oa::Optimizer

base Optimizer

ClC++ public headers

Public Methods

oa::F32 oa::Optimizer::getLr()

oa::U64 oa::Optimizer::getStep()

oa::Status oa::Optimizer::loadFrom(Engine & inEngine, const ModelFile & inFile)

oa::F32 oa::Optimizer::lr()

void oa::Optimizer::notifyProgramReplay(oa::U64 inCount = 1)

oa::Status oa::Optimizer::saveTo(Engine & inEngine, ModelFile & outFile)

void oa::Optimizer::setLr(oa::F32 inLr)

void oa::Optimizer::step()

oa::Status oa::Optimizer::validateLoad(const ModelFile & inFile)

void oa::Optimizer::zeroGrad()

Constructor & Destructor Documentation

void oa::Optimizer::~Optimizer()

No public source comment is attached to this declaration.

Public Method Documentation

oa::F32 oa::Optimizer::getLr()

No public source comment is attached to this declaration.

Returns

oa::F32

The declared return value.

oa::U64 oa::Optimizer::getStep()

current step count.

Returns

oa::U64

The declared return value.

oa::Status oa::Optimizer::loadFrom(
	Engine & inEngine,
	const ModelFile & inFile
)

No public source comment is attached to this declaration.

Parameters

inEngine

Engine &

—

inFile

const ModelFile &

—

Returns

oa::Status

The declared return value.

oa::F32 oa::Optimizer::lr()

No public source comment is attached to this declaration.

Returns

oa::F32

The declared return value.

void oa::Optimizer::notifyProgramReplay(
	oa::U64 inCount = 1
)

A compiled training program executes the already-recorded optimizer kernels without calling step() again. Keep the host-visible logical step aligned so schedules and checkpoints observe the same state as eager execution.

Parameters

inCount

oa::U64

Default: 1

Returns

void

The declared return value.

oa::Status oa::Optimizer::saveTo(
	Engine & inEngine,
	ModelFile & outFile
)

Persistence — write/read optimizer state (moments, step count, hyperparams) into a ModelFile section. Default no-op so SGD/Adam compile until they implement.

Parameters

inEngine

Engine &

—

outFile

ModelFile &

—

Returns

oa::Status

The declared return value.

void oa::Optimizer::setLr(
	oa::F32 inLr
)

get/set learning rate.

Parameters

inLr

oa::F32

—

Returns

void

The declared return value.

void oa::Optimizer::step()

apply accumulated gradients to parameters.

Returns

void

The declared return value.

oa::Status oa::Optimizer::validateLoad(
	const ModelFile & inFile
)

No public source comment is attached to this declaration.

Parameters

inFile

const ModelFile &

—

Returns

oa::Status

The declared return value.

void oa::Optimizer::zeroGrad()

Zero all gradients.

Returns

void

The declared return value.