Skip to content

P
Paddle

PaddlePaddle / Paddle
1 年多 前同步成功

通知 2302
Star 20931
Fork 5422
P
Paddle
未验证 提交 de8c4627 编写于 作者: Q QI JUN 提交者： GitHub
浏览文件
操作

Update new op docs (#6505) 

* update docs about how to add a new operator
上级 61ec0b95
隐藏空白更改
内联 并排
Showing with 80 addition and 111 deletion
doc/howto/dev/new_op_cn.md
浏览文件 @ de8c4627
...@@ -30,8 +30,8 @@ ...@@ -30,8 +30,8 @@
-------------- | :---------------------- -------------- | :----------------------
OpProtoMake定义 | `.cc`文件，Backward Op不需要定义OpProtoMake OpProtoMake定义 | `.cc`文件，Backward Op不需要定义OpProtoMake
Op定义 | `.cc`文件 Op定义 | `.cc`文件
Kernel实现 | CPU、GPU共享Kernel实现在`.h`文件中，否则，CPU 实现在`.cc`文件中，GPU 实现在`.cu`文件中。 Kernel实现 | CPU、CUDA共享Kernel实现在`.h`文件中，否则，CPU 实现在`.cc`文件中，CUDA 实现在`.cu`文件中。
注册Op | Op注册实现在`.cc`文件；Kernel注册CPU实现在`.cc`文件中，GPU实现在`.cu`文件中 注册Op | Op注册实现在`.cc`文件；Kernel注册CPU实现在`.cc`文件中，CUDA实现在`.cu`文件中
实现新的op都添加至目录[paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)下，文件命名以`*_op.h`（如有） 、 `*_op.cc` 、`*_op.cu`（如有）结尾。**系统会根据文件名自动构建op和其对应的Python扩展。** 实现新的op都添加至目录[paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)下，文件命名以`*_op.h`（如有） 、 `*_op.cc` 、`*_op.cu`（如有）结尾。**系统会根据文件名自动构建op和其对应的Python扩展。**
...@@ -153,7 +153,7 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs, ...@@ -153,7 +153,7 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs,
`MulKernel`继承自`framework::OpKernel`，带有下面两个模板参数: `MulKernel`继承自`framework::OpKernel`，带有下面两个模板参数:
- `typename Place`: 表示设备类型，不同设备(CPU、GPU)共享同一个Kernel时，需加该模板参数，不共享则不加，一个不共享的例子是[`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43)。 - `typename DeviceContext`: 表示设备类型，不同设备(CPU、CUDA)共享同一个Kernel时，需加该模板参数，不共享则不加，一个不共享的例子是[`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43)。
- `typename T` : 表示数据类型，如`float`, `double`等。 - `typename T` : 表示数据类型，如`float`, `double`等。
...@@ -165,7 +165,7 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs, ...@@ -165,7 +165,7 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs,
下面是 `MulKernel` `Compute`的实现： 下面是 `MulKernel` `Compute`的实现：
```cpp ```cpp
template <typename Place, typename T> template <typename DeviceContext, typename T>
class MulKernel : public framework::OpKernel { class MulKernel : public framework::OpKernel {
public: public:
void Compute(const framework::ExecutionContext& context) const override { void Compute(const framework::ExecutionContext& context) const override {
...@@ -173,18 +173,16 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs, ...@@ -173,18 +173,16 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs,
auto* Y = context.Input<Tensor>("Y"); auto* Y = context.Input<Tensor>("Y");
auto* Z = context.Output<Tensor>("Out"); auto* Z = context.Output<Tensor>("Out");
Z->mutable_data<T>(context.GetPlace()); Z->mutable_data<T>(context.GetPlace());
auto* device_context = auto& device_context = context.template device_context<DeviceContext>();
