Trainer responsible of fitting a neural network on predefined data. More...

#include <LibtorchANNTrainer.h>

Inheritance diagram for LibtorchANNTrainer:

Public Types
typedef DataFileName	DataFileParameterType

Public Member Functions
	LibtorchANNTrainer (const InputParameters &parameters)
	Construct using input parameters.

virtual void	preTrain () override
	Contains processes which are executed before the training loop.

virtual void	train () override
	Contains processes which are executed for every sample in the training loop.

virtual void	postTrain () override
	Contains processes which are executed after the training loop.

virtual void	initialize () final

virtual void	execute () final

virtual void	finalize () final

virtual void	threadJoin (const UserObject &) final

virtual Real	spatialValue (const Point &) const

virtual const std::vector< Point >	spatialPoints () const

void	setPrimaryThreadCopy (UserObject *primary)

UserObject *	primaryThreadCopy ()

SubProblem &	getSubProblem () const

bool	shouldDuplicateInitialExecution () const

void	gatherSum (T &value)

void	gatherMax (T &value)

void	gatherMin (T &value)

void	gatherProxyValueMax (T1 &proxy, T2 &value)

void	gatherProxyValueMin (T1 &proxy, T2 &value)

std::set< UserObjectName >	getDependObjects () const

const std::set< std::string > &	getRequestedItems () override

const std::set< std::string > &	getSuppliedItems () override

unsigned int	systemNumber () const

virtual bool	needThreadedCopy () const

virtual bool	enabled () const

std::shared_ptr< MooseObject >	getSharedPtr ()

std::shared_ptr< const MooseObject >	getSharedPtr () const

bool	isKokkosObject () const

MooseApp &	getMooseApp () const

const std::string &	type () const

const std::string &	name () const

std::string	typeAndName () const

MooseObjectParameterName	uniqueParameterName (const std::string &parameter_name) const

MooseObjectName	uniqueName () const

const InputParameters &	parameters () const

const hit::Node *	getHitNode () const

bool	hasBase () const

const std::string &	getBase () const

const T &	getParam (const std::string &name) const

std::vector< std::pair< T1, T2 > >	getParam (const std::string &param1, const std::string &param2) const

const T *	queryParam (const std::string &name) const

const T &	getRenamedParam (const std::string &old_name, const std::string &new_name) const

T	getCheckedPointerParam (const std::string &name, const std::string &error_string="") const

bool	isParamValid (const std::string &name) const

bool	isParamSetByUser (const std::string &name) const

void	connectControllableParams (const std::string &parameter, const std::string &object_type, const std::string &object_name, const std::string &object_parameter) const

void	paramError (const std::string &param, Args... args) const

void	paramWarning (const std::string &param, Args... args) const

void	paramWarning (const std::string &param, Args... args) const

void	paramInfo (const std::string &param, Args... args) const

std::string	messagePrefix (const bool hit_prefix=true) const

std::string	errorPrefix (const std::string &) const

void	mooseError (Args &&... args) const

void	mooseDocumentedError (const std::string &repo_name, const unsigned int issue_num, Args &&... args) const

void	mooseErrorNonPrefixed (Args &&... args) const

void	mooseWarning (Args &&... args) const

void	mooseWarning (Args &&... args) const

void	mooseWarningNonPrefixed (Args &&... args) const

void	mooseWarningNonPrefixed (Args &&... args) const

void	mooseDeprecated (Args &&... args) const

void	mooseDeprecated (Args &&... args) const

void	mooseDeprecatedNoTrace (Args &&... args) const

void	mooseInfo (Args &&... args) const

void	callMooseError (std::string msg, const bool with_prefix, const hit::Node *node=nullptr, const bool show_trace=true) const

std::string	getDataFileName (const std::string &param) const

std::string	getDataFileNameByName (const std::string &relative_path) const

std::string	getDataFilePath (const std::string &relative_path) const

virtual void	initialSetup ()

virtual void	timestepSetup ()

virtual void	jacobianSetup ()

virtual void	residualSetup ()

virtual void	customSetup (const ExecFlagType &)

const ExecFlagEnum &	getExecuteOnEnum () const

UserObjectName	getUserObjectName (const std::string &param_name) const

const T &	getUserObject (const std::string &param_name, bool is_dependency=true) const

const T &	getUserObjectByName (const UserObjectName &object_name, bool is_dependency=true) const

const UserObjectBase &	getUserObjectBase (const std::string &param_name, bool is_dependency=true) const

const UserObjectBase &	getUserObjectBaseByName (const UserObjectName &object_name, bool is_dependency=true) const

const std::vector< MooseVariableScalar * > &	getCoupledMooseScalarVars ()

const std::set< TagID > &	getScalarVariableCoupleableVectorTags () const

const std::set< TagID > &	getScalarVariableCoupleableMatrixTags () const

const GenericMaterialProperty< T, is_ad > &	getGenericMaterialProperty (const std::string &name, MaterialData &material_data, const unsigned int state=0)

const GenericMaterialProperty< T, is_ad > &	getGenericMaterialProperty (const std::string &name, const unsigned int state=0)

const GenericMaterialProperty< T, is_ad > &	getGenericMaterialProperty (const std::string &name, const unsigned int state=0)

const MaterialProperty< T > &	getMaterialProperty (const std::string &name, MaterialData &material_data, const unsigned int state=0)

const MaterialProperty< T > &	getMaterialProperty (const std::string &name, const unsigned int state=0)

const MaterialProperty< T > &	getMaterialProperty (const std::string &name, const unsigned int state=0)

const ADMaterialProperty< T > &	getADMaterialProperty (const std::string &name, MaterialData &material_data)

const ADMaterialProperty< T > &	getADMaterialProperty (const std::string &name)

const ADMaterialProperty< T > &	getADMaterialProperty (const std::string &name)

const MaterialProperty< T > &	getMaterialPropertyOld (const std::string &name, MaterialData &material_data)

const MaterialProperty< T > &	getMaterialPropertyOld (const std::string &name)

const MaterialProperty< T > &	getMaterialPropertyOld (const std::string &name)

const MaterialProperty< T > &	getMaterialPropertyOlder (const std::string &name, MaterialData &material_data)

const MaterialProperty< T > &	getMaterialPropertyOlder (const std::string &name)

const MaterialProperty< T > &	getMaterialPropertyOlder (const std::string &name)

const GenericMaterialProperty< T, is_ad > &	getGenericMaterialPropertyByName (const MaterialPropertyName &name, MaterialData &material_data, const unsigned int state)

const GenericMaterialProperty< T, is_ad > &	getGenericMaterialPropertyByName (const MaterialPropertyName &name, const unsigned int state=0)

const GenericMaterialProperty< T, is_ad > &	getGenericMaterialPropertyByName (const MaterialPropertyName &name, const unsigned int state=0)

const MaterialProperty< T > &	getMaterialPropertyByName (const MaterialPropertyName &name, MaterialData &material_data, const unsigned int state=0)

const MaterialProperty< T > &	getMaterialPropertyByName (const MaterialPropertyName &name, const unsigned int state=0)

const MaterialProperty< T > &	getMaterialPropertyByName (const MaterialPropertyName &name, const unsigned int state=0)

const ADMaterialProperty< T > &	getADMaterialPropertyByName (const MaterialPropertyName &name, MaterialData &material_data)

const ADMaterialProperty< T > &	getADMaterialPropertyByName (const MaterialPropertyName &name)

const ADMaterialProperty< T > &	getADMaterialPropertyByName (const MaterialPropertyName &name)

const MaterialProperty< T > &	getMaterialPropertyOldByName (const MaterialPropertyName &name, MaterialData &material_data)

const MaterialProperty< T > &	getMaterialPropertyOldByName (const MaterialPropertyName &name)

const MaterialProperty< T > &	getMaterialPropertyOldByName (const MaterialPropertyName &name)

const MaterialProperty< T > &	getMaterialPropertyOlderByName (const MaterialPropertyName &name, MaterialData &material_data)

const MaterialProperty< T > &	getMaterialPropertyOlderByName (const MaterialPropertyName &name)

const MaterialProperty< T > &	getMaterialPropertyOlderByName (const MaterialPropertyName &name)

Moose::Kokkos::MaterialProperty< T, dimension >	getKokkosMaterialPropertyByName (const std::string &prop_name_in)

Moose::Kokkos::MaterialProperty< T, dimension >	getKokkosMaterialPropertyOldByName (const std::string &prop_name)

Moose::Kokkos::MaterialProperty< T, dimension >	getKokkosMaterialPropertyOlderByName (const std::string &prop_name)

Moose::Kokkos::MaterialProperty< T, dimension >	getKokkosMaterialProperty (const std::string &name)

Moose::Kokkos::MaterialProperty< T, dimension >	getKokkosMaterialPropertyOld (const std::string &name)

Moose::Kokkos::MaterialProperty< T, dimension >	getKokkosMaterialPropertyOlder (const std::string &name)

std::pair< const MaterialProperty< T > *, std::set< SubdomainID > >	getBlockMaterialProperty (const MaterialPropertyName &name)

std::pair< Moose::Kokkos::MaterialProperty< T, dimension >, std::set< SubdomainID > >	getKokkosBlockMaterialProperty (const MaterialPropertyName &name)

const GenericMaterialProperty< T, is_ad > &	getGenericZeroMaterialProperty (const std::string &name)

const GenericMaterialProperty< T, is_ad > &	getGenericZeroMaterialProperty ()

const GenericMaterialProperty< T, is_ad > &	getGenericZeroMaterialPropertyByName (const std::string &prop_name)

const MaterialProperty< T > &	getZeroMaterialProperty (Ts... args)

std::set< SubdomainID >	getMaterialPropertyBlocks (const std::string &name)

std::vector< SubdomainName >	getMaterialPropertyBlockNames (const std::string &name)

std::set< BoundaryID >	getMaterialPropertyBoundaryIDs (const std::string &name)

std::vector< BoundaryName >	getMaterialPropertyBoundaryNames (const std::string &name)

void	checkBlockAndBoundaryCompatibility (std::shared_ptr< MaterialBase > discrete)

std::unordered_map< SubdomainID, std::vector< MaterialBase * > >	buildRequiredMaterials (bool allow_stateful=true)

void	statefulPropertiesAllowed (bool)

virtual bool	getMaterialPropertyCalled () const

virtual const std::unordered_set< unsigned int > &	getMatPropDependencies () const

virtual void	resolveOptionalProperties ()

const GenericMaterialProperty< T, is_ad > &	getPossiblyConstantGenericMaterialPropertyByName (const MaterialPropertyName &prop_name, MaterialData &material_data, const unsigned int state)

bool	isImplicit ()

