ActsPlugins::GnnPipeline Class Reference

Graph Neural Network pipeline for track finding. More...

#include <ActsPlugins/Gnn/GnnPipeline.hpp>

Public Member Functions
	GnnPipeline (std::shared_ptr< GraphConstructionBase > graphConstructor, std::vector< std::shared_ptr< EdgeClassificationBase > > edgeClassifiers, std::shared_ptr< TrackBuildingBase > trackBuilder, std::unique_ptr< const Acts::Logger > logger, bool shrinkNodes=false)
	Construct GNN pipeline.
std::vector< std::vector< int > >	run (std::vector< float > &features, const std::vector< std::uint64_t > &moduleIds, std::vector< int > &spacePointIDs, Device device, const GnnHook &hook={}, GnnTiming *timing=nullptr) const
	Run the GNN pipeline.

Detailed Description

Graph Neural Network pipeline for track finding.

Constructor & Destructor Documentation

GnnPipeline()

ActsPlugins::GnnPipeline::GnnPipeline	(	std::shared_ptr< GraphConstructionBase >	graphConstructor,
		std::vector< std::shared_ptr< EdgeClassificationBase > >	edgeClassifiers,
		std::shared_ptr< TrackBuildingBase >	trackBuilder,
		std::unique_ptr< const Acts::Logger >	logger,
		bool	shrinkNodes = false )

Construct GNN pipeline.

Parameters

graphConstructor	Graph construction stage
edgeClassifiers	Edge classification stages
trackBuilder	Track building stage
logger	Logger instance
shrinkNodes	If true, unused nodes are removed from the graph before each edge classification stage

Member Function Documentation

run()

std::vector< std::vector< int > > ActsPlugins::GnnPipeline::run	(	std::vector< float > &	features,
		const std::vector< std::uint64_t > &	moduleIds,
		std::vector< int > &	spacePointIDs,
		Device	device,
		const GnnHook &	hook = {},
		GnnTiming *	timing = nullptr ) const

Run the GNN pipeline.

Parameters

features	Input feature vector
moduleIds	Module identifiers for each space point
spacePointIDs	Space point identifiers
device	Device to run on (CPU/GPU)
hook	Optional callback hook for pipeline execution
timing	Optional timing output

Returns

Vector of track candidates (each track is a vector of space point IDs)