Skip to content

P
Paddle

BaiXuePrincess / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
提交 f9440e66 编写于 作者: T Travis CI
浏览文件
操作

Deploy to GitHub Pages: b884bc33

上级 4f591e47
展开全部 隐藏空白更改
内联 并排
Showing with 96 addition and 14 deletion
develop/doc/_sources/design/register_grad_op.md.txt
浏览文件 @ f9440e66
......@@ -33,22 +33,45 @@ The mapping relationship between an operator and its gradient operators is a fun
```cpp
// (OpDesc) --> vector<OpDesc>
using GradOpDescMaker = std::function<std::vector<OpDesc>(const OpDesc&)>;
std::function<std::vector<OpDescBind>(const OpDescBind&)>;
```
The function take a `OpDesc` of the forward operator and return one or many gradient operator descriptions.
The function takes an `OpDescBind` of the forward operator and returns one or many gradient operator descriptions. `OpDescBind` is a C++ wrapper for protobuf message `OpDesc` to manipulate `OpDesc` fast.
The `GradOpDescMaker` will be registered in `OpInfo`, to replace `grad_op_type_` field. The `OpInfo` should be
```cpp
struct OpInfo {
GradOpDescMaker grad_op_maker_;
std::function<std::vector<std::unique_ptr<OpDescBind>>(const OpDescBind&)> grad_op_maker_;
...
};
```
The `grad_op_maker_ ` is `nullptr` if the operator does not have associated gradient operators.
We propose a base class called `GradOpDescMakerBase` to let operator developers generate `Gradient Operators` easily. The public interface of that class is
```cpp
class GradOpDescMakerBase {
public:
GradOpDescMakerBase(const OpDescBind& );
virtual std::vector<std::unique_ptr<OpDescBind>> operator()()const = 0;
};
```
We can convert `GradOpDescMakerBase` to `std::function<std::vector<std::unique_ptr<OpDescBind>>(const OpDescBind&)>` by
```cpp
using GradOpMaker = ...;
std::function<std::vector<OpDescBind>(const OpDescBind&)> func;
func = [] (const OpDescBind& fwd_op) {
GradOpMaker maker(fwd_op);
return maker();
};
```
We can write many helper functions since the `GradOpDescMakerBase` is a class now. The basic helper functions get the variables of `Input`, `Output`, `InputGradient` and `OutputGradient` in the forwarding operator.
We should chagne register macros at the same time. In the current solution, there is no difference between forwarding operators and backward operators. So `REGISTER_OP` just register one operator. If the `REGISTER_OPERATOR ` contains `OpProtoAndCheckerMaker` and `GradOpDescMaker`, we just list them in the same macro. It can be done by a macro contains `__VA_ARGS__`.
The user interface should be
......
develop/doc/design/register_grad_op.html
浏览文件 @ f9440e66
......@@ -206,18 +206,36 @@
<span id="proposed-solution"></span><h2>Proposed Solution<a class="headerlink" href="#proposed-solution" title="Permalink to this headline"></a></h2>
<p>The mapping relationship between an operator and its gradient operators is a function. The interface of that function is:</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="c1">// (OpDesc) --&gt; vector&lt;OpDesc&gt;</span>
<span class="k">using</span> <span class="n">GradOpDescMaker</span> <span class="o">=</span> <span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDesc</span><span class="o">&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDesc</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span><span class="p">;</span>
<span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span><span class="p">;</span>
</pre></div>
</div>
<p>The function take a <code class="docutils literal"><span class="pre">OpDesc</span></code> of the forward operator and return one or many gradient operator descriptions.</p>
<p>The function takes an <code class="docutils literal"><span class="pre">OpDescBind</span></code> of the forward operator and returns one or many gradient operator descriptions. <code class="docutils literal"><span class="pre">OpDescBind</span></code> is a C++ wrapper for protobuf message <code class="docutils literal"><span class="pre">OpDesc</span></code> to manipulate <code class="docutils literal"><span class="pre">OpDesc</span></code> fast.</p>
<p>The <code class="docutils literal"><span class="pre">GradOpDescMaker</span></code> will be registered in <code class="docutils literal"><span class="pre">OpInfo</span></code>, to replace <code class="docutils literal"><span class="pre">grad_op_type_</span></code> field. The <code class="docutils literal"><span class="pre">OpInfo</span></code> should be</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="k">struct</span> <span class="n">OpInfo</span> <span class="p">{</span>