Moose::StateArg	determineState () const

virtual void	subdomainSetup () override

virtual void	subdomainSetup () override

bool	hasUserObject (const std::string &param_name) const

bool	hasUserObject (const std::string &param_name) const

bool	hasUserObject (const std::string &param_name) const

bool	hasUserObject (const std::string &param_name) const

bool	hasUserObjectByName (const UserObjectName &object_name) const

bool	hasUserObjectByName (const UserObjectName &object_name) const

bool	hasUserObjectByName (const UserObjectName &object_name) const

bool	hasUserObjectByName (const UserObjectName &object_name) const

const GenericOptionalMaterialProperty< T, is_ad > &	getGenericOptionalMaterialProperty (const std::string &name, const unsigned int state=0)

const GenericOptionalMaterialProperty< T, is_ad > &	getGenericOptionalMaterialProperty (const std::string &name, const unsigned int state=0)

const OptionalMaterialProperty< T > &	getOptionalMaterialProperty (const std::string &name, const unsigned int state=0)

const OptionalMaterialProperty< T > &	getOptionalMaterialProperty (const std::string &name, const unsigned int state=0)

const OptionalADMaterialProperty< T > &	getOptionalADMaterialProperty (const std::string &name)

const OptionalADMaterialProperty< T > &	getOptionalADMaterialProperty (const std::string &name)

const OptionalMaterialProperty< T > &	getOptionalMaterialPropertyOld (const std::string &name)

const OptionalMaterialProperty< T > &	getOptionalMaterialPropertyOld (const std::string &name)

const OptionalMaterialProperty< T > &	getOptionalMaterialPropertyOlder (const std::string &name)

const OptionalMaterialProperty< T > &	getOptionalMaterialPropertyOlder (const std::string &name)

MaterialBase &	getMaterial (const std::string &name)

MaterialBase &	getMaterial (const std::string &name)

MaterialBase &	getMaterialByName (const std::string &name, bool no_warn=false)

MaterialBase &	getMaterialByName (const std::string &name, bool no_warn=false)

bool	hasMaterialProperty (const std::string &name)

bool	hasMaterialProperty (const std::string &name)

bool	hasMaterialPropertyByName (const std::string &name)

bool	hasMaterialPropertyByName (const std::string &name)

bool	hasADMaterialProperty (const std::string &name)

bool	hasADMaterialProperty (const std::string &name)

bool	hasADMaterialPropertyByName (const std::string &name)

bool	hasADMaterialPropertyByName (const std::string &name)

bool	hasKokkosMaterialProperty (const std::string &name)

bool	hasKokkosMaterialProperty (const std::string &name)

bool	hasKokkosMaterialPropertyByName (const std::string &name)

bool	hasKokkosMaterialPropertyByName (const std::string &name)

bool	hasGenericMaterialProperty (const std::string &name)

bool	hasGenericMaterialProperty (const std::string &name)

bool	hasGenericMaterialPropertyByName (const std::string &name)

bool	hasGenericMaterialPropertyByName (const std::string &name)

const Function &	getFunction (const std::string &name) const

const Function &	getFunctionByName (const FunctionName &name) const

bool	hasFunction (const std::string &param_name) const

bool	hasFunctionByName (const FunctionName &name) const

Moose::Kokkos::Function	getKokkosFunction (const std::string &name) const

const T &	getKokkosFunction (const std::string &name) const

Moose::Kokkos::Function	getKokkosFunctionByName (const FunctionName &name) const

const T &	getKokkosFunctionByName (const FunctionName &name) const

bool	hasKokkosFunction (const std::string &param_name) const

bool	hasKokkosFunctionByName (const FunctionName &name) const

bool	isDefaultPostprocessorValue (const std::string &param_name, const unsigned int index=0) const

bool	hasPostprocessor (const std::string &param_name, const unsigned int index=0) const

bool	hasPostprocessorByName (const PostprocessorName &name) const

std::size_t	coupledPostprocessors (const std::string &param_name) const

const PostprocessorName &	getPostprocessorName (const std::string &param_name, const unsigned int index=0) const

const VectorPostprocessorValue &	getVectorPostprocessorValue (const std::string &param_name, const std::string &vector_name) const

const VectorPostprocessorValue &	getVectorPostprocessorValue (const std::string &param_name, const std::string &vector_name, bool needs_broadcast) const

const VectorPostprocessorValue &	getVectorPostprocessorValueByName (const VectorPostprocessorName &name, const std::string &vector_name) const

const VectorPostprocessorValue &	getVectorPostprocessorValueByName (const VectorPostprocessorName &name, const std::string &vector_name, bool needs_broadcast) const

const VectorPostprocessorValue &	getVectorPostprocessorValueOld (const std::string &param_name, const std::string &vector_name) const

const VectorPostprocessorValue &	getVectorPostprocessorValueOld (const std::string &param_name, const std::string &vector_name, bool needs_broadcast) const

const VectorPostprocessorValue &	getVectorPostprocessorValueOldByName (const VectorPostprocessorName &name, const std::string &vector_name) const

const VectorPostprocessorValue &	getVectorPostprocessorValueOldByName (const VectorPostprocessorName &name, const std::string &vector_name, bool needs_broadcast) const

const ScatterVectorPostprocessorValue &	getScatterVectorPostprocessorValue (const std::string &param_name, const std::string &vector_name) const

const ScatterVectorPostprocessorValue &	getScatterVectorPostprocessorValueByName (const VectorPostprocessorName &name, const std::string &vector_name) const

const ScatterVectorPostprocessorValue &	getScatterVectorPostprocessorValueOld (const std::string &param_name, const std::string &vector_name) const

const ScatterVectorPostprocessorValue &	getScatterVectorPostprocessorValueOldByName (const VectorPostprocessorName &name, const std::string &vector_name) const

bool	hasVectorPostprocessor (const std::string &param_name, const std::string &vector_name) const

bool	hasVectorPostprocessor (const std::string &param_name) const

bool	hasVectorPostprocessorByName (const VectorPostprocessorName &name, const std::string &vector_name) const

bool	hasVectorPostprocessorByName (const VectorPostprocessorName &name) const

const VectorPostprocessorName &	getVectorPostprocessorName (const std::string &param_name) const

T &	getSampler (const std::string &name)

Sampler &	getSampler (const std::string &name)

T &	getSamplerByName (const SamplerName &name)

Sampler &	getSamplerByName (const SamplerName &name)

virtual void	meshChanged ()

virtual void	meshDisplaced ()

PerfGraph &	perfGraph ()

const PostprocessorValue &	getPostprocessorValue (const std::string &param_name, const unsigned int index=0) const

const PostprocessorValue &	getPostprocessorValue (const std::string &param_name, const unsigned int index=0) const

const PostprocessorValue &	getPostprocessorValueOld (const std::string &param_name, const unsigned int index=0) const

const PostprocessorValue &	getPostprocessorValueOld (const std::string &param_name, const unsigned int index=0) const

const PostprocessorValue &	getPostprocessorValueOlder (const std::string &param_name, const unsigned int index=0) const

const PostprocessorValue &	getPostprocessorValueOlder (const std::string &param_name, const unsigned int index=0) const

virtual const PostprocessorValue &	getPostprocessorValueByName (const PostprocessorName &name) const

virtual const PostprocessorValue &	getPostprocessorValueByName (const PostprocessorName &name) const

const PostprocessorValue &	getPostprocessorValueOldByName (const PostprocessorName &name) const

const PostprocessorValue &	getPostprocessorValueOldByName (const PostprocessorName &name) const

const PostprocessorValue &	getPostprocessorValueOlderByName (const PostprocessorName &name) const

const PostprocessorValue &	getPostprocessorValueOlderByName (const PostprocessorName &name) const

bool	isVectorPostprocessorDistributed (const std::string &param_name) const

bool	isVectorPostprocessorDistributed (const std::string &param_name) const

bool	isVectorPostprocessorDistributedByName (const VectorPostprocessorName &name) const

bool	isVectorPostprocessorDistributedByName (const VectorPostprocessorName &name) const

const Distribution &	getDistribution (const std::string &name) const

const T &	getDistribution (const std::string &name) const

const Distribution &	getDistribution (const std::string &name) const

const T &	getDistribution (const std::string &name) const

const Distribution &	getDistributionByName (const DistributionName &name) const

const T &	getDistributionByName (const std::string &name) const

const Distribution &	getDistributionByName (const DistributionName &name) const

const T &	getDistributionByName (const std::string &name) const

const Parallel::Communicator &	comm () const

processor_id_type	n_processors () const

processor_id_type	processor_id () const

const std::string &	modelMetaDataName () const
	Accessor for the name of the model meta data.

const FileName &	getModelDataFileName () const
	Get the associated filename.

bool	hasModelData () const
	Check if we need to load model data (if the filename parameter is used)

template<>
SurrogateModel &	getSurrogateModel (const std::string &name) const

template<>
SurrogateTrainerBase &	getSurrogateTrainer (const std::string &name) const

template<>
SurrogateModel &	getSurrogateModelByName (const UserObjectName &name) const

template<>
SurrogateTrainerBase &	getSurrogateTrainerByName (const UserObjectName &name) const


template<typename T , typename... Args>
T &	declareModelData (const std::string &data_name, Args &&... args)
	Declare model data for loading from file as well as restart.