const_cast<platform::DeviceContext*>(context.device_context_); math::matmul<DeviceContext, T>(*X, false, *Y, false, 1, Z, 0, device_context);
math::matmul<Place, T>(*X, false, *Y, false, 1, Z, 0, device_context);
} }
}; };
```
需要注意：**不同设备(CPU、GPU)共享一个Op定义，是否则共享同一个`OpKernel`，取决于`Compute`调用的函数是否支持不同设备。** 需要注意：**不同设备(CPU、CUDA)共享一个Op定义，是否则共享同一个`OpKernel`，取决于`Compute`调用的函数是否支持不同设备。**
`MulOp`的CPU、GPU实现共享同一个`Kernel`。`OpKernel`不共享的例子可以参考：[`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43)。 `MulOp`的CPU、CUDA实现共享同一个`Kernel`。`OpKernel`不共享的例子可以参考：[`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43)。
为了使`OpKernel`的计算过程书写更加简单，并且CPU、GPU的代码可以复用，我们通常借助 Eigen unsupported Tensor模块来实现`Compute`接口。关于在PaddlePaddle中如何使用Eigen库，请参考[使用文档](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/howto/dev/use_eigen_cn.md)。 为了使`OpKernel`的计算过程书写更加简单，并且CPU、CUDA的代码可以复用，我们通常借助 Eigen unsupported Tensor模块来实现`Compute`接口。关于在PaddlePaddle中如何使用Eigen库，请参考[使用文档](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/howto/dev/use_eigen_cn.md)。
到此，前向Op实现完成。接下来，需要在`.cc`文件中注册该op和kernel。 到此，前向Op实现完成。接下来，需要在`.cc`文件中注册该op和kernel。
...@@ -197,9 +195,9 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs, ...@@ -197,9 +195,9 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs,
```cpp ```cpp
namespace ops = paddle::operators; namespace ops = paddle::operators;
REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulOpGrad); REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulOpGrad);
REGISTER_OP_CPU_KERNEL(mul, ops::MulKernel<paddle::platform::CPUPlace, float>); REGISTER_OP_CPU_KERNEL(mul, ops::MulKernel<paddle::platform::CPUDeviceContext, float>);
REGISTER_OP_CPU_KERNEL(mul_grad, REGISTER_OP_CPU_KERNEL(mul_grad,
ops::MulGradKernel<paddle::platform::CPUPlace, float>); ops::MulGradKernel<paddle::platform::CPUDeviceContext, float>);
``` ```
在上面的代码中： 在上面的代码中：
...@@ -209,17 +207,17 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs, ...@@ -209,17 +207,17 @@ MulOp(const std::string &type, const framework::VariableNameMap &inputs,
- `REGISTER_OP_CPU_KERNEL` ：注册`ops::MulKernel`类，并特化模板参数为`paddle::platform::CPUPlace`和`float`类型，同理，注册`ops::MulGradKernel`类。 - `REGISTER_OP_CPU_KERNEL` ：注册`ops::MulKernel`类，并特化模板参数为`paddle::platform::CPUPlace`和`float`类型，同理，注册`ops::MulGradKernel`类。
- 在 `.cu`文件中注册GPU Kernel。 - 在 `.cu`文件中注册CUDA Kernel。
- 请注意，如果GPU Kernel的实现基于Eigen unsupported模块，那么在 `.cu`的开始请加上宏定义 `#define EIGEN_USE_GPU`，代码示例如下： - 请注意，如果CUDA Kernel的实现基于Eigen unsupported模块，那么在 `.cu`的开始请加上宏定义 `#define EIGEN_USE_GPU`，代码示例如下：
```cpp ```cpp
// if use Eigen unsupported module before include head files // if use Eigen unsupported module before include head files
// #define EIGEN_USE_GPU #define EIGEN_USE_GPU
namespace ops = paddle::operators; namespace ops = paddle::operators;
