+<dd><p><em>list</em> &#8211; The list of metrics variables. They will be calculate
+every mini-batch</p>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">reset</code><span class="sig-paren">(</span><em>executor</em>, <em>reset_program=None</em><span class="sig-paren">)</span></dt>
+<dd><p>reset metric states at the begin of each pass/user specified batch</p>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">eval</code><span class="sig-paren">(</span><em>executor</em>, <em>eval_program=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Evaluate the statistics merged by multiple mini-batches.</p>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">create_state</code><span class="sig-paren">(</span><em>suffix</em>, <em>dtype</em>, <em>shape</em><span class="sig-paren">)</span></dt>
+<dd><p>Create state variable.</p>
+<p>NOTE: It is not a public API.</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">参数:</th><td class="field-body"><ul class="first last simple">
+<li><strong>suffix</strong> (<em>str</em>) &#8211; the state suffix.</li>
+<li><strong>dtype</strong> (<em>str|core.DataType</em>) &#8211; the state data type</li>
+<li><strong>shape</strong> (<em>tuple|list</em>) &#8211; the shape of state</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+<p>Returns: State variable</p>
+</dd></dl>
+
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/api/v2/fluid/initializer.html
+++ b/develop/doc_cn/api/v2/fluid/initializer.html
@@ -239,21 +239,90 @@
 <h1>Initializer<a class="headerlink" href="#initializer" title="永久链接至标题">¶</a></h1>
 <div class="section" id="id1">
 <h2>Initializer<a class="headerlink" href="#id1" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">Initializer</code></dt>
+<dd><p>Base class for variable initializers</p>
+<p>Defines the common interface of variable initializers.
+They add operations to the init program that are used
+to initialize variables. Users should not use this class
+directly, but need to use one of its implementations.</p>
+</dd></dl>
+
 </div>
 <div class="section" id="constantinitializer">
 <h2>ConstantInitializer<a class="headerlink" href="#constantinitializer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">ConstantInitializer</code><span class="sig-paren">(</span><em>value=0.0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the constant initializer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="uniforminitializer">
 <h2>UniformInitializer<a class="headerlink" href="#uniforminitializer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">UniformInitializer</code><span class="sig-paren">(</span><em>low=-1.0</em>, <em>high=1.0</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the random uniform distribution initializer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="normalinitializer">
 <h2>NormalInitializer<a class="headerlink" href="#normalinitializer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">NormalInitializer</code><span class="sig-paren">(</span><em>loc=0.0</em>, <em>scale=1.0</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the  random Normal(Gaussian) distribution initializer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="xavierinitializer">
 <h2>XavierInitializer<a class="headerlink" href="#xavierinitializer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">XavierInitializer</code><span class="sig-paren">(</span><em>uniform=True</em>, <em>fan_in=None</em>, <em>fan_out=None</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the Xavier initializer</p>
+<p>This class implements the Xavier weight initializer from the paper
+Understanding the difficulty of training deep feedforward neural
+networks[1] by Xavier Glorot and Yoshua Bengio.</p>
+<p>This initializer is designed to keep the scale of the gradients
+approximately same in all the layers. In case of Uniform distribution,
+the range is [-x, x], where x = sqrt(6 / (fan_in + fan_out)).
+In case of Normal distribution, the mean is 0 and the standard deviation
+is sqrt(2/ (fan_in + fan_out)).</p>
+<p class="rubric">References</p>
+<dl class="docutils">
+<dt>[1] Understanding the difficulty of training deep feedforward neural</dt>
+<dd>networks. International conference on artificial intelligence and
+statistics.