template<typename T , typename... Args>
const T &	getModelData (const std::string &data_name, Args &&... args) const
	Retrieve model data from the interface.


template<typename T = SurrogateModel>
T &	getSurrogateModel (const std::string &name) const
	Get a SurrogateModel/Trainer with a given name.

template<typename T = SurrogateTrainerBase>
T &	getSurrogateTrainer (const std::string &name) const


template<typename T = SurrogateModel>
T &	getSurrogateModelByName (const UserObjectName &name) const
	Get a sampler with a given name.

template<typename T = SurrogateTrainerBase>
T &	getSurrogateTrainerByName (const UserObjectName &name) const

Static Public Member Functions
static InputParameters	validParams ()

static void	callMooseError (MooseApp const app, const InputParameters &params, std::string msg, const bool with_prefix, const hit::Node node, const bool show_trace=true)

static void	sort (typename std::vector< T > &vector)

static void	sortDFS (typename std::vector< T > &vector)

static void	cyclicDependencyError (CyclicDependencyException< T2 > &e, const std::string &header, NameFunc &&name_func)

static void	cyclicDependencyError (CyclicDependencyException< T2 > &e, const std::string &header)

Public Attributes
	usingCombinedWarningSolutionWarnings

const ConsoleStream	_console

Static Public Attributes
static const std::string	type_param

static const std::string	name_param

static const std::string	unique_name_param

static const std::string	app_param

static const std::string	moose_base_param

static const std::string	kokkos_object_param

static constexpr PropertyValue::id_type	default_property_id

static constexpr PropertyValue::id_type	zero_property_id

static constexpr auto	SYSTEM

static constexpr auto	NAME

Protected Member Functions
template<typename T >
const T &	getTrainingData (const ReporterName &rname)

const std::vector< Real > &	getSamplerData () const

const std::vector< Real > &	getPredictorData () const

unsigned int	getCurrentSampleSize () const

unsigned int	getLocalSampleSize () const

virtual std::vector< Real >	evaluateModelError (const SurrogateModel &surr)

virtual void	addPostprocessorDependencyHelper (const PostprocessorName &name) const override

virtual void	addVectorPostprocessorDependencyHelper (const VectorPostprocessorName &name) const override

virtual void	addUserObjectDependencyHelper (const UserObjectBase &uo) const override

void	addReporterDependencyHelper (const ReporterName &reporter_name) override

void	flagInvalidSolutionInternal (const InvalidSolutionID invalid_solution_id) const

InvalidSolutionID	registerInvalidSolutionInternal (const std::string &message, const bool warning) const

const ReporterContextBase &	getReporterContextBaseByName (const ReporterName &reporter_name) const

const ReporterName &	getReporterName (const std::string &param_name) const

T &	declareRestartableData (const std::string &data_name, Args &&... args)

ManagedValue< T >	declareManagedRestartableDataWithContext (const std::string &data_name, void *context, Args &&... args)

const T &	getRestartableData (const std::string &data_name) const

T &	declareRestartableDataWithContext (const std::string &data_name, void *context, Args &&... args)

T &	declareRecoverableData (const std::string &data_name, Args &&... args)

T &	declareRestartableDataWithObjectName (const std::string &data_name, const std::string &object_name, Args &&... args)

T &	declareRestartableDataWithObjectNameWithContext (const std::string &data_name, const std::string &object_name, void *context, Args &&... args)

std::string	restartableName (const std::string &data_name) const

const T &	getMeshProperty (const std::string &data_name, const std::string &prefix)

const T &	getMeshProperty (const std::string &data_name)

bool	hasMeshProperty (const std::string &data_name, const std::string &prefix) const

bool	hasMeshProperty (const std::string &data_name, const std::string &prefix) const

bool	hasMeshProperty (const std::string &data_name) const

bool	hasMeshProperty (const std::string &data_name) const

std::string	meshPropertyName (const std::string &data_name) const

PerfID	registerTimedSection (const std::string &section_name, const unsigned int level) const

PerfID	registerTimedSection (const std::string &section_name, const unsigned int level, const std::string &live_message, const bool print_dots=true) const

std::string	timedSectionName (const std::string &section_name) const

bool	isCoupledScalar (const std::string &var_name, unsigned int i=0) const

unsigned int	coupledScalarComponents (const std::string &var_name) const

unsigned int	coupledScalar (const std::string &var_name, unsigned int comp=0) const

libMesh::Order	coupledScalarOrder (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarValue (const std::string &var_name, unsigned int comp=0) const

const ADVariableValue &	adCoupledScalarValue (const std::string &var_name, unsigned int comp=0) const

const GenericVariableValue< is_ad > &	coupledGenericScalarValue (const std::string &var_name, unsigned int comp=0) const

const GenericVariableValue< false > &	coupledGenericScalarValue (const std::string &var_name, const unsigned int comp) const

const GenericVariableValue< true > &	coupledGenericScalarValue (const std::string &var_name, const unsigned int comp) const

const VariableValue &	coupledVectorTagScalarValue (const std::string &var_name, TagID tag, unsigned int comp=0) const

const VariableValue &	coupledMatrixTagScalarValue (const std::string &var_name, TagID tag, unsigned int comp=0) const

const VariableValue &	coupledScalarValueOld (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarValueOlder (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarDot (const std::string &var_name, unsigned int comp=0) const

const ADVariableValue &	adCoupledScalarDot (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarDotDot (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarDotOld (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarDotDotOld (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarDotDu (const std::string &var_name, unsigned int comp=0) const

const VariableValue &	coupledScalarDotDotDu (const std::string &var_name, unsigned int comp=0) const

const MooseVariableScalar *	getScalarVar (const std::string &var_name, unsigned int comp) const

virtual void	checkMaterialProperty (const std::string &name, const unsigned int state)

virtual void	getKokkosMaterialPropertyHook (const std::string &, const unsigned int)

void	markMatPropRequested (const std::string &)

MaterialPropertyName	getMaterialPropertyName (const std::string &name) const

void	checkExecutionStage ()

const T &	getReporterValue (const std::string &param_name, const std::size_t time_index=0)

const T &	getReporterValue (const std::string &param_name, ReporterMode mode, const std::size_t time_index=0)

const T &	getReporterValue (const std::string &param_name, const std::size_t time_index=0)

const T &	getReporterValue (const std::string &param_name, ReporterMode mode, const std::size_t time_index=0)

const T &	getReporterValueByName (const ReporterName &reporter_name, const std::size_t time_index=0)

const T &	getReporterValueByName (const ReporterName &reporter_name, ReporterMode mode, const std::size_t time_index=0)

const T &	getReporterValueByName (const ReporterName &reporter_name, const std::size_t time_index=0)

const T &	getReporterValueByName (const ReporterName &reporter_name, ReporterMode mode, const std::size_t time_index=0)

bool	hasReporterValue (const std::string &param_name) const

bool	hasReporterValue (const std::string &param_name) const

bool	hasReporterValue (const std::string &param_name) const

bool	hasReporterValue (const std::string &param_name) const

bool	hasReporterValueByName (const ReporterName &reporter_name) const

bool	hasReporterValueByName (const ReporterName &reporter_name) const

bool	hasReporterValueByName (const ReporterName &reporter_name) const

bool	hasReporterValueByName (const ReporterName &reporter_name) const

const GenericMaterialProperty< T, is_ad > *	defaultGenericMaterialProperty (const std::string &name)

const GenericMaterialProperty< T, is_ad > *	defaultGenericMaterialProperty (const std::string &name)

const MaterialProperty< T > *	defaultMaterialProperty (const std::string &name)

const MaterialProperty< T > *	defaultMaterialProperty (const std::string &name)

const ADMaterialProperty< T > *	defaultADMaterialProperty (const std::string &name)

const ADMaterialProperty< T > *	defaultADMaterialProperty (const std::string &name)

Static Protected Member Functions
static std::string	meshPropertyName (const std::string &data_name, const std::string &prefix)

Protected Attributes
const Moose::CoordinateSystemType &	_coord_sys

const THREAD_ID	_tid

SubProblem &	_subproblem

FEProblemBase &	_fe_problem

SystemBase &	_sys

Assembly &	_assembly

const bool	_duplicate_initial_execution

std::set< std::string >	_depend_uo

const bool &	_enabled

MooseApp &	_app

Factory &	_factory

ActionFactory &	_action_factory

const std::string &	_type

const std::string &	_name

const InputParameters &	_pars

const ExecFlagEnum &	_execute_enum

const ExecFlagType &	_current_execute_flag

MooseApp &	_restartable_app

const std::string	_restartable_system_name

const THREAD_ID	_restartable_tid

const bool	_restartable_read_only

FEProblemBase &	_mci_feproblem

FEProblemBase &	_mdi_feproblem

MooseApp &	_pg_moose_app

const std::string	_prefix

FEProblemBase &	_sc_fe_problem

const THREAD_ID	_sc_tid

const Real &	_real_zero

const VariableValue &	_scalar_zero

const Point &	_point_zero

const InputParameters &	_mi_params

const std::string	_mi_name

const MooseObjectName	_mi_moose_object_name

FEProblemBase &	_mi_feproblem

SubProblem &	_mi_subproblem

const THREAD_ID	_mi_tid

const bool	_is_kokkos_object

const Moose::MaterialDataType	_material_data_type

MaterialData &	_material_data

bool	_stateful_allowed

bool	_get_material_property_called

std::vector< std::unique_ptr< PropertyValue > >	_default_properties

std::unordered_set< unsigned int >	_material_property_dependencies

const MaterialPropertyName	_get_suffix

const bool	_use_interpolated_state

const InputParameters &	_ti_params

FEProblemBase &	_ti_feproblem

bool	_is_implicit

Real &	_t

const Real &	_t_old

int &	_t_step

Real &	_dt

Real &	_dt_old

bool	_is_transient

const Parallel::Communicator &	_communicator


Sampler &	_sampler

dof_id_type	_row
	During training loop, this is the row index of the data.

dof_id_type	_local_row
	During training loop, this is the local row index of the data.

const Real *	_rval
	Response value.

const std::vector< Real > *	_rvecval
	Vector response value.

std::vector< const Real * >	_pvals
	Predictor values from reporters.

std::vector< unsigned int >	_pcols
	Columns from sampler for predictors.

unsigned int	_n_dims
	Dimension of predictor data - either _sampler.getNumberOfCols() or _pvals.size() + _pcols.size().

unsigned int &	_n_outputs
	The number of outputs.

