Base class for LBANN SGD-family training algorithms. More...

#include <sgd_training_algorithm.hpp>

Inheritance diagram for lbann::SGDTrainingAlgorithm:

Collaboration diagram for lbann::SGDTrainingAlgorithm:

Public Member Functions
	SGDTrainingAlgorithm (std::string name, std::unique_ptr< SGDTerminationCriteria > stop, bool suppress_timer_output)
	Construct with a name. More...

	SGDTrainingAlgorithm (const SGDTrainingAlgorithm &other)=delete

SGDTrainingAlgorithm &	operator= (const SGDTrainingAlgorithm &other)=delete

	SGDTrainingAlgorithm (SGDTrainingAlgorithm &&other)=default

SGDTrainingAlgorithm &	operator= (SGDTrainingAlgorithm &&other)=default

virtual	~SGDTrainingAlgorithm ()=default

std::string	get_type () const override

void	apply (ExecutionContext &c, model &model, data_coordinator &dc, execution_mode mode) override

void	train (SGDExecutionContext &c, model &model, data_coordinator &dc, SGDTerminationCriteria const &term)

void	evaluate (SGDExecutionContext &c, model &model, data_coordinator &dc, execution_mode mode, SGDTerminationCriteria const &term)

std::unique_ptr< SGDExecutionContext >	get_new_execution_context () const
	Get a default-initialized execution context. More...

Public Member Functions inherited from lbann::TrainingAlgorithm
	TrainingAlgorithm (std::string name)
	Constructor. More...

virtual	~TrainingAlgorithm ()=default

std::string const &	get_name () const noexcept
	A user-defined string identifying the algorithm object. More...

void	apply (model &model, data_coordinator &dc)
	Apply the algorithm to the given model. More...

void	setup_models (std::vector< observer_ptr< model >> const &models, size_t max_mini_batch_size, const std::vector< El::Grid *> &grids)
	Setup a collection of models. More...

std::unique_ptr< ExecutionContext >	get_new_execution_context () const
	Get a default-initialized execution context that fits this training algorithm. More...

Protected Member Functions
bool	train_mini_batch (SGDExecutionContext &c, model &model, data_coordinator &dc, ScopeTimer timer)

bool	evaluate_mini_batch (SGDExecutionContext &c, model &model, data_coordinator &dc, execution_mode mode, ScopeTimer timer)

void	do_train_begin_cbs (model &model, ScopeTimer timer)

void	do_train_end_cbs (model &model, ScopeTimer timer)

void	do_evaluate_begin_cbs (model &model, execution_mode mode, ScopeTimer timer)

void	do_evaluate_end_cbs (model &model, execution_mode mode, ScopeTimer timer)

void	do_epoch_begin_cbs (model &model, ScopeTimer timer)

void	do_epoch_end_cbs (model &model, ScopeTimer timer)

void	do_batch_begin_cbs (model &model, execution_mode mode, ScopeTimer timer)

void	do_batch_end_cbs (model &model, execution_mode mode, ScopeTimer timer)

SGDExecutionContext *	do_get_new_execution_context () const override
	Covariant return-friendly implementation of `get_new_exection_context()`. More...

Protected Member Functions inherited from lbann::TrainingAlgorithm
	TrainingAlgorithm (const TrainingAlgorithm &other)=delete

TrainingAlgorithm &	operator= (const TrainingAlgorithm &other)=delete

	TrainingAlgorithm (TrainingAlgorithm &&other)=default

TrainingAlgorithm &	operator= (TrainingAlgorithm &&other)=default

Private Attributes
TimerMap	m_timers

std::unique_ptr< SGDTerminationCriteria >	m_stopping_criteria

SGDExecutionContext	m_validation_context

size_t	m_validation_epochs

bool	m_suppress_timer = false
	Suppress timer output. More...

Detailed Description

Base class for LBANN SGD-family training algorithms.

Definition at line 50 of file sgd_training_algorithm.hpp.

Constructor & Destructor Documentation

◆ SGDTrainingAlgorithm() [1/3]

lbann::SGDTrainingAlgorithm::SGDTrainingAlgorithm	(	std::string	name,
		std::unique_ptr< SGDTerminationCriteria >	stop,
		bool	suppress_timer_output
	)

Construct with a name.

◆ SGDTrainingAlgorithm() [2/3]

lbann::SGDTrainingAlgorithm::SGDTrainingAlgorithm ( const SGDTrainingAlgorithm & other )

delete

◆ SGDTrainingAlgorithm() [3/3]

lbann::SGDTrainingAlgorithm::SGDTrainingAlgorithm ( SGDTrainingAlgorithm && other )

default

◆ ~SGDTrainingAlgorithm()

virtual lbann::SGDTrainingAlgorithm::~SGDTrainingAlgorithm ( )

virtualdefault

Member Function Documentation

◆ apply()

void lbann::SGDTrainingAlgorithm::apply	(	ExecutionContext &	c,
		model &	model,
		data_coordinator &	dc,
		execution_mode	mode
	)

overridevirtual

Apply the training algorithm to the model with the provided context and execution mode

Implements lbann::TrainingAlgorithm.

◆ do_batch_begin_cbs()

void lbann::SGDTrainingAlgorithm::do_batch_begin_cbs	(	model &	model,
		execution_mode	mode,
		ScopeTimer	timer
	)

protected

Execute callbacks at start of mini-batch.

◆ do_batch_end_cbs()

void lbann::SGDTrainingAlgorithm::do_batch_end_cbs	(	model &	model,
		execution_mode	mode,
		ScopeTimer	timer
	)

protected

Execute callbacks at end of mini-batch.