+(<a class="reference external" href="http://proceedings.mlr.press/v9/glorot10a.html">http://proceedings.mlr.press/v9/glorot10a.html</a>)</dd>
+</dl>
+</dd></dl>
+
 </div>
 <div class="section" id="msrainitializer">
 <h2>MSRAInitializer<a class="headerlink" href="#msrainitializer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.initializer.</code><code class="descname">MSRAInitializer</code><span class="sig-paren">(</span><em>uniform=True</em>, <em>fan_in=None</em>, <em>seed=0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the MSRA initializer a.k.a. Kaiming Initializer</p>
+<p>This class implements the weight initialization from the paper
+Delving Deep into Rectifiers: Surpassing Human-Level Performance on
+ImageNet Classification[1] by Kaiming He, Xiangyu Zhang, Shaoqing Ren
+and Jian Sun. This is a robust initialization method that particularly
+considers the rectifier nonlinearities. In case of Uniform distribution,
+the range is [-x, x], where x = sqrt(6 / fan_in). In case of Normal
+distribution, the mean is 0 and the standard deviation
+is sqrt(2/ fan_in).</p>
+<p class="rubric">References</p>
+<dl class="docutils">
+<dt>[1] Delving Deep into Rectifiers: Surpassing Human-Level Performance</dt>
+<dd>on ImageNet Classification
+(<a class="reference external" href="https://arxiv.org/abs/1502.01852">https://arxiv.org/abs/1502.01852</a>)</dd>
+</dl>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/api/v2/fluid/io.html
+++ b/develop/doc_cn/api/v2/fluid/io.html
@@ -239,6 +239,23 @@
 <h1>IO<a class="headerlink" href="#io" title="永久链接至标题">¶</a></h1>
 <div class="section" id="is-parameter">
 <h2>is_parameter<a class="headerlink" href="#is-parameter" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.io.</code><code class="descname">is_parameter</code><span class="sig-paren">(</span><em>var</em><span class="sig-paren">)</span></dt>
+<dd><p>Check whether the variable is a Parameter.</p>
+<p>This function checks whether the input variable is a Parameter.</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">参数:</th><td class="field-body"><strong>var</strong> &#8211; The input variable.</td>
+</tr>
+<tr class="field-even field"><th class="field-name">返回:</th><td class="field-body">boolean result whether the variable is a Parameter.</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/api/v2/fluid/layers.html
+++ b/develop/doc_cn/api/v2/fluid/layers.html
--- a/develop/doc_cn/api/v2/fluid/nets.html
+++ b/develop/doc_cn/api/v2/fluid/nets.html
@@ -239,18 +239,118 @@
 <h1>Nets<a class="headerlink" href="#nets" title="永久链接至标题">¶</a></h1>
 <div class="section" id="simple-img-conv-pool">
 <h2>simple_img_conv_pool<a class="headerlink" href="#simple-img-conv-pool" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">simple_img_conv_pool</code><span class="sig-paren">(</span><em>input</em>, <em>num_filters</em>, <em>filter_size</em>, <em>pool_size</em>, <em>pool_stride</em>, <em>act</em>, <em>param_attr=None</em>, <em>pool_type='max'</em>, <em>use_cudnn=True</em><span class="sig-paren">)</span></dt>
+<dd></dd></dl>
+
 </div>
 <div class="section" id="img-conv-group">
 <h2>img_conv_group<a class="headerlink" href="#img-conv-group" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">img_conv_group</code><span class="sig-paren">(</span><em>input</em>, <em>conv_num_filter</em>, <em>pool_size</em>, <em>conv_padding=1</em>, <em>conv_filter_size=3</em>, <em>conv_act=None</em>, <em>param_attr=None</em>, <em>conv_with_batchnorm=False</em>, <em>conv_batchnorm_drop_rate=None</em>, <em>pool_stride=1</em>, <em>pool_type=None</em>, <em>use_cudnn=True</em><span class="sig-paren">)</span></dt>
+<dd><p>Image Convolution Group, Used for vgg net.</p>
+</dd></dl>
+
 </div>
 <div class="section" id="sequence-conv-pool">
 <h2>sequence_conv_pool<a class="headerlink" href="#sequence-conv-pool" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">sequence_conv_pool</code><span class="sig-paren">(</span><em>input</em>, <em>num_filters</em>, <em>filter_size</em>, <em>param_attr=None</em>, <em>act='sigmoid'</em>, <em>pool_type='max'</em><span class="sig-paren">)</span></dt>
+<dd></dd></dl>
+
 </div>
 <div class="section" id="glu">
 <h2>glu<a class="headerlink" href="#glu" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">glu</code><span class="sig-paren">(</span><em>input</em>, <em>dim=-1</em><span class="sig-paren">)</span></dt>
+<dd><p>The gated linear unit composed by split, sigmoid activation and elementwise
+multiplication. Specifically, Split the input into two equal sized parts
+<span class="math">\(a\)</span> and <span class="math">\(b\)</span> along the given dimension and then compute as
+following:</p>
+<blockquote>
+<div><div class="math">