Static Protected Attributes
static const std::string	_interpolated_old

static const std::string	_interpolated_older

Private Member Functions
void	checkIntegrity () const

void	executeTraining ()

std::vector< Real >	crossValidate ()

void	updatePredictorRow ()

const Moose::FunctionBase *	getKokkosFunctionByNameHelper (const FunctionName &name) const

const UserObjectBase &	getUserObjectFromFEProblem (const UserObjectName &object_name, const THREAD_ID tid=0) const

const T &	castUserObject (const UserObjectBase &uo_base, const std::string &param_name="") const

void	mooseObjectError (const std::string &param_name, std::stringstream &oss) const

const std::string &	userObjectType (const UserObjectBase &uo) const

const std::string &	userObjectName (const UserObjectBase &uo) const

const PostprocessorName &	getPostprocessorNameInternal (const std::string &param_name, const unsigned int index, const bool allow_default_value=true) const

bool	isDefaultPostprocessorValueByName (const PostprocessorName &name) const

PostprocessorValue	getDefaultPostprocessorValueByName (const PostprocessorName &name) const

void	checkParam (const std::string &param_name, const unsigned int index=std::numeric_limits< unsigned int >::max()) const

bool	postprocessorsAdded () const

const VectorPostprocessorValue &	getVectorPostprocessorByNameHelper (const VectorPostprocessorName &name, const std::string &vector_name, bool broadcast, std::size_t t_index) const

const VectorPostprocessorContext< VectorPostprocessorValue > &	getVectorPostprocessorContextByNameHelper (const VectorPostprocessorName &name, const std::string &vector_name) const

bool	vectorPostprocessorsAdded () const

bool	reportersAdded () const

void	possiblyCheckHasReporter (const ReporterName &reporter_name, const std::string &param_name="") const

RestartableDataValue &	registerRestartableDataOnApp (std::unique_ptr< RestartableDataValue > data, THREAD_ID tid) const

void	registerRestartableNameWithFilterOnApp (const std::string &name, Moose::RESTARTABLE_FILTER filter)

RestartableData< T > &	declareRestartableDataHelper (const std::string &data_name, void *context, Args &&... args) const

virtual std::string	meshPropertyPrefix (const std::string &data_name) const

const RestartableDataValue &	getMeshPropertyInternal (const std::string &data_name, const std::string &prefix) const

void	mooseErrorInternal (Args &&... args) const

const VariableValue *	getDefaultValue (const std::string &var_name) const

const ADVariableValue *	getADDefaultValue (const std::string &var_name) const

void	checkVar (const std::string &var_name) const

void	validateExecutionerType (const std::string &name, const std::string &fn_name) const

Moose::MaterialDataType	getMaterialDataType (const std::set< BoundaryID > &boundary_ids) const

unsigned int	getMaxQps () const

void	addConsumedPropertyName (const MooseObjectName &obj_name, const std::string &prop_name)

const PostprocessorValue &	getPostprocessorValueInternal (const std::string &param_name, unsigned int index, std::size_t t_index) const

const PostprocessorValue &	getPostprocessorValueInternal (const std::string &param_name, unsigned int index, std::size_t t_index) const

const PostprocessorValue &	getPostprocessorValueByNameInternal (const PostprocessorName &name, std::size_t t_index) const

const PostprocessorValue &	getPostprocessorValueByNameInternal (const PostprocessorName &name, std::size_t t_index) const

void	possiblyCheckHasVectorPostprocessor (const std::string &param_name, const std::string &vector_name) const

void	possiblyCheckHasVectorPostprocessor (const std::string &param_name, const std::string &vector_name) const

void	possiblyCheckHasVectorPostprocessorByName (const VectorPostprocessorName &name, const std::string &vector_name) const

void	possiblyCheckHasVectorPostprocessorByName (const VectorPostprocessorName &name, const std::string &vector_name) const

Static Private Member Functions
static const hit::Node *	getHitNode (const InputParameters &params)

static std::string	messagePrefix (const InputParameters &params, const bool hit_prefix)

Private Attributes
const std::vector< Real > &	_predictor_row
	Data from the current predictor row.

std::vector< Real >	_flattened_data
	The gathered data in a flattened form to be able to convert easily to torch::Tensor.

std::vector< Real >	_flattened_response
	The gathered response in a flattened form to be able to convert easily to torch::Tensor.

std::vector< unsigned int > &	_num_neurons_per_layer
	Number of neurons within the hidden layers (the length of this vector should be the same as _num_hidden_layers)

std::vector< std::string > &	_activation_function
	Activation functions for each hidden layer.

const std::string	_nn_filename
	Name of the pytorch output file.

const bool	_read_from_file
	Switch indicating if an already existing neural net should be read from a file or not.

Moose::LibtorchTrainingOptions	_optim_options
	The struct which contains the information for the training of the neural net.

std::shared_ptr< Moose::LibtorchArtificialNeuralNet > &	_nn
	Pointer to the neural net object (initialized as null)

const bool	_standardize_input
	If the training output should be standardized (scaled and shifted)

const bool	_standardize_output
	If the training output should be standardized (scaled and shifted)

StochasticTools::Standardizer &	_input_standardizer
	Standardizer for use with input (x)

StochasticTools::Standardizer &	_output_standardizer
	Standardizer for use with output response (y)

std::vector< Real >	_row_data
	Sampler data for the current row.

std::vector< Real >	_predictor_data
	Predictor data for current row - can be combination of Sampler and Reporter values.

const bool	_skip_unconverged
	Whether or not we are skipping samples that have unconverged solutions.

const bool *	_converged
	Whether or not the current sample has a converged solution.

unsigned int	_current_sample_size
	Number of samples used to train the model.

unsigned int	_local_sample_size
	Number of samples (locally) used to train the model.

std::unordered_map< ReporterName, std::shared_ptr< TrainingDataBase > >	_training_data
	Vector of reporter names and their corresponding values (to be filled by getTrainingData)

UserObject *	_primary_thread_copy

std::set< std::string >	_supplied_uo

const ParallelParamObject &	_parent

const MooseBase &	_si_moose_base

const FEProblemBase *	_si_problem

ExecFlagEnum	_empty_execute_enum

const MooseObject &	_fni_object

const InputParameters &	_fni_params

FEProblemBase &	_fni_feproblem

const THREAD_ID	_fni_tid

const MooseObject &	_uoi_moose_object

const FEProblemBase &	_uoi_feproblem

const THREAD_ID	_uoi_tid

const MooseObject &	_ppi_moose_object

const InputParameters &	_ppi_params

const FEProblemBase &	_ppi_feproblem

std::map< PostprocessorName, std::unique_ptr< PostprocessorValue > >	_default_values

const bool	_broadcast_by_default

const MooseObject &	_vpi_moose_object

const FEProblemBase &	_vpi_feproblem

const THREAD_ID	_vpi_tid

const InputParameters &	_ri_params

FEProblemBase &	_ri_fe_problem_base

const ReporterData &	_ri_reporter_data

const MooseObject &	_ri_moose_object

const InputParameters &	_dni_params

FEProblemBase &	_dni_feproblem

const MooseObject *const	_dni_moose_object_ptr

const InputParameters &	_si_params

FEProblemBase &	_si_feproblem

THREAD_ID	_si_tid

const RestartableDataMapName	_metaname

std::string	_restartable_name

MooseApp &	_meta_data_app

const MooseObject *const	_meta_data_object

const InputParameters &	_sc_parameters

const std::string &	_sc_name

const bool	_sc_is_implicit

std::unordered_map< std::string, std::vector< MooseVariableScalar * > >	_coupled_scalar_vars

std::unordered_map< std::string, std::unique_ptr< VariableValue > >	_default_value

std::unordered_map< std::string, std::unique_ptr< ADVariableValue > >	_dual_default_value

std::vector< MooseVariableScalar * >	_coupled_moose_scalar_vars

std::unordered_map< std::string, std::vector< MooseVariableFieldBase * > >	_sc_coupled_vars

std::set< TagID >	_sc_coupleable_vector_tags

std::set< TagID >	_sc_coupleable_matrix_tags

const MooseObject &	_mi_moose_object

const bool	_mi_boundary_restricted

const std::set< SubdomainID > &	_mi_block_ids

const std::set< BoundaryID > &	_mi_boundary_ids

std::vector< std::unique_ptr< OptionalMaterialPropertyProxyBase< MaterialPropertyInterface > > >	_optional_property_proxies

const std::string	_ti_name

const MooseObject &	_model_object
	Reference to the MooseObject that uses this interface.

const std::string	_model_meta_data_name
	The model meta data name.

PublicRestartable	_model_restartable
	Member for interfacing with the framework's restartable system.

const InputParameters &	_smi_params
	Parameters of the object with this interface.

FEProblemBase &	_smi_feproblem
	Reference to FEProblemBase instance.

const THREAD_ID	_smi_tid
	Thread ID.


std::vector< dof_id_type >	_skip_indices

const MooseEnum &	_cv_type
	Type of cross validation to perform - for now, just 'none' (no CV) or 'k_fold'.

const unsigned int &	_n_splits
	Number of splits (k) to split sampler data into.

const unsigned int &	_cv_n_trials
	Number of repeated trials of cross validation to perform.

const unsigned int &	_cv_seed
	Seed used for _cv_generator.

MooseRandom	_cv_generator
	Random number generator used for shuffling sampler rows during splitting.

const SurrogateModel *	_cv_surrogate
	SurrogateModel used to evaluate model error relative to test points.

const bool	_doing_cv
	Set to true if cross validation is being performed, controls behavior in execute().

std::vector< std::vector< Real > > &	_cv_trial_scores
	RMSE scores from each CV trial - can be grabbed by VPP or Reporter.