REGISTER_OP_GPU_KERNEL(mul, ops::MulKernel<paddle::platform::GPUPlace, float>); REGISTER_OP_CUDA_KERNEL(mul, ops::MulKernel<paddle::platform::CUDADeviceContext, float>);
REGISTER_OP_GPU_KERNEL(mul_grad, REGISTER_OP_CUDA_KERNEL(mul_grad,
ops::MulGradKernel<paddle::platform::GPUPlace, float>); ops::MulGradKernel<paddle::platform::CUDADeviceContext, float>);
``` ```
### 5. 编译 ### 5. 编译
...@@ -236,71 +234,55 @@ make mul_op ...@@ -236,71 +234,55 @@ make mul_op
## 实现单元测试 ## 实现单元测试
单测包括对比前向Op不同设备(CPU、GPU)的实现、对比反向OP不同设备(CPU、GPU)的实现、反向Op的梯度测试。下面介绍介绍[`MulOp`的单元测试](https://github.com/PaddlePaddle/Paddle/blob/develop/python/paddle/v2/framework/tests/test_mul_op.py)。 单测包括对比前向Op不同设备(CPU、CUDA)的实现、对比反向OP不同设备(CPU、CUDA)的实现、反向Op的梯度测试。下面介绍介绍[`MulOp`的单元测试](https://github.com/PaddlePaddle/Paddle/blob/develop/python/paddle/v2/framework/tests/test_mul_op.py)。
### 前向Operator单元测试
前向Op单元测试继承自`unittest.TestCase`，并定义元类`__metaclass__ = OpTestMeta`。各项更加具体的单元测试在`OpTestMeta`里完成。测试前向Operator，需要： Op单元测试继承自`OpTest`。各项更加具体的单元测试在`TestMulOp`里完成。测试Operator，需要：
1. 在`setUp`函数定义输入、输出，以及相关的属性参数。 1. 在`setUp`函数定义输入、输出，以及相关的属性参数。
2. 生成随机的输入数据。 2. 生成随机的输入数据。
3. 在Python脚本中实现与前向operator相同的计算逻辑，得到输出值，与operator前向计算的输出进行对比。 3. 在Python脚本中实现与前向operator相同的计算逻辑，得到输出值，与operator前向计算的输出进行对比。
4. 反向计算已经自动集成进测试框架，直接调用相应接口即可。
```python ```python
import unittest import unittest
import numpy as np import numpy as np
from gradient_checker import GradientChecker, create_op from op_test import OpTest
from op_test_util import OpTestMeta
class TestMulOp(unittest.TestCase):
__metaclass__ = OpTestMeta
class TestMulOp(OpTest):
def setUp(self): def setUp(self):
self.type = "mul" self.op_type = "mul"
self.inputs = { self.inputs = {
'X': np.random.random((32, 84)).astype("float32"), 'X': np.random.random((32, 84)).astype("float32"),
'Y': np.random.random((84, 100)).astype("float32") 'Y': np.random.random((84, 100)).astype("float32")
} }
self.outputs = {'Out': np.dot(self.inputs['X'], self.inputs['Y'])} self.outputs = {'Out': np.dot(self.inputs['X'], self.inputs['Y'])}
```
上面的代码首先导入依赖的包，下面是对`setUp`函数中操作的重要变量的详细解释： def test_check_output(self):
self.check_output()
- `self.type = "mul" ` : 定义类型，与operator注册时注册的类型一致。
- `self.inputs` : 定义输入，类型为`numpy.array`，并初始化。
- `self.outputs` : 定义输出，并在Python脚本中完成与operator同样的计算逻辑，返回Python端的计算结果。
### 反向Operator单元测试
反向Op单元测试继承自`GradientChecker`，而`GradientChecker`继承自`unittest.TestCase`，因此，**反向单元测试函数需要以`test_`开头**。 def test_check_grad_normal(self):
self.check_grad(['X', 'Y'], 'Out', max_relative_error=0.5)
```python def test_check_grad_ingore_x(self):
class TestMulGradOp(GradientChecker): self.check_grad(
def setUp(self): ['Y'], 'Out', max_relative_error=0.5, no_grad_set=set("X"))
self.op = create_op("mul")
self.inputs = {
'X': np.random.random((32, 84)).astype("float32"),
'Y': np.random.random((84, 100)).astype("float32")
}
def test_check_grad_normal(self): def test_check_grad_ingore_y(self):