+\[{GLU}(a, b)= a \otimes \sigma(b)\]</div>
+</div></blockquote>
+<p>Refer to <a class="reference external" href="https://arxiv.org/pdf/1612.08083.pdf">Language Modeling with Gated Convolutional Networks</a>.</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">参数:</th><td class="field-body"><ul class="first simple">
+<li><strong>input</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+<li><strong>dim</strong> (<em>int</em>) &#8211; The dimension along which to split. If <span class="math">\(dim &lt; 0\)</span>, the
+dimension to split along is <span class="math">\(rank(input) + dim\)</span>.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">返回:</th><td class="field-body"><p class="first">The Tensor variable with half the size of input.</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">返回类型:</th><td class="field-body"><p class="first last">Variable</p>
+</td>
+</tr>
+</tbody>
+</table>
+<p class="rubric">Examples</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># x is a Tensor variable with shape [3, 6, 9]</span>
+<span class="n">fluid</span><span class="o">.</span><span class="n">nets</span><span class="o">.</span><span class="n">glu</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">dim</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>  <span class="c1"># shape of output: [3, 3, 9]</span>
+</pre></div>
+</div>
+</dd></dl>
+
 </div>
 <div class="section" id="dot-product-attention">
 <h2>dot_product_attention<a class="headerlink" href="#dot-product-attention" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.nets.</code><code class="descname">dot_product_attention</code><span class="sig-paren">(</span><em>querys</em>, <em>keys</em>, <em>values</em><span class="sig-paren">)</span></dt>
+<dd><p>The dot-product attention.</p>
+<p>Attention mechanism can be seen as mapping a query and a set of key-value
+pairs to an output. The output is computed as a weighted sum of the values,
+where the weight assigned to each value is computed by a compatibility
+function (dot-product here) of the query with the corresponding key.</p>
+<p>The dot-product attention can be implemented through (batch) matrix
+multipication as follows:</p>
+<blockquote>
+<div><div class="math">
+\[Attention(Q, K, V)= softmax(QK^\mathrm{T})V\]</div>
+</div></blockquote>
+<p>Refer to <a class="reference external" href="https://arxiv.org/pdf/1706.03762.pdf">Attention Is All You Need</a>.</p>
+<p>Note that batch data containing sequences with different lengths is not
+supported by this because of the (batch) matrix multipication.</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">参数:</th><td class="field-body"><ul class="first simple">
+<li><strong>query</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+<li><strong>key</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+<li><strong>value</strong> (<em>Variable</em>) &#8211; The input variable which is a Tensor or LoDTensor.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">返回:</th><td class="field-body"><p class="first">The Tensor variables representing the output and attention scores.</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">返回类型:</th><td class="field-body"><p class="first last">tuple</p>
+</td>
+</tr>
+</tbody>
+</table>
+<p class="rubric">Examples</p>
+<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># Suppose q, k, v are tensor variables with the following shape:</span>
+<span class="c1"># q: [3, 5, 9], k: [3, 6, 9], v: [3, 6, 10]</span>
+<span class="n">out</span><span class="p">,</span> <span class="n">attn_scores</span> <span class="o">=</span> <span class="n">fluid</span><span class="o">.</span><span class="n">nets</span><span class="o">.</span><span class="n">dot_product_attention</span><span class="p">(</span><span class="n">q</span><span class="p">,</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span>
+<span class="n">out</span><span class="o">.</span><span class="n">shape</span>  <span class="c1"># [3, 5, 10]</span>
+<span class="n">attn_scores</span><span class="o">.</span><span class="n">shape</span>  <span class="c1"># [3, 5, 6]</span>
+</pre></div>
+</div>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/api/v2/fluid/optimizer.html
+++ b/develop/doc_cn/api/v2/fluid/optimizer.html
@@ -239,24 +239,105 @@
 <h1>Optimizer<a class="headerlink" href="#optimizer" title="永久链接至标题">¶</a></h1>
 <div class="section" id="id1">
 <h2>Optimizer<a class="headerlink" href="#id1" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">Optimizer</code><span class="sig-paren">(</span><em>global_step=None</em>, <em>regularization=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Optimizer Base class.</p>
+<p>Define the common interface of an optimizer.
