Tensor Class Reference

Tensor Class Reference#

6 min read time

Applies to Linux

rocCV: roccv::Tensor Class Reference

#include <tensor.hpp>

Public Types
using	Requirements = TensorRequirements

Public Member Functions
const eDeviceType	device () const
	Returns the location (device or host) of the tensor data. More...

const DataType &	dtype () const
	Returns the data type of the tensor. More...

TensorData	exportData () const
	Exports the tensor data of the tensor. More...

template<typename DerivedTensorData >
DerivedTensorData	exportData () const
	Exports tensor data and casts it to a specified tensor data object. More...

const TensorLayout &	layout () const
	Returns the layout of the tensor. More...

Tensor &	operator= (const Tensor &other)

int	rank () const
	Returns the rank of the tensor (i.e. the number of dimensions) More...

Tensor	reshape (const TensorShape &new_shape) const
	Creates a view of this tensor with a new shape and layout. More...

Tensor	reshape (const TensorShape &new_shape, const DataType &new_dtype) const
	Creates a vew of this tensor with a new shape, layout, and data type. The number of bytes allocated must match the original tensor. More...

const TensorShape &	shape () const
	Returns the shape of the tensor. More...

const int64_t	shape (int d) const &
	Retrieves a specific dimension size from the tensor shape. More...

	Tensor (const Tensor &other)=delete

	Tensor (const TensorRequirements &reqs)
	Constructs a Tensor object given a list of requirements. Creating a tensor through this constructor will automatically allocate the required amount of space on either the device or host. More...

	Tensor (const TensorRequirements &reqs, std::shared_ptr< TensorStorage > data)
	Constructs a Tensor object given a list of requirements and the underlying data as a TensorStorage pointer. This constructor will not automatically allocate data. More...

	Tensor (const TensorShape &shape, DataType dtype, const eDeviceType device=eDeviceType::GPU)
	Constructs a tensor object and allocates the appropriate amount of memory on the specified device. More...

	Tensor (int num_images, Size2D image_size, ImageFormat fmt, eDeviceType device=eDeviceType::GPU)
	Constructs a tensor using image-based requirements and allocates the appropriate amount of memory on the specified device. More...

	Tensor (Tensor &&other)

Static Public Member Functions
static Requirements	CalcRequirements (const TensorShape &shape, DataType dtype, const eDeviceType device=eDeviceType::GPU)
	Calculates tensor requirements. This essentially wraps the provided parameters into a TensorRequirements object. More...

static Requirements	CalcRequirements (int num_images, Size2D image_size, ImageFormat fmt, eDeviceType device=eDeviceType::GPU)
	Calculates tensor requirements using image-based parameters. More...

Member Typedef Documentation

◆ Requirements

using roccv::Tensor::Requirements = TensorRequirements

Constructor & Destructor Documentation

◆ Tensor() [1/6]

roccv::Tensor::Tensor ( const TensorRequirements & reqs )

explicit

Constructs a Tensor object given a list of requirements. Creating a tensor through this constructor will automatically allocate the required amount of space on either the device or host.

Parameters

[in] reqs An object representing the requirements for this tensor.

◆ Tensor() [2/6]

roccv::Tensor::Tensor	(	const TensorRequirements &	reqs,
		std::shared_ptr< TensorStorage >	data
	)

explicit

Constructs a Tensor object given a list of requirements and the underlying data as a TensorStorage pointer. This constructor will not automatically allocate data.

Parameters

[in]	reqs	An object representing the requirements for this tensor.
[in]	data	A TensorStorage object for the tensor's underlying data.

◆ Tensor() [3/6]

roccv::Tensor::Tensor	(	const TensorShape &	shape,
		DataType	dtype,
		const eDeviceType	device = `eDeviceType::GPU`
	)

explicit

Constructs a tensor object and allocates the appropriate amount of memory on the specified device.

Parameters

[in]	shape	The shape describing the tensor.
[in]	dtype	The underlying datatype of the tensor.
[in]	device	The device the tensor should be allocated on.

◆ Tensor() [4/6]

roccv::Tensor::Tensor	(	int	num_images,
		Size2D	image_size,
		ImageFormat	fmt,
		eDeviceType	device = `eDeviceType::GPU`
	)

explicit

Constructs a tensor using image-based requirements and allocates the appropriate amount of memory on the specified device.

Parameters

[in]	num_images	The number of images in the batch.
[in]	image_size	The size for images in the batch.
[in]	fmt	The format of the underlying image data.
[in]	device	The device the tensor should be allocated on.