Detailed Description

Trainer responsible of fitting a neural network on predefined data.

Definition at line 23 of file LibtorchANNTrainer.h.

Constructor & Destructor Documentation

◆ LibtorchANNTrainer()

LibtorchANNTrainer::LibtorchANNTrainer ( const InputParameters & parameters )

Construct using input parameters.

Definition at line 66 of file LibtorchANNTrainer.C.

  : SurrogateTrainer(parameters),
    _predictor_row(getPredictorData()),
    _num_neurons_per_layer(declareModelData<std::vector<unsigned int>>(
        "num_neurons_per_layer", getParam<std::vector<unsigned int>>("num_neurons_per_layer"))),
    _activation_function(declareModelData<std::vector<std::string>>(
        "activation_function", getParam<std::vector<std::string>>("activation_function"))),
    _nn_filename(getParam<std::string>("nn_filename")),
    _read_from_file(getParam<bool>("read_from_file")),
    _nn(declareModelData<std::shared_ptr<Moose::LibtorchArtificialNeuralNet>>("nn")),
    _standardize_input(getParam<bool>("standardize_input")),
    _standardize_output(getParam<bool>("standardize_output")),
    _input_standardizer(declareModelData<StochasticTools::Standardizer>("input_standardizer")),
    _output_standardizer(declareModelData<StochasticTools::Standardizer>("output_standardizer"))
{
  // Fixing the RNG seed to make sure every experiment is the same.
  // Otherwise sampling / stochastic gradient descent would be different.
  torch::manual_seed(getParam<unsigned int>("seed"));
 
  _optim_options.optimizer_type = "adam";
  _optim_options.learning_rate = getParam<Real>("learning_rate");
  _optim_options.num_epochs = getParam<unsigned int>("num_epochs");
  _optim_options.num_batches = getParam<unsigned int>("num_batches");
  _optim_options.rel_loss_tol = getParam<Real>("rel_loss_tol");
  _optim_options.print_loss = getParam<unsigned int>("print_epoch_loss") > 0;
  _optim_options.print_epoch_loss = getParam<unsigned int>("print_epoch_loss");
  _optim_options.parallel_processes = getParam<unsigned int>("max_processes");
}

Member Function Documentation

◆ checkIntegrity()

void SurrogateTrainer::checkIntegrity ( ) const

privateinherited

Definition at line 196 of file SurrogateTrainer.C.

{
  // Check that the number of sampler columns hasn't changed
  if (_row_data.size() != _sampler.getNumberOfCols())
    mooseError("Number of sampler columns has changed.");
 
  // Check that training data is correctly sized
  for (auto & pair : _training_data)
  {
    dof_id_type rsize = pair.second->size();
    dof_id_type nrow =
        pair.second->isDistributed() ? _sampler.getNumberOfLocalRows() : _sampler.getNumberOfRows();
    if (rsize != nrow)
      mooseError("Reporter value ",
                 pair.first,
                 " of size ",
                 rsize,
                 " does not match sampler size (",
                 nrow,
                 ").");
  }
}

Referenced by SurrogateTrainer::executeTraining().

◆ crossValidate()

std::vector< Real > SurrogateTrainer::crossValidate ( )

privateinherited

Definition at line 252 of file SurrogateTrainer.C.

{
  std::vector<Real> cv_score(1, 0.0);
 
  // Get skipped indices for each split
  dof_id_type n_rows = _sampler.getNumberOfRows();
  std::vector<std::vector<dof_id_type>> split_indices;
  if (processor_id() == 0)
  {
    std::vector<dof_id_type> indices_flat(n_rows);
    std::iota(indices_flat.begin(), indices_flat.end(), 0);
    MooseUtils::shuffle(indices_flat, _cv_generator, 0);
 
    split_indices.resize(_n_splits);
    for (const auto & k : make_range(_n_splits))
    {
      const dof_id_type num_ind = n_rows / _n_splits + (k < (n_rows % _n_splits) ? 1 : 0);
      split_indices[k].insert(split_indices[k].begin(),
                              std::make_move_iterator(indices_flat.begin()),
                              std::make_move_iterator(indices_flat.begin() + num_ind));
      std::sort(split_indices[k].begin(), split_indices[k].end());
      indices_flat.erase(indices_flat.begin(), indices_flat.begin() + num_ind);
    }
  }
 
  std::vector<dof_id_type> split_ids_buffer;
  for (const auto & k : make_range(_n_splits))
  {
    if (processor_id() == 0)
      split_ids_buffer = split_indices[k];
    _communicator.broadcast(split_ids_buffer, 0);
 
    _current_sample_size = _sampler.getNumberOfRows() - split_ids_buffer.size();
 
    auto first = std::lower_bound(
        split_ids_buffer.begin(), split_ids_buffer.end(), _sampler.getLocalRowBegin());
    auto last = std::upper_bound(
        split_ids_buffer.begin(), split_ids_buffer.end(), _sampler.getLocalRowEnd());
    _skip_indices.insert(_skip_indices.begin(), first, last);
 
    _local_sample_size = _sampler.getNumberOfLocalRows() - _skip_indices.size();
 
    // Train the model
    executeTraining();
 
    // Evaluate the model
    std::vector<Real> split_mse(1, 0.0);
    std::vector<Real> row_mse(1, 0.0);
 
    auto skipped_row = _skip_indices.begin();
 
    for (dof_id_type p = _sampler.getLocalRowBegin(); p < _sampler.getLocalRowEnd(); ++p)
    {
      const std::vector<Real> row = _sampler.getNextLocalRow();
      if (skipped_row != _skip_indices.end() && p == *skipped_row)
      {
        for (unsigned int i = 0; i < _row_data.size(); ++i)
          _row_data[i] = row[i];
 
        for (auto & pair : _training_data)
          pair.second->setCurrentIndex(
              (pair.second->isDistributed() ? p - _sampler.getLocalRowBegin() : p));
 
        updatePredictorRow();
 
        row_mse = evaluateModelError(*_cv_surrogate);
 
        // Expand split_mse if needed.
        split_mse.resize(row_mse.size(), 0.0);
 
        // Increment errors
        for (unsigned int r = 0; r < split_mse.size(); ++r)
          split_mse[r] += row_mse[r];
 
        skipped_row++;
      }
    }
    gatherSum(split_mse);
 
    // Expand cv_score if necessary.
    cv_score.resize(split_mse.size(), 0.0);
 
    for (auto r : make_range(split_mse.size()))
      cv_score[r] += split_mse[r] / n_rows;
 
    _skip_indices.clear();
  }
 
  for (auto r : make_range(cv_score.size()))
    cv_score[r] = std::sqrt(cv_score[r]);
 
  return cv_score;
}

Referenced by SurrogateTrainer::execute().

◆ declareModelData()

template<typename T , typename... Args>

T & RestartableModelInterface::declareModelData	(	const std::string &	data_name,
		Args &&...	args
	)

inherited

Declare model data for loading from file as well as restart.

Definition at line 73 of file RestartableModelInterface.h.

{
  return _model_restartable.declareRestartableData<T>(data_name, std::forward<Args>(args)...);
}

◆ evaluateModelError()

std::vector< Real > SurrogateTrainer::evaluateModelError ( const SurrogateModel & surr )

protectedvirtualinherited

Definition at line 347 of file SurrogateTrainer.C.

{
  std::vector<Real> error(1, 0.0);
 
  if (_rval)
  {
    Real model_eval = surr.evaluate(_predictor_data);
    error[0] = MathUtils::pow(model_eval - (*_rval), 2);
  }
  else if (_rvecval)
  {
    error.resize(_rvecval->size());
 
    // Evaluate for vector response.
    std::vector<Real> model_eval(error.size());
    surr.evaluate(_predictor_data, model_eval);
    for (auto r : make_range(_rvecval->size()))
      error[r] = MathUtils::pow(model_eval[r] - (*_rvecval)[r], 2);
  }
 
  return error;
}

Referenced by SurrogateTrainer::crossValidate().

◆ execute()

void SurrogateTrainer::execute ( )

finalvirtualinherited

Implements GeneralUserObject.

Definition at line 176 of file SurrogateTrainer.C.

{
  if (_doing_cv)
    for (const auto & trial : make_range(_cv_n_trials))
    {
      std::vector<Real> trial_score = crossValidate();
 
      // Expand _cv_trial_scores with more columns if necessary, then insert values.
      for (unsigned int r = _cv_trial_scores.size(); r < trial_score.size(); ++r)
        _cv_trial_scores.push_back(std::vector<Real>(_cv_n_trials, 0.0));
      for (auto r : make_range(trial_score.size()))
        _cv_trial_scores[r][trial] = trial_score[r];
    }
 
  _current_sample_size = _sampler.getNumberOfRows();
  _local_sample_size = _sampler.getNumberOfLocalRows();
  executeTraining();
}

◆ executeTraining()

void SurrogateTrainer::executeTraining ( )

privateinherited

Definition at line 220 of file SurrogateTrainer.C.

{
  checkIntegrity();
  _row = _sampler.getLocalRowBegin();
  _local_row = 0;
 
  preTrain();
 
  for (_row = _sampler.getLocalRowBegin(); _row < _sampler.getLocalRowEnd(); ++_row)
  {
    // Need to do this manually in order to keep the iterators valid
    const std::vector<Real> data = _sampler.getNextLocalRow();
    for (unsigned int i = 0; i < _row_data.size(); ++i)
      _row_data[i] = data[i];
 
    // Set training data
    for (auto & pair : _training_data)
      pair.second->setCurrentIndex((pair.second->isDistributed() ? _local_row : _row));
 
    updatePredictorRow();
 
    if ((!_skip_unconverged || *_converged) &&
        std::find(_skip_indices.begin(), _skip_indices.end(), _row) == _skip_indices.end())
      train();
 
    _local_row++;
  }
 
  postTrain();
}

Referenced by SurrogateTrainer::crossValidate(), and SurrogateTrainer::execute().