<span class="n">GradOpDescMaker</span> <span class="n">grad_op_maker_</span><span class="p">;</span>
<span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">unique_ptr</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span> <span class="n">grad_op_maker_</span><span class="p">;</span>
<span class="p">...</span>
<span class="p">};</span>
</pre></div>
</div>
<p>The <code class="docutils literal"><span class="pre">grad_op_maker_</span></code> is <code class="docutils literal"><span class="pre">nullptr</span></code> if the operator does not have associated gradient operators.</p>
<p>We propose a base class called <code class="docutils literal"><span class="pre">GradOpDescMakerBase</span></code> to let operator developers generate <code class="docutils literal"><span class="pre">Gradient</span> <span class="pre">Operators</span></code> easily. The public interface of that class is</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">GradOpDescMakerBase</span> <span class="p">{</span>
<span class="k">public</span><span class="o">:</span>
<span class="n">GradOpDescMakerBase</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span> <span class="p">);</span>
<span class="k">virtual</span> <span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">unique_ptr</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;&gt;</span> <span class="k">operator</span><span class="p">()()</span><span class="k">const</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
<span class="p">};</span>
</pre></div>
</div>
<p>We can convert <code class="docutils literal"><span class="pre">GradOpDescMakerBase</span></code> to <code class="docutils literal"><span class="pre">std::function&lt;std::vector&lt;std::unique_ptr&lt;OpDescBind&gt;&gt;(const</span> <span class="pre">OpDescBind&amp;)&gt;</span></code> by</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="k">using</span> <span class="n">GradOpMaker</span> <span class="o">=</span> <span class="p">...;</span>
<span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span> <span class="n">func</span><span class="p">;</span>
<span class="n">func</span> <span class="o">=</span> <span class="p">[]</span> <span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span> <span class="n">fwd_op</span><span class="p">)</span> <span class="p">{</span>
<span class="n">GradOpMaker</span> <span class="n">maker</span><span class="p">(</span><span class="n">fwd_op</span><span class="p">);</span>
<span class="k">return</span> <span class="nf">maker</span><span class="p">();</span>
<span class="p">};</span>
</pre></div>
</div>
<p>We can write many helper functions since the <code class="docutils literal"><span class="pre">GradOpDescMakerBase</span></code> is a class now. The basic helper functions get the variables of <code class="docutils literal"><span class="pre">Input</span></code>, <code class="docutils literal"><span class="pre">Output</span></code>, <code class="docutils literal"><span class="pre">InputGradient</span></code> and <code class="docutils literal"><span class="pre">OutputGradient</span></code> in the forwarding operator.</p>
<p>We should chagne register macros at the same time. In the current solution, there is no difference between forwarding operators and backward operators. So <code class="docutils literal"><span class="pre">REGISTER_OP</span></code> just register one operator. If the <code class="docutils literal"><span class="pre">REGISTER_OPERATOR</span></code> contains <code class="docutils literal"><span class="pre">OpProtoAndCheckerMaker</span></code> and <code class="docutils literal"><span class="pre">GradOpDescMaker</span></code>, we just list them in the same macro. It can be done by a macro contains <code class="docutils literal"><span class="pre">__VA_ARGS__</span></code>.</p>
<p>The user interface should be</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDesc</span><span class="o">&gt;</span> <span class="n">MinusOpGradMaker</span><span class="p">(</span><span class="n">OpDesc</span><span class="p">)</span> <span class="p">{...}</span>
......
develop/doc/searchindex.js
浏览文件 @ f9440e66
因为 它太大了无法显示 source diff 。你可以改为 查看blob。
develop/doc_cn/_sources/design/register_grad_op.md.txt
浏览文件 @ f9440e66
......@@ -33,22 +33,45 @@ The mapping relationship between an operator and its gradient operators is a fun
```cpp
// (OpDesc) --> vector<OpDesc>
using GradOpDescMaker = std::function<std::vector<OpDesc>(const OpDesc&)>;
std::function<std::vector<OpDescBind>(const OpDescBind&)>;
```
The function take a `OpDesc` of the forward operator and return one or many gradient operator descriptions.