# mul op will enlarge the relative error self.check_grad(
self.check_grad(['X', 'Y'], 'Out', max_relative_error=0.5) ['X'], 'Out', max_relative_error=0.5, no_grad_set=set('Y'))
def test_check_grad_ingore_x(self): ```
self.check_grad(
['Y'], 'Out', max_relative_error=0.5, no_grad_set=set("X"))
def test_check_grad_ingore_y(self): 上面的代码首先导入依赖的包，下面是对`setUp`函数中操作的重要变量的详细解释：
self.check_grad(
['X'], 'Out', max_relative_error=0.5, no_grad_set=set('Y'))
```
下面解释代码中一些关键的地方: - `self.op_type = "mul" ` : 定义类型，与operator注册时注册的类型一致。
- `self.inputs` : 定义输入，类型为`numpy.array`，并初始化。
- `self.outputs` : 定义输出，并在Python脚本中完成与operator同样的计算逻辑，返回Python端的计算结果。
- 调用`create_op("mul")`创建反向Op对应的前向Op。 而反向测试中：
- `test_check_grad_normal`中调用`check_grad`使用数值法检测梯度正确性和稳定性。 - `test_check_grad_normal`中调用`check_grad`使用数值法检测梯度正确性和稳定性。
- 第一个参数`["X", "Y"]` : 指定对输入变量`X`、`Y`做梯度检测。 - 第一个参数`["X", "Y"]` : 指定对输入变量`X`、`Y`做梯度检测。
- 第二个参数`"Out"` : 指定前向网络最终的输出目标变量`Out`。 - 第二个参数`"Out"` : 指定前向网络最终的输出目标变量`Out`。
...@@ -328,5 +310,5 @@ ctest -R test_mul_op ...@@ -328,5 +310,5 @@ ctest -R test_mul_op
- 为每个Op创建单独的`*_op.h`（如有）、`*_op.cc`和`*_op.cu`（如有）。不允许一个文件中包含多个Op，这将会导致编译出错。 - 为每个Op创建单独的`*_op.h`（如有）、`*_op.cc`和`*_op.cu`（如有）。不允许一个文件中包含多个Op，这将会导致编译出错。
- 注册Op时的类型名，需要和该Op的名字一样。即不允许在`A_op.cc`里面，注册`REGISTER_OP(B, ...)`等，这将会导致单元测试出错。 - 注册Op时的类型名，需要和该Op的名字一样。即不允许在`A_op.cc`里面，注册`REGISTER_OP(B, ...)`等，这将会导致单元测试出错。
- 如果Op没有实现GPU Kernel，请不要创建空的`*_op.cu`，这将会导致单元测试出错。 - 如果Op没有实现CUDA Kernel，请不要创建空的`*_op.cu`，这将会导致单元测试出错。
- 如果多个Op依赖一些共用的函数，可以创建非`*_op.*`格式的文件来存放，如`gather.h`文件。 - 如果多个Op依赖一些共用的函数，可以创建非`*_op.*`格式的文件来存放，如`gather.h`文件。
doc/howto/dev/new_op_en.md
浏览文件 @ de8c4627
...@@ -28,8 +28,8 @@ An operator can be differentiated by whether in has kernel methods. An operator ...@@ -28,8 +28,8 @@ An operator can be differentiated by whether in has kernel methods. An operator
-------------- | :---------------------- -------------- | :----------------------
OpProtoMake definition | `.cc`files, Backward Op does not need an OpProtoMake interface. OpProtoMake definition | `.cc`files, Backward Op does not need an OpProtoMake interface.
Op definition | `.cc` files Op definition | `.cc` files
Kernel implementation | The kernel methods shared between CPU and GPU are defined in `.h` files. CPU-specific kernels live in `.cc` files, while GPU-specific kernels are implemented in `.cu`files. Kernel implementation | The kernel methods shared between CPU and CUDA are defined in `.h` files. CPU-specific kernels live in `.cc` files, while CUDA-specific kernels are implemented in `.cu`files.
Registering the Op | Ops are registered in `.cc` files; For Kernel registration, `.cc` files contain the CPU implementation, while `.cu` files contain the GPU implementation. Registering the Op | Ops are registered in `.cc` files; For Kernel registration, `.cc` files contain the CPU implementation, while `.cu` files contain the CUDA implementation.
