apollo::prediction::network::RnnModel类 参考

RnnModel is a derived class from NetModel, it has a specific layers structure. 更多...

#include <rnn_model.h>

类 apollo::prediction::network::RnnModel 继承关系图:

apollo::prediction::network::RnnModel 的协作图:

Public 成员函数
void	Run (const std::vector< Eigen::MatrixXf > &inputs, Eigen::MatrixXf *output) const override
	Compute the model output from inputs according to a defined layers' flow
void	SetState (const std::vector< Eigen::MatrixXf > &states) override
	Set the internal state of a network model
void	State (std::vector< Eigen::MatrixXf > *states) const override
	Access to the internal state of a network model
void	ResetState () const override
	Set the internal state of a model
Public 成员函数 继承自 apollo::prediction::network::NetModel
	NetModel ()=default
	Constructor
virtual	~NetModel ()=default
	Destructor
bool	LoadModel (const NetParameter &net_parameter)
	Load network parameters from a protobuf message
std::string	PerformanceString () const
	Shows the performance information of a network
const std::string &	Name () const
	Name of a network model
int	Id () const
	Id of a network model
bool	IsOk () const
	Indicate the state of a network model

额外继承的成员函数
Protected 属性 继承自 apollo::prediction::network::NetModel
std::vector< std::unique_ptr< Layer > >	layers_
NetParameter	net_parameter_
bool	ok_ = false

详细描述

RnnModel is a derived class from NetModel, it has a specific layers structure.

在文件 rnn_model.h 第 37 行定义.

成员函数说明

ResetState()

void apollo::prediction::network::RnnModel::ResetState ( ) const

overridevirtual

Set the internal state of a model

参数

A	specified internal state in a vector of Eigen::MatrixXf

重载 apollo::prediction::network::NetModel .

在文件 rnn_model.cc 第 74 行定义.

                                {
  layers_[4]->ResetState();
  layers_[5]->ResetState();
}

Run()

void apollo::prediction::network::RnnModel::Run ( const std::vector< Eigen::MatrixXf > &  inputs, Eigen::MatrixXf *  output  ) const

overridevirtual

Compute the model output from inputs according to a defined layers' flow

参数

Inputs	to the network
Output	of the network will be returned

实现了 apollo::prediction::network::NetModel.

在文件 rnn_model.cc 第 25 行定义.

                                                {
  Eigen::MatrixXf inp1;
  Eigen::MatrixXf inp2;
  layers_[0]->Run({inputs[0]}, &inp1);
  layers_[1]->Run({inputs[1]}, &inp2);
 
  Eigen::MatrixXf bn1;
  Eigen::MatrixXf bn2;
  layers_[2]->Run({inp1}, &bn1);
  layers_[3]->Run({inp2}, &bn2);
 
  Eigen::MatrixXf lstm1;
  Eigen::MatrixXf lstm2;
  layers_[4]->Run({bn1}, &lstm1);
  layers_[5]->Run({bn2}, &lstm2);
 
  Eigen::MatrixXf merge;
  Eigen::MatrixXf dense1;
  Eigen::MatrixXf act1;
  layers_[6]->Run({lstm1, lstm2}, &merge);
  layers_[7]->Run({merge}, &dense1);
  layers_[8]->Run({dense1}, &bn1);
  layers_[9]->Run({bn1}, &act1);
 
  Eigen::MatrixXf dense2;
  Eigen::MatrixXf prob;
  layers_[10]->Run({act1}, &dense2);
  layers_[12]->Run({dense2}, &bn1);
  layers_[14]->Run({bn1}, &prob);
 
  Eigen::MatrixXf acc;
  layers_[11]->Run({act1}, &dense2);
  layers_[13]->Run({dense2}, &bn1);
  layers_[15]->Run({bn1}, &acc);
 
  output->resize(1, 2);
  *output << prob, acc;
}

SetState()

void apollo::prediction::network::RnnModel::SetState ( const std::vector< Eigen::MatrixXf > &  states )

overridevirtual

Set the internal state of a network model

参数

A	specified internal state in a vector of Eigen::MatrixXf

重载 apollo::prediction::network::NetModel .

在文件 rnn_model.cc 第 65 行定义.

                                                              {
  layers_[4]->SetState(states);
  layers_[5]->ResetState();
}

State()

void apollo::prediction::network::RnnModel::State ( std::vector< Eigen::MatrixXf > *  states ) const

overridevirtual

Access to the internal state of a network model

返回

Internal state in a vector of Eigen::MatrixXf of the model

重载 apollo::prediction::network::NetModel .

在文件 rnn_model.cc 第 70 行定义.

                                                           {
  layers_[4]->State(states);
}

---

该类的文档由以下文件生成:

• modules/prediction/network/rnn_model/rnn_model.h
• modules/prediction/network/rnn_model/rnn_model.cc