nvneural::WeightsData< TVectorData > Class Template Reference

Standard IWeightsData implementation containing a vector of values in CPU memory. More...

#include <WeightsData.h>

Inheritance diagram for nvneural::WeightsData< TVectorData >:

Public Member Functions
NetworkBackendId	backendId () const noexcept
	Returns the backend for which this weights data is relevant. More...
const void *	data () const noexcept
	Returns a pointer to the first element of the weights buffer. More...
TensorDimension	dimension () const noexcept
	Returns the size of the weights buffer in elements. More...
bool	memManagedExternally () const noexcept
	Returns true if the memory is managed outside NvNeural and should not be freed or reallocated by the framework. More...
TensorFormat	tensorFormat () const noexcept
	Returns the tensor format for the weights buffer. More...
	WeightsData (TensorFormat format, TensorDimension dimension, std::vector< TVectorData > &&container)
	Creates a new WeightsData object. More...
Public Member Functions inherited from nvneural::refobj::RefObjectBase< refobj::Implements< IWeightsData > >
IRefObject::RefCount	addRef () const noexcept
	Increment the object's reference count. More...
const void *	queryInterface (IRefObject::TypeId interfaceId) const noexcept
	This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.
void *	queryInterface (IRefObject::TypeId interfaceId) noexcept
	Retrieves a new object interface pointer. More...
	RefObjectBase ()
	Default constructor. Logs object creation.
IRefObject::RefCount	release () const noexcept
	Decrements the object's reference count and destroys the object if the reference count reaches zero. More...

Detailed Description

template<typename TVectorData>
class nvneural::WeightsData< TVectorData >

Standard IWeightsData implementation containing a vector of values in CPU memory.

Typical weights loader implementations will move the data vector into this object.

Template Parameters

TVectorData

Element type contained in the vector

Constructor & Destructor Documentation

WeightsData()

template<typename TVectorData >

WeightsData::WeightsData	(	TensorFormat	format,
		TensorDimension	dimension,
		std::vector< TVectorData > &&	container
	)

Creates a new WeightsData object.

Parameters

format	Tensor format (data type and layout); type is not inferred from the container parameter
dimension	Tensor dimension of the weights data
container	Rvalue reference to vector storage containing the entire weights tensor

Member Function Documentation

backendId()

template<typename TVectorData >

NetworkBackendId WeightsData::backendId

noexcept

Returns the backend for which this weights data is relevant.

data()

template<typename TVectorData >

const void * WeightsData::data

noexcept

Returns a pointer to the first element of the weights buffer.

The address space for this pointer depends on the backend:

• Cpu: Pointer is in CPU memory.
• Cuda: Pointer is a CUdeviceptr.

dimension()

template<typename TVectorData >

TensorDimension WeightsData::dimension

noexcept

Returns the size of the weights buffer in elements.

memManagedExternally()

template<typename TVectorData >

bool WeightsData::memManagedExternally

noexcept

Returns true if the memory is managed outside NvNeural and should not be freed or reallocated by the framework.

tensorFormat()

template<typename TVectorData >

TensorFormat WeightsData::tensorFormat

noexcept

Returns the tensor format for the weights buffer.

---

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

• Host/Inc/WeightsData.h
• Host/Src/WeightsData.cpp