ceras::view_4d< T > Struct Template Reference

#include <tensor.hpp>

Inheritance diagram for ceras::view_4d< T >:

Public Member Functions
constexpr	view_4d (T *data=nullptr, unsigned long batch_size=0, unsigned long row=0, unsigned long col=0, unsigned long channel=0) noexcept
constexpr auto	operator[] (unsigned long index) noexcept
constexpr auto	operator[] (unsigned long index) const noexcept

Public Attributes
T *	data_
	The pointer to the start position of the 1-D array. More...
unsigned long	batch_size_
	The batch size of the 4-D tensor, also the first dimension of the tensor. More...
unsigned long	row_
	The row of the 4-D tensor, also the second dimension of the tensor. More...
unsigned long	col_
	The column of the 4-D tensor, also the third dimension of the tensor. More...
unsigned long	channel_
	The channel of the 4-D tensor, also the last dimension of the tensor. More...

Detailed Description

template<typename T>
struct ceras::view_4d< T >

A class viewing a 1-D array as a 4-D tensor. This class is useful when treating an array as a typical 4-D tensor in a neural network, with a shape of [batch_size, row, column, channel].

Constructor & Destructor Documentation

view_4d()

template<typename T >

constexpr ceras::view_4d< T >::view_4d ( T *  data = nullptr, unsigned long  batch_size = 0, unsigned long  row = 0, unsigned long  col = 0, unsigned long  channel = 0  )

inlineconstexprnoexcept

Constructor of view_4d

Parameters

data	The raw pointer to the start position of the 1-D array.
batch_size	The first dimension of the 4-D tensor, also for the batch size in the CNN layers.
row	The second dimension of the 4-D tensor, also for the row in the CNN layers.
col	The third dimension of the 4-D tensor, also for the column in the CNN layers.
channel	The last dimension of the 4-D tensor, also for the channel in the CNN layers.

Member Function Documentation

operator[]() [1/2]

template<typename T >

constexpr auto ceras::view_4d< T >::operator[] ( unsigned long  index ) const

inlineconstexprnoexcept

Giving a view_3d interface for operator [].

Parameters

index

The first dimension of the 4-D tensor.

Example usage:

std::vector<float> array;
array.resize( 16*8*8*3 );
// operations on `array`
auto t = view_4d{ array.data(), 16, 8, 8, 3 };
float v0123 = t[0][1][2][3];

operator[]() [2/2]

template<typename T >

constexpr auto ceras::view_4d< T >::operator[] ( unsigned long  index )

inlineconstexprnoexcept

Giving a view_3d interface for operator [].

Parameters

index

The first dimension of the 4-D tensor.

Example usage:

std::vector<float> array;
array.resize( 16*8*8*3 );
auto t = view_4d{ array.data(), 16, 8, 8, 3 };
t[0][1][2][3] = 1.0;

Member Data Documentation

batch_size_

template<typename T >

unsigned long ceras::view_4d< T >::batch_size_

The batch size of the 4-D tensor, also the first dimension of the tensor.

channel_

template<typename T >

unsigned long ceras::view_4d< T >::channel_

The channel of the 4-D tensor, also the last dimension of the tensor.

col_

template<typename T >

unsigned long ceras::view_4d< T >::col_

The column of the 4-D tensor, also the third dimension of the tensor.

data_

template<typename T >

T* ceras::view_4d< T >::data_

The pointer to the start position of the 1-D array.

row_

template<typename T >

unsigned long ceras::view_4d< T >::row_

The row of the 4-D tensor, also the second dimension of the tensor.

---

The documentation for this struct was generated from the following file:

• /home/feng/workspace/github.repo/ceras/include/tensor.hpp