+User should not use this class directly,
+but need to use one of it&#8217;s implementation.</p>
+<dl class="method">
+<dt>
+<code class="descname">create_optimization_pass</code><span class="sig-paren">(</span><em>parameters_and_grads</em>, <em>loss</em>, <em>startup_program=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Add optimization operators to update gradients to variables.</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">参数:</th><td class="field-body"><ul class="first simple">
+<li><strong>loss</strong> &#8211; the target that this optimization is for.</li>
+<li><strong>parameters_and_grads</strong> &#8211; a list of (variable, gradient) pair to update.</li>
+</ul>
+</td>
+</tr>
+<tr class="field-even field"><th class="field-name">返回:</th><td class="field-body"><p class="first">a list of operators that will complete one step of
+optimization. This will include parameter update ops, global step
+update ops and any other custom ops required by subclasses to manage
+their internal state.
+:param startup_program:</p>
+</td>
+</tr>
+<tr class="field-odd field"><th class="field-name">返回类型:</th><td class="field-body"><p class="first last">return_op_list</p>
+</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+
+<dl class="method">
+<dt>
+<code class="descname">minimize</code><span class="sig-paren">(</span><em>loss</em>, <em>startup_program=None</em>, <em>parameter_list=None</em>, <em>no_grad_set=None</em><span class="sig-paren">)</span></dt>
+<dd><p>Add operations to minimize <cite>loss</cite> by updating <cite>parameter_list</cite>.</p>
+<p>This method combines interface <cite>append_backward()</cite> and
+<cite>create_optimization_pass()</cite> into one.</p>
+</dd></dl>
+
+</dd></dl>
+
 </div>
 <div class="section" id="sgdoptimizer">
 <h2>SGDOptimizer<a class="headerlink" href="#sgdoptimizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">SGDOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple SGD optimizer without any state.</p>
+</dd></dl>
+
 </div>
 <div class="section" id="momentumoptimizer">
 <h2>MomentumOptimizer<a class="headerlink" href="#momentumoptimizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">MomentumOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>momentum</em>, <em>use_nesterov=False</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple Momentum optimizer with velocity state</p>
+</dd></dl>
+
 </div>
 <div class="section" id="adagradoptimizer">
 <h2>AdagradOptimizer<a class="headerlink" href="#adagradoptimizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">AdagradOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>epsilon=1e-06</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple Adagrad optimizer with moment state</p>
+</dd></dl>
+
 </div>
 <div class="section" id="adamoptimizer">
 <h2>AdamOptimizer<a class="headerlink" href="#adamoptimizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">AdamOptimizer</code><span class="sig-paren">(</span><em>learning_rate=0.001</em>, <em>beta1=0.9</em>, <em>beta2=0.999</em>, <em>epsilon=1e-08</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the Adam Optimizer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="adamaxoptimizer">
 <h2>AdamaxOptimizer<a class="headerlink" href="#adamaxoptimizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">AdamaxOptimizer</code><span class="sig-paren">(</span><em>learning_rate=0.001</em>, <em>beta1=0.9</em>, <em>beta2=0.999</em>, <em>epsilon=1e-08</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the Adamax Optimizer</p>
+</dd></dl>
+
 </div>
 <div class="section" id="decayedadagradoptimizer">
 <h2>DecayedAdagradOptimizer<a class="headerlink" href="#decayedadagradoptimizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.optimizer.</code><code class="descname">DecayedAdagradOptimizer</code><span class="sig-paren">(</span><em>learning_rate</em>, <em>decay=0.95</em>, <em>epsilon=1e-06</em>, <em>**kwargs</em><span class="sig-paren">)</span></dt>
+<dd><p>Simple Decayed Adagrad optimizer with moment state</p>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/api/v2/fluid/profiler.html
+++ b/develop/doc_cn/api/v2/fluid/profiler.html
@@ -239,6 +239,35 @@
 <h1>Profiler<a class="headerlink" href="#profiler" title="永久链接至标题">¶</a></h1>
 <div class="section" id="id1">
 <h2>Profiler<a class="headerlink" href="#id1" title="永久链接至标题">¶</a></h2>
+<dl class="function">
+<dt>
+<code class="descclassname">paddle.v2.fluid.profiler.</code><code class="descname">cuda_profiler</code><span class="sig-paren">(</span><em>*args</em>, <em>**kwds</em><span class="sig-paren">)</span></dt>
+<dd><p>The CUDA profiler.
