ROOT::Experimental::Internal::ML::RClusterLoader Class Reference

Loads TTree/RNTuple clusters from one or more RDataFrames into RFlat2DMatrix buffers for ML training and validation.

Overview

At construction the loader scans the cluster boundaries of every provided RDataFrame and stores them as a flat list of RClusterRange objects. SplitDataset() then partitions those ranges into training and validation sets according to validationSplit.

The split strategy depends on whether shuffling is enabled or not

• Unshuffled: one cut is made so that the first (1 - validationSplit) fraction of entries goes to training. At most one cluster is split at the boundary.
• Shuffled: each cluster is split proportionally (according to validationSplit) so both sets draw entries from every part of the dataset. ShuffleTrainingClusters() and ShuffleValidationClusters() re-order the cluster lists at the start of each epoch. A second shuffling step, at the entries level, happens inside LoadTrainingClusterInto() and LoadValidationClusterInto() when loading the data into the tensors.

Filtered RDataFrames

When any RDataFrame carries a filter, the true entry count is not known until the computation graph is executed. In this case SplitDataset() is a no-op and the split is discovered lazily inside LoadTrainingClusterInto() during the first epoch. After the first epoch FinaliseSplitDiscovery() marks the split as stable and all subsequent epochs use the same pre-computed ranges.

Definition at line 80 of file RClusterLoader.hxx.

Public Member Functions
	RClusterLoader (std::vector< ROOT::RDF::RNode > &rdfs, const std::vector< std::string > &cols, const std::vector< std::size_t > &vecSizes, float vecPadding, float validationSplit, bool shuffle, std::size_t setSeed)
void	FinaliseSplitDiscovery ()
	Mark the train/val split as finalised after the first epoch.
std::size_t	GetNmTotalClusters () const
std::size_t	GetNumChunkCols () const
std::size_t	GetNumTrainingClusters () const
std::size_t	GetNumTrainingEntries () const
std::size_t	GetNumValidationClusters () const
std::size_t	GetNumValidationEntries () const
const std::vector< RClusterRange > &	GetTrainingClusters () const
const std::vector< RClusterRange > &	GetValidationClusters () const
bool	IsSplitDiscovered () const
void	LoadClusterInto (RFlat2DMatrix &dest, std::size_t rdfIdx, std::uint64_t startRow, std::uint64_t endRow)
std::size_t	LoadTrainingClusterInto (RFlat2DMatrix &dest, std::size_t rdfIdx, std::uint64_t startRow, std::uint64_t endRow)
	Load one training cluster and return the number of rows written.
void	LoadValidationClusterInto (RFlat2DMatrix &dest, std::size_t rdfIdx, std::uint64_t startRow, std::uint64_t endRow)
	Load one validation cluster at the end of dest.
void	ShuffleTrainingClusters (std::size_t epochIdx)
	Re-order training clusters for the upcoming epoch.
void	ShuffleValidationClusters (std::size_t epochIdx)
	Re-order validation clusters for the upcoming epoch.
void	SplitDataset ()
	Distribute the clusters into training and validation datasets No-op for filtered RDataFrames, the split is discovered lazily during the first epoch.

Private Attributes
std::size_t	fAccumulatedFilteredForTrain {0}
std::size_t	fAccumulatedFilteredForVal {0}
std::vector< RClusterRange >	fAllClusters
std::vector< std::string >	fCols
bool	fIsFiltered {false}
std::size_t	fNumChunkCols
std::size_t	fNumCols
std::size_t	fNumTrainingEntries {0}
std::size_t	fNumValidationEntries {0}
std::vector< ROOT::RDF::RNode > &	fRdfs
std::vector< std::size_t >	fRdfSizes
std::size_t	fSetSeed
bool	fShuffle
bool	fSplitDiscovered {false}
std::size_t	fSumVecSizes
std::size_t	fTotalEntries {0}
std::vector< RClusterRange >	fTrainingClusters
std::vector< RClusterRange >	fValidationClusters
float	fValidationSplit
float	fVecPadding
std::vector< std::size_t >	fVecSizes

#include </github/home/ROOT-CI/src/tree/ml/src/RClusterLoader.hxx>

Constructor & Destructor Documentation

RClusterLoader()

ROOT::Experimental::Internal::ML::RClusterLoader::RClusterLoader	(	std::vector< ROOT::RDF::RNode > &	rdfs,
		const std::vector< std::string > &	cols,
		const std::vector< std::size_t > &	vecSizes,
		float	vecPadding,
		float	validationSplit,
		bool	shuffle,
		std::size_t	setSeed )

Definition at line 5 of file RClusterLoader.cxx.

Member Function Documentation

FinaliseSplitDiscovery()

void ROOT::Experimental::Internal::ML::RClusterLoader::FinaliseSplitDiscovery

(

)

Mark the train/val split as finalised after the first epoch.

Definition at line 285 of file RClusterLoader.cxx.

GetNmTotalClusters()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::GetNmTotalClusters ( ) const

inline

Definition at line 148 of file RClusterLoader.hxx.

GetNumChunkCols()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::GetNumChunkCols ( ) const

inline

Definition at line 135 of file RClusterLoader.hxx.

GetNumTrainingClusters()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::GetNumTrainingClusters ( ) const

inline

Definition at line 143 of file RClusterLoader.hxx.

GetNumTrainingEntries()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::GetNumTrainingEntries ( ) const

inline

Definition at line 133 of file RClusterLoader.hxx.