New Operator implementations are added to the list [paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators), with file names in the format `*_op.h` (if applicable), `*_op.cc`, `*_op.cu` (if applicable).** The system will use the naming scheme to automatically build operators and their corresponding Python extensions. ** New Operator implementations are added to the list [paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators), with file names in the format `*_op.h` (if applicable), `*_op.cc`, `*_op.cu` (if applicable).** The system will use the naming scheme to automatically build operators and their corresponding Python extensions. **
...@@ -151,7 +151,7 @@ Usually `OpProtoMaker` and `Op`'s type definitions are written in `.cc` files, w ...@@ -151,7 +151,7 @@ Usually `OpProtoMaker` and `Op`'s type definitions are written in `.cc` files, w
`MulKernel` inherits `framework::OpKernel`, which includes the following templates: `MulKernel` inherits `framework::OpKernel`, which includes the following templates:
- `typename Place` denotes device type. When different devices, namely the CPU and the GPU, share the same kernel, this template needs to be added. If they don't share kernels, this must not be added. An example of a non-sharing kernel is [`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43). - `typename DeviceContext` denotes device context type. When different devices, namely the CPUDeviceContext and the CUDADeviceContext, share the same kernel, this template needs to be added. If they don't share kernels, this must not be added. An example of a non-sharing kernel is [`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43).
- `typename T` denotes data type, such as `float` or `double`. - `typename T` denotes data type, such as `float` or `double`.
...@@ -163,7 +163,7 @@ Usually `OpProtoMaker` and `Op`'s type definitions are written in `.cc` files, w ...@@ -163,7 +163,7 @@ Usually `OpProtoMaker` and `Op`'s type definitions are written in `.cc` files, w
`MulKernel`'s implementation of `Compute` is as follows: `MulKernel`'s implementation of `Compute` is as follows:
```cpp ```cpp
template <typename Place, typename T> template <typename DeviceContext, typename T>
class MulKernel : public framework::OpKernel { class MulKernel : public framework::OpKernel {
public: public:
void Compute(const framework::ExecutionContext& context) const override { void Compute(const framework::ExecutionContext& context) const override {
...@@ -171,16 +171,15 @@ Usually `OpProtoMaker` and `Op`'s type definitions are written in `.cc` files, w ...@@ -171,16 +171,15 @@ Usually `OpProtoMaker` and `Op`'s type definitions are written in `.cc` files, w
auto* Y = context.Input<Tensor>("Y"); auto* Y = context.Input<Tensor>("Y");
auto* Z = context.Output<Tensor>("Out"); auto* Z = context.Output<Tensor>("Out");
Z->mutable_data<T>(context.GetPlace()); Z->mutable_data<T>(context.GetPlace());
auto* device_context = auto& device_context = context.template device_context<DeviceContext>();
const_cast<platform::DeviceContext*>(context.device_context_); math::matmul<DeviceContext, T>(*X, false, *Y, false, 1, Z, 0, device_context);
math::matmul<Place, T>(*X, false, *Y, false, 1, Z, 0, device_context);
} }
}; };
``` ```
Note that **different devices (CPU, GPU)share an Op definition; whether or not they share the same `OpKernel` depends on whether `Compute` calls functions that support both devices.** Note that **different devices (CPU, CUDA)share an Op definition; whether or not they share the same `OpKernel` depends on whether `Compute` calls functions that support both devices.**
`MulOp`'s CPU and GPU share the same `Kernel`. A non-sharing `OpKernel` example can be seen in [`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43). `MulOp`'s CPU and CUDA share the same `Kernel`. A non-sharing `OpKernel` example can be seen in [`OnehotCrossEntropyOpKernel`](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/cross_entropy_op.h#L43).
To ease the writing of `OpKernel` compute, and for reusing code cross-device, [`Eigen-unsupported Tensor`](https://bitbucket.org/eigen/eigen/src/default/unsupported/Eigen/CXX11/src/Tensor/README.md?fileviewer=file-view-default) module is used to implement `Compute` interface. To learn about how the Eigen library is used in PaddlePaddle, please see [usage document](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/howto/dev/use_eigen_cn.md). To ease the writing of `OpKernel` compute, and for reusing code cross-device, [`Eigen-unsupported Tensor`](https://bitbucket.org/eigen/eigen/src/default/unsupported/Eigen/CXX11/src/Tensor/README.md?fileviewer=file-view-default) module is used to implement `Compute` interface. To learn about how the Eigen library is used in PaddlePaddle, please see [usage document](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/howto/dev/use_eigen_cn.md).