◆ finalize()

virtual void SurrogateTrainer::finalize ( )

inlinefinalvirtualinherited

Reimplemented from SurrogateTrainerBase.

Definition at line 63 of file SurrogateTrainer.h.

63{}

◆ getCurrentSampleSize()

unsigned int SurrogateTrainer::getCurrentSampleSize ( ) const

inlineprotectedinherited

Definition at line 102 of file SurrogateTrainer.h.

102{ return _current_sample_size; };

Referenced by PolynomialChaosTrainer::postTrain(), and PolynomialChaosTrainer::preTrain().

◆ getLocalSampleSize()

unsigned int SurrogateTrainer::getLocalSampleSize ( ) const

inlineprotectedinherited

Definition at line 107 of file SurrogateTrainer.h.

107{ return _local_sample_size; };

Referenced by preTrain(), GaussianProcessTrainer::preTrain(), and NearestPointTrainer::preTrain().

◆ getModelData()

template<typename T , typename... Args>

const T & RestartableModelInterface::getModelData	(	const std::string &	data_name,
		Args &&...	args
	)		const

inherited

Retrieve model data from the interface.

Definition at line 80 of file RestartableModelInterface.h.

{
  return _model_restartable.getRestartableData<T>(data_name, std::forward<Args>(args)...);
}

◆ getModelDataFileName()

const FileName & RestartableModelInterface::getModelDataFileName ( ) const

inherited

Get the associated filename.

Definition at line 33 of file RestartableModelInterface.C.

{
  return _model_object.getParam<FileName>("filename");
}

◆ getPredictorData()

const std::vector< Real > & SurrogateTrainer::getPredictorData ( ) const

inlineprotectedinherited

Definition at line 97 of file SurrogateTrainer.h.

97{ return _predictor_data; };

◆ getSamplerData()

const std::vector< Real > & SurrogateTrainer::getSamplerData ( ) const

inlineprotectedinherited

Definition at line 92 of file SurrogateTrainer.h.

92{ return _row_data; };

◆ getSurrogateModel() [1/2]

template<typename T >

T & SurrogateModelInterface::getSurrogateModel ( const std::string & name ) const

inherited

Get a SurrogateModel/Trainer with a given name.

Parameters

name	The name of the parameter key of the sampler to retrieve

Returns: The sampler with name associated with the parameter 'name'

Definition at line 81 of file SurrogateModelInterface.h.

{
  return getSurrogateModelByName<T>(_smi_params.get<UserObjectName>(name));
}

Referenced by SurrogateTrainer::initialize().

◆ getSurrogateModel() [2/2]

template<>

SurrogateModel & SurrogateModelInterface::getSurrogateModel ( const std::string & name ) const

inherited

Definition at line 46 of file SurrogateModelInterface.C.

{
  return getSurrogateModelByName<SurrogateModel>(_smi_params.get<UserObjectName>(name));
}

◆ getSurrogateModelByName() [1/2]

template<typename T >

T & SurrogateModelInterface::getSurrogateModelByName ( const UserObjectName & name ) const

inherited

Get a sampler with a given name.

Parameters

name	The name of the sampler to retrieve

Returns: The sampler with name 'name'

Definition at line 88 of file SurrogateModelInterface.h.

{
  std::vector<T *> models;
  _smi_feproblem.theWarehouse()
      .query()
      .condition<AttribName>(name)
      .condition<AttribSystem>("SurrogateModel")
      .queryInto(models);
  if (models.empty())
    mooseError("Unable to find a SurrogateModel object of type " + std::string(typeid(T).name()) +
               " with the name '" + name + "'");
  return *(models[0]);
}

Referenced by CrossValidationScores::CrossValidationScores(), EvaluateSurrogate::EvaluateSurrogate(), and InverseMapping::initialSetup().

◆ getSurrogateModelByName() [2/2]

template<>

SurrogateModel & SurrogateModelInterface::getSurrogateModelByName ( const UserObjectName & name ) const

inherited

Definition at line 31 of file SurrogateModelInterface.C.

{
  std::vector<SurrogateModel *> models;
  _smi_feproblem.theWarehouse()
      .query()
      .condition<AttribName>(name)
      .condition<AttribSystem>("SurrogateModel")
      .queryInto(models);
  if (models.empty())
    mooseError("Unable to find a SurrogateModel object with the name '" + name + "'");
  return *(models[0]);
}

◆ getSurrogateTrainer() [1/2]

template<typename T >

T & SurrogateModelInterface::getSurrogateTrainer ( const std::string & name ) const

inherited

Definition at line 104 of file SurrogateModelInterface.h.

{
  return getSurrogateTrainerByName<T>(_smi_params.get<UserObjectName>(name));
}

◆ getSurrogateTrainer() [2/2]

template<>

SurrogateTrainerBase & SurrogateModelInterface::getSurrogateTrainer ( const std::string & name ) const

inherited

Definition at line 60 of file SurrogateModelInterface.C.

{
  return getSurrogateTrainerByName<SurrogateTrainerBase>(_smi_params.get<UserObjectName>(name));
}

◆ getSurrogateTrainerByName() [1/2]

template<typename T >

T & SurrogateModelInterface::getSurrogateTrainerByName ( const UserObjectName & name ) const

inherited

Definition at line 111 of file SurrogateModelInterface.h.

{
  SurrogateTrainerBase * base_ptr =
      &_smi_feproblem.getUserObject<SurrogateTrainerBase>(name, _smi_tid);
  T * obj_ptr = dynamic_cast<T *>(base_ptr);
  if (!obj_ptr)
    mooseError("Failed to find a SurrogateTrainer object of type " + std::string(typeid(T).name()) +
                   " with the name '",
               name,
               "' for the desired type.");
  return *obj_ptr;
}

Referenced by SurrogateTrainerOutput::output().

◆ getSurrogateTrainerByName() [2/2]

template<>

SurrogateTrainerBase & SurrogateModelInterface::getSurrogateTrainerByName ( const UserObjectName & name ) const

inherited

Definition at line 53 of file SurrogateModelInterface.C.

{
  return _smi_feproblem.getUserObject<SurrogateTrainerBase>(name);
}

◆ getTrainingData()

template<typename T >

const T & SurrogateTrainer::getTrainingData ( const ReporterName & rname )

protectedinherited

Definition at line 208 of file SurrogateTrainer.h.

{
  auto it = _training_data.find(rname);
  if (it != _training_data.end())
  {
    auto data = std::dynamic_pointer_cast<TrainingData<T>>(it->second);
    if (!data)
      mooseError("Reporter value ", rname, " already exists but is of different type.");
    return data->get();
  }
  else
  {
    const std::vector<T> & rval = getReporterValueByName<std::vector<T>>(rname);
    _training_data[rname] = std::make_shared<TrainingData<T>>(rval);
    return std::dynamic_pointer_cast<TrainingData<T>>(_training_data[rname])->get();
  }
}

◆ hasModelData()

bool RestartableModelInterface::hasModelData ( ) const

inherited

Check if we need to load model data (if the filename parameter is used)

Definition at line 39 of file RestartableModelInterface.C.

{
  return _model_object.isParamValid("filename");
}

◆ initialize()

void SurrogateTrainer::initialize ( )

finalvirtualinherited

Reimplemented from SurrogateTrainerBase.

Definition at line 153 of file SurrogateTrainer.C.

{
  // Figure out if data is distributed
  for (auto & pair : _training_data)
  {
    const ReporterName & name = pair.first;
    TrainingDataBase & data = *pair.second;
 
    const auto & mode = _fe_problem.getReporterData().getReporterMode(name);
    if (mode == REPORTER_MODE_DISTRIBUTED || (mode == REPORTER_MODE_ROOT && processor_id() != 0))
      data.isDistributed() = true;
    else if (mode == REPORTER_MODE_REPLICATED ||
             (mode == REPORTER_MODE_ROOT && processor_id() == 0))
      data.isDistributed() = false;
    else
      mooseError("Predictor reporter value ", name, " is not of supported mode.");
  }
 
  if (_doing_cv)
    _cv_surrogate = &getSurrogateModel("cv_surrogate");
}

◆ modelMetaDataName()

const std::string & RestartableModelInterface::modelMetaDataName ( ) const

inlineinherited

Accessor for the name of the model meta data.

Definition at line 47 of file RestartableModelInterface.h.

47{ return _model_meta_data_name; }

RestartableModelInterface::_model_meta_data_name

const std::string _model_meta_data_name

The model meta data name.

Definition RestartableModelInterface.h:61

Referenced by MappingOutput::output(), and SurrogateTrainerOutput::output().

◆ postTrain()

void LibtorchANNTrainer::postTrain ( )

overridevirtual

Contains processes which are executed after the training loop.

Reimplemented from SurrogateTrainer.

Definition at line 115 of file LibtorchANNTrainer.C.

