Per-optimizer scratch state (moments, directions, iteration counter) backing the update step. More...

#include <optimizer.h>

Public Member Functions
	OptimizerData ()=default

virtual	~OptimizerData ()=default

virtual void	print () const
	Prints the optimizer scratch state for debugging.

void	set (const vector< Shape > &slot_shapes, Device device=Device::CPU)
	Allocates a buffer with the requested slot shapes on the target device and refreshes the views.

Public Attributes
Buffer	data

vector< TensorView >	views

VectorR	potential_parameters

VectorR	training_direction

float	initial_learning_rate = 0.0f

Index	iteration = 0

Detailed Description

Per-optimizer scratch state (moments, directions, iteration counter) backing the update step.

Constructor & Destructor Documentation

◆ OptimizerData()

opennn::OptimizerData::OptimizerData ( )

default

◆ ~OptimizerData()

virtual opennn::OptimizerData::~OptimizerData ( )

virtualdefault

Member Function Documentation

◆ print()

virtual void opennn::OptimizerData::print ( ) const

virtual

Prints the optimizer scratch state for debugging.

◆ set()

void opennn::OptimizerData::set	(	const vector< Shape > &	slot_shapes,
		Device	device = Device::CPU )

Allocates a buffer with the requested slot shapes on the target device and refreshes the views.

Member Data Documentation

◆ data

Buffer opennn::OptimizerData::data

◆ initial_learning_rate

float opennn::OptimizerData::initial_learning_rate = 0.0f

◆ iteration

Index opennn::OptimizerData::iteration = 0

◆ potential_parameters

VectorR opennn::OptimizerData::potential_parameters

◆ training_direction

VectorR opennn::OptimizerData::training_direction

◆ views

vector<TensorView> opennn::OptimizerData::views

Public Member Functions

Public Attributes