...@@ -196,9 +195,9 @@ The definition of its corresponding backward operator, if applicable, is similar ...@@ -196,9 +195,9 @@ The definition of its corresponding backward operator, if applicable, is similar
```cpp ```cpp
namespace ops = paddle::operators; namespace ops = paddle::operators;
REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulOpGrad); REGISTER_OP(mul, ops::MulOp, ops::MulOpMaker, mul_grad, ops::MulOpGrad);
REGISTER_OP_CPU_KERNEL(mul, ops::MulKernel<paddle::platform::CPUPlace, float>); REGISTER_OP_CPU_KERNEL(mul, ops::MulKernel<paddle::platform::CPUDeviceContext, float>);
REGISTER_OP_CPU_KERNEL(mul_grad, REGISTER_OP_CPU_KERNEL(mul_grad,
ops::MulGradKernel<paddle::platform::CPUPlace, float>); ops::MulGradKernel<paddle::platform::CPUDeviceContext, float>);
``` ```
In that code block, In that code block,
...@@ -208,17 +207,17 @@ The definition of its corresponding backward operator, if applicable, is similar ...@@ -208,17 +207,17 @@ The definition of its corresponding backward operator, if applicable, is similar
- `REGISTER_OP_CPU_KERNEL` registers `ops::MulKernel` class and specialized template types `paddle::platform::CPUPlace` and `float`, which also registers `ops::MulGradKernel`. - `REGISTER_OP_CPU_KERNEL` registers `ops::MulKernel` class and specialized template types `paddle::platform::CPUPlace` and `float`, which also registers `ops::MulGradKernel`.
- Registering GPU Kernel in `.cu` files - Registering CUDA Kernel in `.cu` files
- Note that if GPU Kernel is implemented using the `Eigen unsupported` module, then on top of `.cu`, a macro definition `#define EIGEN_USE_GPU` is needed, such as - Note that if CUDA Kernel is implemented using the `Eigen unsupported` module, then on top of `.cu`, a macro definition `#define EIGEN_USE_GPU` is needed, such as
```cpp ```cpp
// if use Eigen unsupported module before include head files // if use Eigen unsupported module before include head files
#define EIGEN_USE_GPU #define EIGEN_USE_GPU
namespace ops = paddle::operators; namespace ops = paddle::operators;
REGISTER_OP_GPU_KERNEL(mul, ops::MulKernel<paddle::platform::GPUPlace, float>); REGISTER_OP_CUDA_KERNEL(mul, ops::MulKernel<paddle::platform::CUDADeviceContext, float>);
REGISTER_OP_GPU_KERNEL(mul_grad, REGISTER_OP_CUDA_KERNEL(mul_grad,
ops::MulGradKernel<paddle::platform::GPUPlace, float>); ops::MulGradKernel<paddle::platform::CUDADeviceContext, float>);
``` ```
### 5. Compilation ### 5. Compilation
...@@ -253,62 +252,50 @@ A forward operator unit test inherits `unittest.TestCase` and defines metaclass ...@@ -253,62 +252,50 @@ A forward operator unit test inherits `unittest.TestCase` and defines metaclass
2. Generating random input data. 2. Generating random input data.
3. Implementing the same computation logic in a Python script: 3. Implementing the same computation logic in a Python script.
4. Call check gradient function to check the backward operator.
```python ```python
import unittest import unittest
import numpy as np import numpy as np
from gradient_checker import GradientChecker, create_op from op_test import OpTest
from op_test_util import OpTestMeta
class TestMulOp(unittest.TestCase):
__metaclass__ = OpTestMeta
class TestMulOp(OpTest):
def setUp(self): def setUp(self):
self.type = "mul" self.op_type = "mul"
self.inputs = { self.inputs = {
'X': np.random.random((32, 84)).astype("float32"), 'X': np.random.random((32, 84)).astype("float32"),
'Y': np.random.random((84, 100)).astype("float32") 'Y': np.random.random((84, 100)).astype("float32")
} }
self.outputs = {'Out': np.dot(self.inputs['X'], self.inputs['Y'])} self.outputs = {'Out': np.dot(self.inputs['X'], self.inputs['Y'])}
```
Get its output, and compare it with the forward operator's own output.