{
  _communicator.allgather(_flattened_data);
  _communicator.allgather(_flattened_response);
 
  // Then, we create and load our Tensors
  unsigned int num_samples = _flattened_response.size();
  unsigned int num_inputs = _n_dims;
 
  // We create a neural net (for the definition of the net see the header file)
  _nn = std::make_shared<Moose::LibtorchArtificialNeuralNet>(
      _nn_filename, num_inputs, 1, _num_neurons_per_layer, _activation_function);
 
  if (_read_from_file)
    try
    {
      torch::load(_nn, _nn_filename);
      _console << "Loaded requested .pt file." << std::endl;
    }
    catch (const c10::Error & e)
    {
      mooseError("The requested pytorch file could not be loaded.\n", e.msg());
    }
 
  // The default data type in pytorch is float, while we use double in MOOSE.
  // Therefore, in some cases we have to convert Tensors to double.
  auto options = torch::TensorOptions().dtype(at::kDouble);
  torch::Tensor data_tensor =
      torch::from_blob(_flattened_data.data(), {num_samples, num_inputs}, options).to(at::kDouble);
  torch::Tensor response_tensor =
      torch::from_blob(_flattened_response.data(), {num_samples, 1}, options).to(at::kDouble);
 
  // We standardize the input/output pairs if the user requested it
  if (_standardize_input)
  {
    auto data_std_mean = torch::std_mean(data_tensor, 0);
    auto & data_std = std::get<0>(data_std_mean);
    auto & data_mean = std::get<1>(data_std_mean);
 
    data_tensor = (data_tensor - data_mean) / data_std;
 
    std::vector<Real> converted_data_mean;
    LibtorchUtils::tensorToVector(data_mean, converted_data_mean);
    std::vector<Real> converted_data_std;
    LibtorchUtils::tensorToVector(data_std, converted_data_std);
    _input_standardizer.set(converted_data_mean, converted_data_std);
  }
  else
    _input_standardizer.set(_n_dims);
 
  if (_standardize_output)
  {
    auto response_std_mean = torch::std_mean(response_tensor, 0);
    auto & response_std = std::get<0>(response_std_mean);
    auto & response_mean = std::get<1>(response_std_mean);
 
    response_tensor = (response_tensor - response_mean) / response_std;
 
    std::vector<Real> converted_response_mean;
    LibtorchUtils::tensorToVector(response_mean, converted_response_mean);
    std::vector<Real> converted_response_std;
    LibtorchUtils::tensorToVector(response_std, converted_response_std);
    _output_standardizer.set(converted_response_mean, converted_response_std);
  }
  else
    _output_standardizer.set(1);
 
  // We create a custom data set from our converted data
  Moose::LibtorchDataset my_data(data_tensor, response_tensor);
 
  // We create atrainer for our neral net and train it with the dataset
  Moose::LibtorchArtificialNeuralNetTrainer<> trainer(*_nn, comm());
  trainer.train(my_data, _optim_options);
}

◆ preTrain()

void LibtorchANNTrainer::preTrain ( )

overridevirtual

Contains processes which are executed before the training loop.

Reimplemented from SurrogateTrainer.

Definition at line 96 of file LibtorchANNTrainer.C.

{
  // Resize to number of sample points
  _flattened_data.clear();
  _flattened_response.clear();
  _flattened_data.reserve(getLocalSampleSize() * _n_dims);
  _flattened_response.reserve(getLocalSampleSize());
}

◆ threadJoin()

virtual void SurrogateTrainerBase::threadJoin ( const UserObject & )

inlinefinalvirtualinherited

Reimplemented from GeneralUserObject.

Definition at line 40 of file SurrogateTrainer.h.

40{} // GeneralUserObjects are not threaded

◆ train()

void LibtorchANNTrainer::train ( )

overridevirtual

Contains processes which are executed for every sample in the training loop.

Reimplemented from SurrogateTrainer.

Definition at line 106 of file LibtorchANNTrainer.C.

{
  for (auto & p : _predictor_row)
    _flattened_data.push_back(p);
 
  _flattened_response.push_back(*_rval);
}

◆ updatePredictorRow()

void SurrogateTrainer::updatePredictorRow ( )

privateinherited

Definition at line 371 of file SurrogateTrainer.C.

{
  unsigned int d = 0;
  for (const auto & val : _pvals)
    _predictor_data[d++] = *val;
  for (const auto & col : _pcols)
    _predictor_data[d++] = _row_data[col];
}

Referenced by SurrogateTrainer::crossValidate(), and SurrogateTrainer::executeTraining().

◆ validParams()

InputParameters LibtorchANNTrainer::validParams ( )

static

Definition at line 19 of file LibtorchANNTrainer.C.

{
  InputParameters params = SurrogateTrainer::validParams();
 
  params.addClassDescription("Trains a simple neural network using libtorch.");
 
  params.addRangeCheckedParam<unsigned int>(
      "num_batches", 1, "1<=num_batches", "Number of batches.");
  params.addRangeCheckedParam<unsigned int>(
      "num_epochs", 1, "0<num_epochs", "Number of training epochs.");
  params.addRangeCheckedParam<Real>(
      "rel_loss_tol",
      0,
      "0<=rel_loss_tol<=1",
      "The relative loss where we stop the training of the neural net.");
  params.addParam<std::vector<unsigned int>>(
      "num_neurons_per_layer", std::vector<unsigned int>(), "Number of neurons per layer.");
  params.addParam<std::vector<std::string>>(
      "activation_function",
      std::vector<std::string>({"relu"}),
      "The type of activation functions to use. It is either one value "
      "or one value per hidden layer.");
  params.addParam<std::string>(
      "nn_filename", "net.pt", "Filename used to output the neural net parameters.");
  params.addParam<bool>("read_from_file",
                        false,
                        "Switch to allow reading old trained neural nets for further training.");
  params.addParam<Real>("learning_rate", 0.001, "Learning rate (relaxation).");
  params.addRangeCheckedParam<unsigned int>(
      "print_epoch_loss",
      0,
      "0<=print_epoch_loss",
      "Epoch training loss printing. 0 - no printing, 1 - every epoch, 10 - every 10th epoch.");
  params.addParam<unsigned int>(
      "seed", 11, "Random number generator seed for stochastic optimizers.");
  params.addParam<unsigned int>(
      "max_processes", 1, "The maximum number of parallel processes that the trainer will use.");
 
  params.addParam<bool>(
      "standardize_input", true, "Standardize (center and scale) training inputs (x values)");
  params.addParam<bool>(
      "standardize_output", true, "Standardize (center and scale) training outputs (y values)");
 
  params.suppressParameter<MooseEnum>("response_type");
  return params;
}

Member Data Documentation

◆ _activation_function

std::vector<std::string>& LibtorchANNTrainer::_activation_function

private

Activation functions for each hidden layer.

Definition at line 55 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _converged

const bool* SurrogateTrainer::_converged

privateinherited

Whether or not the current sample has a converged solution.

Definition at line 170 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::executeTraining(), and SurrogateTrainer::SurrogateTrainer().

◆ _current_sample_size

unsigned int SurrogateTrainer::_current_sample_size

privateinherited

Number of samples used to train the model.

Definition at line 173 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::crossValidate(), SurrogateTrainer::execute(), and SurrogateTrainer::getCurrentSampleSize().

◆ _cv_generator

MooseRandom SurrogateTrainer::_cv_generator

privateinherited

Random number generator used for shuffling sampler rows during splitting.

Definition at line 196 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::crossValidate(), and SurrogateTrainer::SurrogateTrainer().

◆ _cv_n_trials

const unsigned int& SurrogateTrainer::_cv_n_trials

privateinherited

Number of repeated trials of cross validation to perform.

Definition at line 192 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::execute().

◆ _cv_seed

const unsigned int& SurrogateTrainer::_cv_seed

privateinherited

Seed used for _cv_generator.

Definition at line 194 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::SurrogateTrainer().

◆ _cv_surrogate

const SurrogateModel* SurrogateTrainer::_cv_surrogate

privateinherited

SurrogateModel used to evaluate model error relative to test points.

Definition at line 198 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::crossValidate(), and SurrogateTrainer::initialize().

◆ _cv_trial_scores

std::vector<std::vector<Real> >& SurrogateTrainer::_cv_trial_scores

privateinherited

RMSE scores from each CV trial - can be grabbed by VPP or Reporter.

Definition at line 202 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::execute().

◆ _cv_type

const MooseEnum& SurrogateTrainer::_cv_type

privateinherited

Type of cross validation to perform - for now, just 'none' (no CV) or 'k_fold'.

Definition at line 188 of file SurrogateTrainer.h.

◆ _doing_cv

const bool SurrogateTrainer::_doing_cv

privateinherited

Set to true if cross validation is being performed, controls behavior in execute().

Definition at line 200 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::execute(), SurrogateTrainer::initialize(), and SurrogateTrainer::SurrogateTrainer().

◆ _flattened_data

std::vector<Real> LibtorchANNTrainer::_flattened_data

private

The gathered data in a flattened form to be able to convert easily to torch::Tensor.

Definition at line 45 of file LibtorchANNTrainer.h.

Referenced by postTrain(), preTrain(), and train().

◆ _flattened_response

std::vector<Real> LibtorchANNTrainer::_flattened_response

private

The gathered response in a flattened form to be able to convert easily to torch::Tensor.

Definition at line 48 of file LibtorchANNTrainer.h.

Referenced by postTrain(), preTrain(), and train().

◆ _input_standardizer

StochasticTools::Standardizer& LibtorchANNTrainer::_input_standardizer

private

Standardizer for use with input (x)

Definition at line 79 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _local_row

dof_id_type SurrogateTrainer::_local_row

protectedinherited

During training loop, this is the local row index of the data.

Definition at line 123 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::executeTraining().

◆ _local_sample_size

unsigned int SurrogateTrainer::_local_sample_size

privateinherited

Number of samples (locally) used to train the model.

Definition at line 176 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::crossValidate(), SurrogateTrainer::execute(), and SurrogateTrainer::getLocalSampleSize().

◆ _model_meta_data_name

const std::string RestartableModelInterface::_model_meta_data_name

privateinherited

The model meta data name.

This is used to store the restartable data within the RestartableDataMap.

Definition at line 61 of file RestartableModelInterface.h.

Referenced by RestartableModelInterface::modelMetaDataName(), and RestartableModelInterface::RestartableModelInterface().

◆ _model_object

const MooseObject& RestartableModelInterface::_model_object

privateinherited

Reference to the MooseObject that uses this interface.

Definition at line 57 of file RestartableModelInterface.h.

Referenced by RestartableModelInterface::getModelDataFileName(), RestartableModelInterface::hasModelData(), and RestartableModelInterface::RestartableModelInterface().

◆ _model_restartable

PublicRestartable RestartableModelInterface::_model_restartable

privateinherited

Member for interfacing with the framework's restartable system.

We need this because we would like to have the capability to handle the model data separately from the other data members used for checkpointing.

Definition at line 68 of file RestartableModelInterface.h.