+This fuctions is used to profile CUDA program by CUDA runtime application
+programming interface. The profiling result will be written into
+<cite>output_file</cite> with Key-Value pair format or Comma separated values format.
+The user can set the output mode by <cite>output_mode</cite> argument and set the
+counters/options for profiling by <cite>config</cite> argument. The default config
+is [&#8216;gpustarttimestamp&#8217;, &#8216;gpustarttimestamp&#8217;, &#8216;gridsize3d&#8217;,
+&#8216;threadblocksize&#8217;, &#8216;streamid&#8217;, &#8216;enableonstart 0&#8217;, &#8216;conckerneltrace&#8217;].</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">参数:</th><td class="field-body"><ul class="first last simple">
+<li><strong>output_file</strong> (<em>string</em>) &#8211; The output file name, the result will be
+written into this file.</li>
+<li><strong>output_mode</strong> (<em>string</em>) &#8211; The output mode has Key-Value pair format and
+Comma separated values format. It should be &#8216;kvp&#8217; or &#8216;csv&#8217;.</li>
+<li><strong>config</strong> (<em>list of string</em>) &#8211; The profiler options and counters can refer
+to &#8220;Compute Command Line Profiler User Guide&#8221;.</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/api/v2/fluid/regularizer.html
+++ b/develop/doc_cn/api/v2/fluid/regularizer.html
@@ -239,12 +239,36 @@
 <h1>Regularizer<a class="headerlink" href="#regularizer" title="永久链接至标题">¶</a></h1>
 <div class="section" id="weightdecayregularizer">
 <h2>WeightDecayRegularizer<a class="headerlink" href="#weightdecayregularizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.regularizer.</code><code class="descname">WeightDecayRegularizer</code></dt>
+<dd><p>Base class for weight decay regularizers</p>
+<p>Defines the common interface of weight-decay regularizers.
+Weight-decay regularizers are added only during the backward
+pass for faster regularization. They add operations to the network
+that correspond to gradient of the regularization function.
+Users should not use this class directly, but need to use one
+of its implementations</p>
+</dd></dl>
+
 </div>
 <div class="section" id="l2decayregularizer">
 <h2>L2DecayRegularizer<a class="headerlink" href="#l2decayregularizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt>
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.regularizer.</code><code class="descname">L2DecayRegularizer</code><span class="sig-paren">(</span><em>regularization_coeff=0.0</em><span class="sig-paren">)</span></dt>
+<dd><p>Implements the L2 Weight Decay Regularization</p>
+</dd></dl>
+
 </div>
-<div class="section" id="l1decayregularizer">
-<h2>L1DecayRegularizer<a class="headerlink" href="#l1decayregularizer" title="永久链接至标题">¶</a></h2>
+<div class="section" id="module-paddle.v2.fluid.regularizer">
+<span id="l1decayregularizer"></span><h2>L1DecayRegularizer<a class="headerlink" href="#module-paddle.v2.fluid.regularizer" title="永久链接至标题">¶</a></h2>
+<dl class="class">
+<dt id="paddle.v2.fluid.regularizer.L1DecayRegularizer">
+<em class="property">class </em><code class="descclassname">paddle.v2.fluid.regularizer.</code><code class="descname">L1DecayRegularizer</code><span class="sig-paren">(</span><em>regularization_coeff=0.0</em><span class="sig-paren">)</span><a class="headerlink" href="#paddle.v2.fluid.regularizer.L1DecayRegularizer" title="永久链接至目标">¶</a></dt>
+<dd><p>Implements the L1 Weight Decay Regularization</p>