◆ Tensor() [5/6]

roccv::Tensor::Tensor ( const Tensor & other )

delete

◆ Tensor() [6/6]

roccv::Tensor::Tensor ( Tensor && other )

Member Function Documentation

◆ CalcRequirements() [1/2]

static Requirements roccv::Tensor::CalcRequirements	(	const TensorShape &	shape,
		DataType	dtype,
		const eDeviceType	device = `eDeviceType::GPU`
	)

static

Calculates tensor requirements. This essentially wraps the provided parameters into a TensorRequirements object.

Parameters

[in]	shape	The desired shape of the tensor.
[in]	dtype	The desired data type of the tensor's raw data.
[in]	device	The deivce the tensor data should belong to.

Returns: A TensorRequirements object representing this tensor's requirements.

◆ CalcRequirements() [2/2]

static Requirements roccv::Tensor::CalcRequirements	(	int	num_images,
		Size2D	image_size,
		ImageFormat	fmt,
		eDeviceType	device = `eDeviceType::GPU`
	)

static

Calculates tensor requirements using image-based parameters.

Parameters

[in]	num_images	The number of images in the batch.
[in]	image_size	The size for images in the batch.
[in]	fmt	The format of the underlying image data.
[in]	device	The deivce the tensor data should belong to.

Returns: A TensorRequirements object representing the tensor's requirements.

◆ device()

const eDeviceType roccv::Tensor::device ( ) const

Returns the location (device or host) of the tensor data.

Returns: The location of the tensor data.

◆ dtype()

const DataType& roccv::Tensor::dtype ( ) const

Returns the data type of the tensor.

Returns: Data type of the tensor

◆ exportData() [1/2]

TensorData roccv::Tensor::exportData ( ) const

Exports the tensor data of the tensor.

Returns: Tensor data of the tensor

◆ exportData() [2/2]

template<typename DerivedTensorData >

DerivedTensorData roccv::Tensor::exportData ( ) const

inline

Exports tensor data and casts it to a specified tensor data object.

Template Parameters

The	tensor data object to cast this tensor's data to

Returns: The tensor data casted to the tensor data object specified

◆ layout()

const TensorLayout& roccv::Tensor::layout ( ) const

Returns the layout of the tensor.

Returns: Layout of the tensor

◆ operator=()

Tensor& roccv::Tensor::operator= ( const Tensor & other )

◆ rank()

int roccv::Tensor::rank ( ) const

Returns the rank of the tensor (i.e. the number of dimensions)

Returns: An integer representing the rank of the tensor

◆ reshape() [1/2]

Tensor roccv::Tensor::reshape ( const TensorShape & new_shape ) const

Creates a view of this tensor with a new shape and layout.

Parameters

[in] new_shape the new shape of the tensor

Returns: Tensor

◆ reshape() [2/2]

Tensor roccv::Tensor::reshape	(	const TensorShape &	new_shape,
		const DataType &	new_dtype
	)		const

Creates a vew of this tensor with a new shape, layout, and data type. The number of bytes allocated must match the original tensor.

Parameters

new_shape	The new tensor shape.
new_dtype	The new data type of the underlying tensor data.

Returns: Tensor

◆ shape() [1/2]

const TensorShape& roccv::Tensor::shape ( ) const

Returns the shape of the tensor.

Returns: Shape of the tensor

◆ shape() [2/2]

const int64_t roccv::Tensor::shape ( int d ) const &

Retrieves a specific dimension size from the tensor shape.

Parameters

[in] d The index of the dimension.

Returns: The size of the specified dimension.

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

/home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-roccv/checkouts/latest/include/core/tensor.hpp

Tensor Class Reference

Tensor Class Reference#

Public Types

Public Member Functions

Static Public Member Functions

Member Typedef Documentation

◆ Requirements

Constructor & Destructor Documentation

◆ Tensor() [1/6]

◆ Tensor() [2/6]

◆ Tensor() [3/6]

◆ Tensor() [4/6]

◆ Tensor() [5/6]

◆ Tensor() [6/6]

Member Function Documentation

◆ CalcRequirements() [1/2]

◆ CalcRequirements() [2/2]

◆ device()

◆ dtype()

◆ exportData() [1/2]

◆ exportData() [2/2]

◆ layout()

◆ operator=()

◆ rank()

◆ reshape() [1/2]

◆ reshape() [2/2]

◆ shape() [1/2]

◆ shape() [2/2]