opennn::Configuration Class Reference

Global singleton holding the OpenNN device and precision configuration. More...

#include <configuration.h>

Classes
struct	Resolved
	Resolved configuration after Auto values are mapped to concrete device and types. More...

Public Member Functions
void	set (Device new_device=Device::Auto, Type new_training_type=Type::Auto, Type new_inference_type=Type::Auto)
	Updates device and precision settings; subsequent resolve() calls reflect the change.
Device	get_device () const
Type	get_training_type () const
Type	get_inference_type () const
const Resolved &	resolve () const
	Returns the cached resolved configuration, computing it on first access.
bool	is_gpu () const
bool	is_cpu () const
bool	is_bf16_training () const
bool	is_bf16_inference () const

Static Public Member Functions
static Configuration &	instance ()
	Returns the process-wide Configuration singleton.

Detailed Description

Global singleton holding the OpenNN device and precision configuration.

Member Function Documentation

get_device()

Device opennn::Configuration::get_device ( ) const

inlinenodiscard

get_inference_type()

Type opennn::Configuration::get_inference_type ( ) const

inlinenodiscard

get_training_type()

Type opennn::Configuration::get_training_type ( ) const

inlinenodiscard

instance()

static Configuration & opennn::Configuration::instance ( )

inlinestaticnodiscard

Returns the process-wide Configuration singleton.

is_bf16_inference()

bool opennn::Configuration::is_bf16_inference ( ) const

inlinenodiscard

is_bf16_training()

bool opennn::Configuration::is_bf16_training ( ) const

inlinenodiscard

is_cpu()

bool opennn::Configuration::is_cpu ( ) const

inlinenodiscard

is_gpu()

bool opennn::Configuration::is_gpu ( ) const

inlinenodiscard

resolve()

const Resolved & opennn::Configuration::resolve ( ) const

inlinenodiscard

Returns the cached resolved configuration, computing it on first access.

set()

void opennn::Configuration::set	(	Device	new_device = Device::Auto,
		Type	new_training_type = Type::Auto,
		Type	new_inference_type = Type::Auto )

Updates device and precision settings; subsequent resolve() calls reflect the change.

Parameters

new_device	Target execution device (Auto picks CUDA when available).
new_training_type	Numeric type used during training.
new_inference_type	Numeric type used during inference.