The function takes an `OpDescBind` of the forward operator and returns one or many gradient operator descriptions. `OpDescBind` is a C++ wrapper for protobuf message `OpDesc` to manipulate `OpDesc` fast.
The `GradOpDescMaker` will be registered in `OpInfo`, to replace `grad_op_type_` field. The `OpInfo` should be
```cpp
struct OpInfo {
GradOpDescMaker grad_op_maker_;
std::function<std::vector<std::unique_ptr<OpDescBind>>(const OpDescBind&)> grad_op_maker_;
...
};
```
The `grad_op_maker_ ` is `nullptr` if the operator does not have associated gradient operators.
We propose a base class called `GradOpDescMakerBase` to let operator developers generate `Gradient Operators` easily. The public interface of that class is
```cpp
class GradOpDescMakerBase {
public:
GradOpDescMakerBase(const OpDescBind& );
virtual std::vector<std::unique_ptr<OpDescBind>> operator()()const = 0;
};
```
We can convert `GradOpDescMakerBase` to `std::function<std::vector<std::unique_ptr<OpDescBind>>(const OpDescBind&)>` by
```cpp
using GradOpMaker = ...;
std::function<std::vector<OpDescBind>(const OpDescBind&)> func;
func = [] (const OpDescBind& fwd_op) {
GradOpMaker maker(fwd_op);
return maker();
};
```
We can write many helper functions since the `GradOpDescMakerBase` is a class now. The basic helper functions get the variables of `Input`, `Output`, `InputGradient` and `OutputGradient` in the forwarding operator.
We should chagne register macros at the same time. In the current solution, there is no difference between forwarding operators and backward operators. So `REGISTER_OP` just register one operator. If the `REGISTER_OPERATOR ` contains `OpProtoAndCheckerMaker` and `GradOpDescMaker`, we just list them in the same macro. It can be done by a macro contains `__VA_ARGS__`.
The user interface should be
......
develop/doc_cn/design/register_grad_op.html
浏览文件 @ f9440e66
......@@ -220,18 +220,36 @@
<span id="proposed-solution"></span><h2>Proposed Solution<a class="headerlink" href="#proposed-solution" title="永久链接至标题"></a></h2>
<p>The mapping relationship between an operator and its gradient operators is a function. The interface of that function is:</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="c1">// (OpDesc) --&gt; vector&lt;OpDesc&gt;</span>
<span class="k">using</span> <span class="n">GradOpDescMaker</span> <span class="o">=</span> <span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDesc</span><span class="o">&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDesc</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span><span class="p">;</span>
<span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span><span class="p">;</span>
</pre></div>
</div>
<p>The function take a <code class="docutils literal"><span class="pre">OpDesc</span></code> of the forward operator and return one or many gradient operator descriptions.</p>
<p>The function takes an <code class="docutils literal"><span class="pre">OpDescBind</span></code> of the forward operator and returns one or many gradient operator descriptions. <code class="docutils literal"><span class="pre">OpDescBind</span></code> is a C++ wrapper for protobuf message <code class="docutils literal"><span class="pre">OpDesc</span></code> to manipulate <code class="docutils literal"><span class="pre">OpDesc</span></code> fast.</p>
<p>The <code class="docutils literal"><span class="pre">GradOpDescMaker</span></code> will be registered in <code class="docutils literal"><span class="pre">OpInfo</span></code>, to replace <code class="docutils literal"><span class="pre">grad_op_type_</span></code> field. The <code class="docutils literal"><span class="pre">OpInfo</span></code> should be</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="k">struct</span> <span class="n">OpInfo</span> <span class="p">{</span>
<span class="n">GradOpDescMaker</span> <span class="n">grad_op_maker_</span><span class="p">;</span>
<span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">unique_ptr</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span> <span class="n">grad_op_maker_</span><span class="p">;</span>
<span class="p">...</span>
<span class="p">};</span>
</pre></div>
</div>
