UE::NNE::FTensorShape Class Reference

#include <NNETypes.h>

Public Member Functions
	FTensorShape ()=default
	FTensorShape (const FTensorShape &OtherShape)=default
TConstArrayView< uint32 >	GetData () const
int32	Rank () const
NNE_API uint64	Volume () const
NNE_API bool	IsCompatibleWith (const FSymbolicTensorShape &SymbolicShape) const
NNE_API bool	operator== (const FTensorShape &OtherShape) const
NNE_API bool	operator!= (const FTensorShape &OtherShape) const
NNE_API void	operator= (const FTensorShape &OtherShape)

Static Public Member Functions
static NNE_API FTensorShape	Make (TConstArrayView< uint32 > Data)
static NNE_API FTensorShape	MakeFromSymbolic (const FSymbolicTensorShape &SymbolicShape)

Static Public Attributes
static constexpr int32	MaxRank = FSymbolicTensorShape::MaxRank

Detailed Description

The concrete shape of a tensor.

Concrete tensor shapes are well defined through strictly positive values and thus have a defined volume.

Constructor & Destructor Documentation

FTensorShape() [1/2]

UE::NNE::FTensorShape::FTensorShape ( )

default

Default construct a tensor shape.

FTensorShape() [2/2]

UE::NNE::FTensorShape::FTensorShape ( const FTensorShape &  OtherShape )

default

Construct this tensor shape from another one.

Parameters

OtherShape

the other tensor shape to copy from.

Member Function Documentation

GetData()

TConstArrayView< uint32 > UE::NNE::FTensorShape::GetData ( ) const

inline

Get the dimensions of the tensor shape.

Returns

a view on the dimensions of the tensor shape.

IsCompatibleWith()

bool UE::NNE::FTensorShape::IsCompatibleWith

(

const FSymbolicTensorShape &

SymbolicShape

)

const

Check if this tensor shape is part of the ensemble of shapes defined by the symbolic shape.

For example [1,2] is compatible with [-1,2] and [1,2] but not with [2,2] or [1,2,3].

Returns

true if this tensor shape is compatible with the symbolic shape, false otherwise.

Make()

FTensorShape UE::NNE::FTensorShape::Make ( TConstArrayView< uint32 >  Data )

static

Construct a tensor shape with the given dimensions.

Parameters

Data	an array of dimensions.

Returns

a tensor shape with the given dimensions.

MakeFromSymbolic()

FTensorShape UE::NNE::FTensorShape::MakeFromSymbolic ( const FSymbolicTensorShape &  SymbolicShape )

static

Construct a tensor shape from a symbolic tensor shape.

Negative dimension values are replaced by 1.

Parameters

SymbolicShape

a symbolic tensor shape.

Returns

a tensor shape with the given dimensions.

operator!=()

bool UE::NNE::FTensorShape::operator!=

(

const FTensorShape &

OtherShape

)

const

Check if the tensor shape is different from another one.

Parameters

OtherShape

the other tensor shape.

Returns

true if the tensor shape is different from the other one, false otherwise.

operator=()

void UE::NNE::FTensorShape::operator=

(

const FTensorShape &

OtherShape

)

Assign this tensor shape from another one.

Parameters

OtherShape

the other tensor shape to copy from.

operator==()

bool UE::NNE::FTensorShape::operator==

(

const FTensorShape &

OtherShape

)

const

Check if the tensor shape is equal to another one.

Parameters

OtherShape

the other tensor shape.

Returns

true if the tensor shape is equal to the other one, false otherwise.

Rank()

int32 UE::NNE::FTensorShape::Rank ( ) const

inline

Get the number of dimensions of the tensor shape, aka the rank.

Returns

the number of dimensions of the tensor shape.

Volume()

uint64 UE::NNE::FTensorShape::Volume

(

)

const

Get the number of elements in this tensor shape.

Returns

the number of element in this tensor shape.

Member Data Documentation

MaxRank

constexpr int32 UE::NNE::FTensorShape::MaxRank = FSymbolicTensorShape::MaxRank

staticconstexpr

The maximum number of dimensions supported by the tensor shape, aka the rank.

---

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

• Engine/Source/Runtime/NNE/Public/NNETypes.h
• Engine/Source/Runtime/NNE/Private/NNETypes.cpp