Referenced by RestartableModelInterface::declareModelData(), and RestartableModelInterface::getModelData().

◆ _n_dims

unsigned int SurrogateTrainer::_n_dims

protectedinherited

Dimension of predictor data - either _sampler.getNumberOfCols() or _pvals.size() + _pcols.size().

Definition at line 133 of file SurrogateTrainer.h.

Referenced by NearestPointTrainer::NearestPointTrainer(), postTrain(), GaussianProcessTrainer::postTrain(), preTrain(), SurrogateTrainer::SurrogateTrainer(), NearestPointTrainer::train(), and PolynomialRegressionTrainer::train().

◆ _n_outputs

unsigned int& SurrogateTrainer::_n_outputs

protectedinherited

The number of outputs.

Definition at line 135 of file SurrogateTrainer.h.

Referenced by GaussianProcessTrainer::GaussianProcessTrainer(), GaussianProcessTrainer::postTrain(), and GaussianProcessTrainer::train().

◆ _n_splits

const unsigned int& SurrogateTrainer::_n_splits

privateinherited

Number of splits (k) to split sampler data into.

Definition at line 190 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::crossValidate(), and SurrogateTrainer::SurrogateTrainer().

◆ _nn

std::shared_ptr<Moose::LibtorchArtificialNeuralNet>& LibtorchANNTrainer::_nn

private

Pointer to the neural net object (initialized as null)

Definition at line 70 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _nn_filename

const std::string LibtorchANNTrainer::_nn_filename

private

Name of the pytorch output file.

This is used for loading and storing already existing data.

Definition at line 59 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _num_neurons_per_layer

std::vector<unsigned int>& LibtorchANNTrainer::_num_neurons_per_layer

private

Number of neurons within the hidden layers (the length of this vector should be the same as _num_hidden_layers)

Definition at line 52 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _optim_options

Moose::LibtorchTrainingOptions LibtorchANNTrainer::_optim_options

private

The struct which contains the information for the training of the neural net.

Definition at line 67 of file LibtorchANNTrainer.h.

Referenced by LibtorchANNTrainer(), and postTrain().

◆ _output_standardizer

StochasticTools::Standardizer& LibtorchANNTrainer::_output_standardizer

private

Standardizer for use with output response (y)

Definition at line 82 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _pcols

std::vector<unsigned int> SurrogateTrainer::_pcols

protectedinherited

Columns from sampler for predictors.

Definition at line 131 of file SurrogateTrainer.h.

Referenced by PolynomialChaosTrainer::PolynomialChaosTrainer(), SurrogateTrainer::SurrogateTrainer(), and SurrogateTrainer::updatePredictorRow().

◆ _predictor_data

std::vector<Real> SurrogateTrainer::_predictor_data

privateinherited

Predictor data for current row - can be combination of Sampler and Reporter values.

Definition at line 164 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::evaluateModelError(), SurrogateTrainer::getPredictorData(), SurrogateTrainer::SurrogateTrainer(), and SurrogateTrainer::updatePredictorRow().

◆ _predictor_row

const std::vector<Real>& LibtorchANNTrainer::_predictor_row

private

Data from the current predictor row.

Definition at line 42 of file LibtorchANNTrainer.h.

Referenced by train().

◆ _pvals

std::vector<const Real *> SurrogateTrainer::_pvals

protectedinherited

Predictor values from reporters.

Definition at line 129 of file SurrogateTrainer.h.

Referenced by PolynomialChaosTrainer::PolynomialChaosTrainer(), SurrogateTrainer::SurrogateTrainer(), and SurrogateTrainer::updatePredictorRow().

◆ _read_from_file

const bool LibtorchANNTrainer::_read_from_file

private

Switch indicating if an already existing neural net should be read from a file or not.

This can be used to load existing torch files (from previous MOOSE or python runs for retraining and further manipulation)

Definition at line 64 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _row

dof_id_type SurrogateTrainer::_row

protectedinherited

During training loop, this is the row index of the data.

Definition at line 121 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::executeTraining(), and PolynomialChaosTrainer::train().

◆ _row_data

std::vector<Real> SurrogateTrainer::_row_data

privateinherited

Sampler data for the current row.

Definition at line 161 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::checkIntegrity(), SurrogateTrainer::crossValidate(), SurrogateTrainer::executeTraining(), SurrogateTrainer::getSamplerData(), and SurrogateTrainer::updatePredictorRow().

◆ _rval

const Real* SurrogateTrainer::_rval

protectedinherited

Response value.

Definition at line 125 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::evaluateModelError(), SurrogateTrainer::SurrogateTrainer(), train(), GaussianProcessTrainer::train(), NearestPointTrainer::train(), and PolynomialRegressionTrainer::train().

◆ _rvecval

const std::vector<Real>* SurrogateTrainer::_rvecval

protectedinherited

Vector response value.

Definition at line 127 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::evaluateModelError(), SurrogateTrainer::SurrogateTrainer(), GaussianProcessTrainer::train(), NearestPointTrainer::train(), and PolynomialRegressionTrainer::train().

◆ _sampler

Sampler& SurrogateTrainer::_sampler

protectedinherited

Sampler being used for training

Definition at line 119 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::checkIntegrity(), SurrogateTrainer::crossValidate(), SurrogateTrainer::execute(), SurrogateTrainer::executeTraining(), GaussianProcessTrainer::GaussianProcessTrainer(), PolynomialChaosTrainer::PolynomialChaosTrainer(), PolynomialRegressionTrainer::PolynomialRegressionTrainer(), and SurrogateTrainer::SurrogateTrainer().

◆ _skip_indices

std::vector<dof_id_type> SurrogateTrainer::_skip_indices

privateinherited

Vector of indices to skip during executeTraining()

Definition at line 186 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::crossValidate(), and SurrogateTrainer::executeTraining().

◆ _skip_unconverged

const bool SurrogateTrainer::_skip_unconverged

privateinherited

Whether or not we are skipping samples that have unconverged solutions.

Definition at line 167 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::executeTraining(), and SurrogateTrainer::SurrogateTrainer().

◆ _smi_feproblem

FEProblemBase& SurrogateModelInterface::_smi_feproblem

privateinherited

Reference to FEProblemBase instance.

Definition at line 73 of file SurrogateModelInterface.h.

Referenced by SurrogateModelInterface::getSurrogateModelByName(), and SurrogateModelInterface::getSurrogateTrainerByName().

◆ _smi_params

const InputParameters& SurrogateModelInterface::_smi_params

privateinherited

Parameters of the object with this interface.

Definition at line 70 of file SurrogateModelInterface.h.

Referenced by SurrogateModelInterface::getSurrogateModel(), and SurrogateModelInterface::getSurrogateTrainer().

◆ _smi_tid

const THREAD_ID SurrogateModelInterface::_smi_tid

privateinherited

Thread ID.

Definition at line 76 of file SurrogateModelInterface.h.

Referenced by SurrogateModelInterface::getSurrogateTrainerByName().

◆ _standardize_input

const bool LibtorchANNTrainer::_standardize_input

private

If the training output should be standardized (scaled and shifted)

Definition at line 73 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _standardize_output

const bool LibtorchANNTrainer::_standardize_output

private

If the training output should be standardized (scaled and shifted)

Definition at line 76 of file LibtorchANNTrainer.h.

Referenced by postTrain().

◆ _training_data

std::unordered_map<ReporterName, std::shared_ptr<TrainingDataBase> > SurrogateTrainer::_training_data

privateinherited

Vector of reporter names and their corresponding values (to be filled by getTrainingData)

Definition at line 179 of file SurrogateTrainer.h.

Referenced by SurrogateTrainer::checkIntegrity(), SurrogateTrainer::crossValidate(), SurrogateTrainer::executeTraining(), SurrogateTrainer::getTrainingData(), and SurrogateTrainer::initialize().

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

stochastic_tools/include/libtorch/surrogates/LibtorchANNTrainer.h
stochastic_tools/src/libtorch/trainers/LibtorchANNTrainer.C

Public Types

Public Member Functions

Static Public Member Functions

Public Attributes

Static Public Attributes

Protected Member Functions

Static Protected Member Functions

Protected Attributes

Static Protected Attributes

Private Member Functions

Static Private Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ LibtorchANNTrainer()

Member Function Documentation

◆ checkIntegrity()

◆ crossValidate()

◆ declareModelData()

◆ evaluateModelError()

◆ execute()

◆ executeTraining()

◆ finalize()

◆ getCurrentSampleSize()

◆ getLocalSampleSize()

◆ getModelData()

◆ getModelDataFileName()

◆ getPredictorData()

◆ getSamplerData()

◆ getSurrogateModel() [1/2]

◆ getSurrogateModel() [2/2]

◆ getSurrogateModelByName() [1/2]

◆ getSurrogateModelByName() [2/2]

◆ getSurrogateTrainer() [1/2]

◆ getSurrogateTrainer() [2/2]

◆ getSurrogateTrainerByName() [1/2]

◆ getSurrogateTrainerByName() [2/2]

◆ getTrainingData()

◆ hasModelData()

◆ initialize()

◆ modelMetaDataName()

◆ postTrain()

◆ preTrain()

◆ threadJoin()

◆ train()

◆ updatePredictorRow()

◆ validParams()

Member Data Documentation

◆ _activation_function

◆ _converged

◆ _current_sample_size

◆ _cv_generator

◆ _cv_n_trials

◆ _cv_seed

◆ _cv_surrogate

◆ _cv_trial_scores

◆ _cv_type

◆ _doing_cv

◆ _flattened_data

◆ _flattened_response

◆ _input_standardizer

◆ _local_row

◆ _local_sample_size

◆ _model_meta_data_name

◆ _model_object

◆ _model_restartable

◆ _n_dims

◆ _n_outputs

◆ _n_splits

◆ _nn

◆ _nn_filename

◆ _num_neurons_per_layer

◆ _optim_options

◆ _output_standardizer

◆ _pcols

◆ _predictor_data

◆ _predictor_row

◆ _pvals

◆ _read_from_file

◆ _row