+</dd></dl>
+
 </div>
 </div>


--- a/develop/doc_cn/genindex.html
+++ b/develop/doc_cn/genindex.html
@@ -236,6 +236,7 @@
 <a href="#B"><strong>B</strong></a>
 | <a href="#C"><strong>C</strong></a>
 | <a href="#L"><strong>L</strong></a>
+ | <a href="#M"><strong>M</strong></a>
 | <a href="#P"><strong>P</strong></a>
 | <a href="#R"><strong>R</strong></a>
 | <a href="#S"><strong>S</strong></a>
@@ -261,12 +262,14 @@
 <h2 id="L">L</h2>
 <table style="width: 100%" class="indextable genindextable"><tr>
  <td style="width: 33%; vertical-align: top;"><ul>
-      <li><a href="api/v2/data/image.html#paddle.v2.image.left_right_flip">left_right_flip() (在 paddle.v2.image 模块中)</a>
+      <li><a href="api/v2/fluid/regularizer.html#paddle.v2.fluid.regularizer.L1DecayRegularizer">L1DecayRegularizer (paddle.v2.fluid.regularizer 中的类)</a>
 </li>
-      <li><a href="api/v2/data/image.html#paddle.v2.image.load_and_transform">load_and_transform() (在 paddle.v2.image 模块中)</a>
+      <li><a href="api/v2/data/image.html#paddle.v2.image.left_right_flip">left_right_flip() (在 paddle.v2.image 模块中)</a>
 </li>
  </ul></td>
  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/data/image.html#paddle.v2.image.load_and_transform">load_and_transform() (在 paddle.v2.image 模块中)</a>
+</li>
      <li><a href="api/v2/data/image.html#paddle.v2.image.load_image">load_image() (在 paddle.v2.image 模块中)</a>
 </li>
      <li><a href="api/v2/data/image.html#paddle.v2.image.load_image_bytes">load_image_bytes() (在 paddle.v2.image 模块中)</a>
@@ -274,8 +277,20 @@
  </ul></td>
 </tr></table>

+<h2 id="M">M</h2>
+<table style="width: 100%" class="indextable genindextable"><tr>
+  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/fluid/evaluator.html#paddle.v2.fluid.evaluator.Evaluator.metrics">metrics (paddle.v2.fluid.evaluator.Evaluator 属性)</a>
+</li>
+  </ul></td>
+</tr></table>
+
 <h2 id="P">P</h2>
 <table style="width: 100%" class="indextable genindextable"><tr>
+  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/fluid/regularizer.html#module-paddle.v2.fluid.regularizer">paddle.v2.fluid.regularizer (模块)</a>
+</li>
+  </ul></td>
  <td style="width: 33%; vertical-align: top;"><ul>
      <li><a href="api/v2/data/image.html#module-paddle.v2.image">paddle.v2.image (模块)</a>
 </li>
@@ -298,6 +313,10 @@
 <table style="width: 100%" class="indextable genindextable"><tr>
  <td style="width: 33%; vertical-align: top;"><ul>
      <li><a href="api/v2/data/image.html#paddle.v2.image.simple_transform">simple_transform() (在 paddle.v2.image 模块中)</a>
+</li>
+  </ul></td>
+  <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="api/v2/fluid/evaluator.html#paddle.v2.fluid.evaluator.Evaluator.states">states (paddle.v2.fluid.evaluator.Evaluator 属性)</a>
 </li>
  </ul></td>
 </tr></table>

--- a/develop/doc_cn/objects.inv
+++ b/develop/doc_cn/objects.inv
--- a/develop/doc_cn/py-modindex.html
+++ b/develop/doc_cn/py-modindex.html
@@ -248,6 +248,11 @@
       <td>
       <code class="xref">paddle</code></td><td>
       <em></em></td></tr>
+     <tr class="cg-1">
+       <td></td>
+       <td>&#160;&#160;&#160;
+       <a href="api/v2/fluid/regularizer.html#module-paddle.v2.fluid.regularizer"><code class="xref">paddle.v2.fluid.regularizer</code></a></td><td>
+       <em></em></td></tr>
     <tr class="cg-1">
       <td></td>
       <td>&#160;&#160;&#160;

--- a/develop/doc_cn/searchindex.js
+++ b/develop/doc_cn/searchindex.js