◆ do_epoch_begin_cbs()

void lbann::SGDTrainingAlgorithm::do_epoch_begin_cbs	(	model &	model,
		ScopeTimer	timer
	)

protected

Execute callbacks at start of epoch.

◆ do_epoch_end_cbs()

void lbann::SGDTrainingAlgorithm::do_epoch_end_cbs	(	model &	model,
		ScopeTimer	timer
	)

protected

Execute callbacks at end of epoch.

◆ do_evaluate_begin_cbs()

void lbann::SGDTrainingAlgorithm::do_evaluate_begin_cbs	(	model &	model,
		execution_mode	mode,
		ScopeTimer	timer
	)

protected

Execute callbacks at start of evaluation.

◆ do_evaluate_end_cbs()

void lbann::SGDTrainingAlgorithm::do_evaluate_end_cbs	(	model &	model,
		execution_mode	mode,
		ScopeTimer	timer
	)

protected

Execute callbacks at end of evaluation.

◆ do_get_new_execution_context()

SGDExecutionContext* lbann::SGDTrainingAlgorithm::do_get_new_execution_context ( ) const

overrideprotectedvirtual

Covariant return-friendly implementation of get_new_exection_context().

Implements lbann::TrainingAlgorithm.

◆ do_train_begin_cbs()

void lbann::SGDTrainingAlgorithm::do_train_begin_cbs	(	model &	model,
		ScopeTimer	timer
	)

protected

Execute callbacks at start of training.

◆ do_train_end_cbs()

void lbann::SGDTrainingAlgorithm::do_train_end_cbs	(	model &	model,
		ScopeTimer	timer
	)

protected

Execute callbacks at end of training.

◆ evaluate()

void lbann::SGDTrainingAlgorithm::evaluate	(	SGDExecutionContext &	c,
		model &	model,
		data_coordinator &	dc,
		execution_mode	mode,
		SGDTerminationCriteria const &	term
	)

Evaluate a model using the forward pass of an SGD solver.

◆ evaluate_mini_batch()

bool lbann::SGDTrainingAlgorithm::evaluate_mini_batch	(	SGDExecutionContext &	c,
		model &	model,
		data_coordinator &	dc,
		execution_mode	mode,
		ScopeTimer	timer
	)

protected

Evaluate model on one step / mini-batch of an SGD forward pass

◆ get_new_execution_context()

std::unique_ptr<SGDExecutionContext> lbann::SGDTrainingAlgorithm::get_new_execution_context ( ) const

Get a default-initialized execution context.

Note: This method participates in the "covariant-smart-pointer-return" pattern. In particular, it hides the base-class method to give the illusion of a covariant return.

◆ get_type()

std::string lbann::SGDTrainingAlgorithm::get_type ( ) const

overridevirtual

Copy training_algorithm.

Implements lbann::TrainingAlgorithm.

◆ operator=() [1/2]

SGDTrainingAlgorithm& lbann::SGDTrainingAlgorithm::operator= ( const SGDTrainingAlgorithm & other )

delete

◆ operator=() [2/2]

SGDTrainingAlgorithm& lbann::SGDTrainingAlgorithm::operator= ( SGDTrainingAlgorithm && other )

default

◆ train()

void lbann::SGDTrainingAlgorithm::train	(	SGDExecutionContext &	c,
		model &	model,
		data_coordinator &	dc,
		SGDTerminationCriteria const &	term
	)

Train a model using an iterative SGD solver.

◆ train_mini_batch()

bool lbann::SGDTrainingAlgorithm::train_mini_batch	(	SGDExecutionContext &	c,
		model &	model,
		data_coordinator &	dc,
		ScopeTimer	timer
	)

protected

Train model on one step / mini-batch of an SGD forward pass

Member Data Documentation

◆ m_stopping_criteria

std::unique_ptr<SGDTerminationCriteria> lbann::SGDTrainingAlgorithm::m_stopping_criteria

private

Definition at line 142 of file sgd_training_algorithm.hpp.

◆ m_suppress_timer

bool lbann::SGDTrainingAlgorithm::m_suppress_timer = false

private

Suppress timer output.

Deprecated:: This is a temporary way to disable timer output. This will be more configurable in the future.

Definition at line 156 of file sgd_training_algorithm.hpp.

◆ m_timers

TimerMap lbann::SGDTrainingAlgorithm::m_timers

private

Definition at line 141 of file sgd_training_algorithm.hpp.

◆ m_validation_context

SGDExecutionContext lbann::SGDTrainingAlgorithm::m_validation_context

private

Definition at line 148 of file sgd_training_algorithm.hpp.

◆ m_validation_epochs

size_t lbann::SGDTrainingAlgorithm::m_validation_epochs

private

Definition at line 149 of file sgd_training_algorithm.hpp.

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

sgd_training_algorithm.hpp

Public Member Functions

Protected Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ SGDTrainingAlgorithm() [1/3]

◆ SGDTrainingAlgorithm() [2/3]

◆ SGDTrainingAlgorithm() [3/3]

◆ ~SGDTrainingAlgorithm()

Member Function Documentation

◆ apply()

◆ do_batch_begin_cbs()

◆ do_batch_end_cbs()

◆ do_epoch_begin_cbs()

◆ do_epoch_end_cbs()

◆ do_evaluate_begin_cbs()

◆ do_evaluate_end_cbs()

◆ do_get_new_execution_context()

◆ do_train_begin_cbs()

◆ do_train_end_cbs()

◆ evaluate()

◆ evaluate_mini_batch()

◆ get_new_execution_context()

◆ get_type()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ train()

◆ train_mini_batch()

Member Data Documentation

◆ m_stopping_criteria

◆ m_suppress_timer

◆ m_timers

◆ m_validation_context

◆ m_validation_epochs