<p>The <code class="docutils literal"><span class="pre">grad_op_maker_</span></code> is <code class="docutils literal"><span class="pre">nullptr</span></code> if the operator does not have associated gradient operators.</p>
<p>We propose a base class called <code class="docutils literal"><span class="pre">GradOpDescMakerBase</span></code> to let operator developers generate <code class="docutils literal"><span class="pre">Gradient</span> <span class="pre">Operators</span></code> easily. The public interface of that class is</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">GradOpDescMakerBase</span> <span class="p">{</span>
<span class="k">public</span><span class="o">:</span>
<span class="n">GradOpDescMakerBase</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span> <span class="p">);</span>
<span class="k">virtual</span> <span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">unique_ptr</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;&gt;</span> <span class="k">operator</span><span class="p">()()</span><span class="k">const</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
<span class="p">};</span>
</pre></div>
</div>
<p>We can convert <code class="docutils literal"><span class="pre">GradOpDescMakerBase</span></code> to <code class="docutils literal"><span class="pre">std::function&lt;std::vector&lt;std::unique_ptr&lt;OpDescBind&gt;&gt;(const</span> <span class="pre">OpDescBind&amp;)&gt;</span></code> by</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="k">using</span> <span class="n">GradOpMaker</span> <span class="o">=</span> <span class="p">...;</span>
<span class="n">std</span><span class="o">::</span><span class="n">function</span><span class="o">&lt;</span><span class="n">std</span><span class="o">::</span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDescBind</span><span class="o">&gt;</span><span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span><span class="p">)</span><span class="o">&gt;</span> <span class="n">func</span><span class="p">;</span>
<span class="n">func</span> <span class="o">=</span> <span class="p">[]</span> <span class="p">(</span><span class="k">const</span> <span class="n">OpDescBind</span><span class="o">&amp;</span> <span class="n">fwd_op</span><span class="p">)</span> <span class="p">{</span>
<span class="n">GradOpMaker</span> <span class="n">maker</span><span class="p">(</span><span class="n">fwd_op</span><span class="p">);</span>
<span class="k">return</span> <span class="nf">maker</span><span class="p">();</span>
<span class="p">};</span>
</pre></div>
</div>
<p>We can write many helper functions since the <code class="docutils literal"><span class="pre">GradOpDescMakerBase</span></code> is a class now. The basic helper functions get the variables of <code class="docutils literal"><span class="pre">Input</span></code>, <code class="docutils literal"><span class="pre">Output</span></code>, <code class="docutils literal"><span class="pre">InputGradient</span></code> and <code class="docutils literal"><span class="pre">OutputGradient</span></code> in the forwarding operator.</p>
<p>We should chagne register macros at the same time. In the current solution, there is no difference between forwarding operators and backward operators. So <code class="docutils literal"><span class="pre">REGISTER_OP</span></code> just register one operator. If the <code class="docutils literal"><span class="pre">REGISTER_OPERATOR</span></code> contains <code class="docutils literal"><span class="pre">OpProtoAndCheckerMaker</span></code> and <code class="docutils literal"><span class="pre">GradOpDescMaker</span></code>, we just list them in the same macro. It can be done by a macro contains <code class="docutils literal"><span class="pre">__VA_ARGS__</span></code>.</p>
<p>The user interface should be</p>
<div class="highlight-cpp"><div class="highlight"><pre><span></span><span class="n">vector</span><span class="o">&lt;</span><span class="n">OpDesc</span><span class="o">&gt;</span> <span class="n">MinusOpGradMaker</span><span class="p">(</span><span class="n">OpDesc</span><span class="p">)</span> <span class="p">{...}</span>
......
develop/doc_cn/searchindex.js
浏览文件 @ f9440e66
此差异已折叠。
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或