The code above first loads required packages. In addition, we have
- `self.type = "mul" ` defines the type that is identical to what the operator's registered type.
- `self.inputs` defines input, with type `numpy.array` and initializes it.
- `self.outputs` defines output and completes the same operator computation in the Python script, and returns its result from the Python script.
### Testing Backward Operators def test_check_output(self):
self.check_output()
def test_check_grad_normal(self):
self.check_grad(['X', 'Y'], 'Out', max_relative_error=0.5)
A backward operator unit test inherits `GradientChecker`, which inherits `unittest.TestCase`. As a result, **a backward operator unit test needs to be have the prefix `test_`**. def test_check_grad_ingore_x(self):
self.check_grad(
['Y'], 'Out', max_relative_error=0.5, no_grad_set=set("X"))
```python def test_check_grad_ingore_y(self):
class TestMulGradOp(GradientChecker): self.check_grad(
def setUp(self): ['X'], 'Out', max_relative_error=0.5, no_grad_set=set('Y'))
self.op = create_op("mul")
self.inputs = {
'X': np.random.random((32, 84)).astype("float32"),
'Y': np.random.random((84, 100)).astype("float32")
}
def test_check_grad_normal(self): ```
# mul op will enlarge the relative error Get its output, and compare it with the forward operator's own output.
self.check_grad(['X', 'Y'], 'Out', max_relative_error=0.5)
def test_check_grad_ingore_x(self): The code above first loads required packages. In addition, we have
self.check_grad(
['Y'], 'Out', max_relative_error=0.5, no_grad_set=set("X"))
def test_check_grad_ingore_y(self): - `self.op_type = "mul" ` defines the type that is identical to what the operator's registered type.
self.check_grad( - `self.inputs` defines input, with type `numpy.array` and initializes it.
['X'], 'Out', max_relative_error=0.5, no_grad_set=set('Y')) - `self.outputs` defines output and completes the same operator computation in the Python script, and returns its result from the Python script.
```
Some key points in the code above include: Some key points in checking gradient above include:
- `create_op("mul")` creates the backward operator's corresponding forward operator.
- `test_normal` calls `check_grad` to validate scaling tests' correctness and stability through numeric methods. - `test_normal` calls `check_grad` to validate scaling tests' correctness and stability through numeric methods.
- The first variable `["X", "Y"]` appoints `X` and `Y` to be scale tested. - The first variable `["X", "Y"]` appoints `X` and `Y` to be scale tested.
- The second variable `"Out"` points to the network's final output target `Out`. - The second variable `"Out"` points to the network's final output target `Out`.
...@@ -338,5 +325,5 @@ ctest -R test_mul_op ...@@ -338,5 +325,5 @@ ctest -R test_mul_op
- Every `*_op.h` (if applicable), `*_op.cc`, and `*_op.cu` (if applicable) must be created for a unique Op. Compiling will fail if multiple operators are included per file. - Every `*_op.h` (if applicable), `*_op.cc`, and `*_op.cu` (if applicable) must be created for a unique Op. Compiling will fail if multiple operators are included per file.
- The type with which an operator is registered needs to be identical to the Op's name. Registering `REGISTER_OP(B, ...)` in `A_op.cc` will cause unit testing failures. - The type with which an operator is registered needs to be identical to the Op's name. Registering `REGISTER_OP(B, ...)` in `A_op.cc` will cause unit testing failures.
- If the operator does not implement a GPU kernel, please refrain from creating an empty `*_op.cu` file, or else unit tests will fail. - If the operator does not implement a CUDA kernel, please refrain from creating an empty `*_op.cu` file, or else unit tests will fail.
- If multiple operators rely on some shared methods, a file NOT named `*_op.*` can be created to store them, such as `gather.h`. - If multiple operators rely on some shared methods, a file NOT named `*_op.*` can be created to store them, such as `gather.h`.
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