GetNumValidationClusters()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::GetNumValidationClusters ( ) const

inline

Definition at line 147 of file RClusterLoader.hxx.

GetNumValidationEntries()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::GetNumValidationEntries ( ) const

inline

Definition at line 134 of file RClusterLoader.hxx.

GetTrainingClusters()

const std::vector< RClusterRange > & ROOT::Experimental::Internal::ML::RClusterLoader::GetTrainingClusters ( ) const

inline

Definition at line 137 of file RClusterLoader.hxx.

GetValidationClusters()

const std::vector< RClusterRange > & ROOT::Experimental::Internal::ML::RClusterLoader::GetValidationClusters ( ) const

inline

Definition at line 141 of file RClusterLoader.hxx.

IsSplitDiscovered()

bool ROOT::Experimental::Internal::ML::RClusterLoader::IsSplitDiscovered ( ) const

inline

Definition at line 129 of file RClusterLoader.hxx.

LoadClusterInto()

void ROOT::Experimental::Internal::ML::RClusterLoader::LoadClusterInto	(	RFlat2DMatrix &	dest,
		std::size_t	rdfIdx,
		std::uint64_t	startRow,
		std::uint64_t	endRow )

Definition at line 162 of file RClusterLoader.cxx.

LoadTrainingClusterInto()

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::LoadTrainingClusterInto	(	RFlat2DMatrix &	dest,
		std::size_t	rdfIdx,
		std::uint64_t	startRow,
		std::uint64_t	endRow )

Load one training cluster and return the number of rows written.

Unfiltered: delegates directly to LoadClusterInto() Filtered, epoch 1 (!fSplitDiscovered):

• On the first call, Count() is called across all RDFs to obtain the total filtered entry count, fNumTrainingEntries and fNumValidationEntries are set as targets.
• A single Foreach on the full raw cluster range loads data and captures rdfentry_ simultaneously. The real train/val boundary is computed from the accumulated filtered count vs the target, then the train sub-range is pushed to fTrainingClusters and the val sub-range to fValidationClusters.
• Only the train rows are written into dest. -All subsequent epochs: delegates directly to LoadClusterInto()

Definition at line 186 of file RClusterLoader.cxx.

LoadValidationClusterInto()

void ROOT::Experimental::Internal::ML::RClusterLoader::LoadValidationClusterInto	(	RFlat2DMatrix &	dest,
		std::size_t	rdfIdx,
		std::uint64_t	startRow,
		std::uint64_t	endRow )

Load one validation cluster at the end of dest.

Definition at line 277 of file RClusterLoader.cxx.

ShuffleTrainingClusters()

void ROOT::Experimental::Internal::ML::RClusterLoader::ShuffleTrainingClusters

(

std::size_t

epochIdx

)

Re-order training clusters for the upcoming epoch.

Definition at line 141 of file RClusterLoader.cxx.

ShuffleValidationClusters()

void ROOT::Experimental::Internal::ML::RClusterLoader::ShuffleValidationClusters

(

std::size_t

epochIdx

)

Re-order validation clusters for the upcoming epoch.

Definition at line 153 of file RClusterLoader.cxx.

SplitDataset()

void ROOT::Experimental::Internal::ML::RClusterLoader::SplitDataset

(

)

Distribute the clusters into training and validation datasets No-op for filtered RDataFrames, the split is discovered lazily during the first epoch.

Definition at line 45 of file RClusterLoader.cxx.

Member Data Documentation

fAccumulatedFilteredForTrain

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fAccumulatedFilteredForTrain {0}

private

Definition at line 105 of file RClusterLoader.hxx.

fAccumulatedFilteredForVal

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fAccumulatedFilteredForVal {0}

private

Definition at line 106 of file RClusterLoader.hxx.

fAllClusters

std::vector<RClusterRange> ROOT::Experimental::Internal::ML::RClusterLoader::fAllClusters

private

Definition at line 95 of file RClusterLoader.hxx.

fCols

std::vector<std::string> ROOT::Experimental::Internal::ML::RClusterLoader::fCols

private

Definition at line 84 of file RClusterLoader.hxx.

fIsFiltered

bool ROOT::Experimental::Internal::ML::RClusterLoader::fIsFiltered {false}

private

Definition at line 103 of file RClusterLoader.hxx.

fNumChunkCols

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fNumChunkCols

private

Definition at line 93 of file RClusterLoader.hxx.

fNumCols

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fNumCols

private

Definition at line 91 of file RClusterLoader.hxx.

fNumTrainingEntries

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fNumTrainingEntries {0}

private

Definition at line 100 of file RClusterLoader.hxx.

fNumValidationEntries

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fNumValidationEntries {0}

private

Definition at line 101 of file RClusterLoader.hxx.

fRdfs

std::vector<ROOT::RDF::RNode>& ROOT::Experimental::Internal::ML::RClusterLoader::fRdfs

private

Definition at line 82 of file RClusterLoader.hxx.

fRdfSizes

std::vector<std::size_t> ROOT::Experimental::Internal::ML::RClusterLoader::fRdfSizes

private

Definition at line 83 of file RClusterLoader.hxx.

fSetSeed

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fSetSeed

private

Definition at line 89 of file RClusterLoader.hxx.

fShuffle

bool ROOT::Experimental::Internal::ML::RClusterLoader::fShuffle

private

Definition at line 88 of file RClusterLoader.hxx.

fSplitDiscovered

bool ROOT::Experimental::Internal::ML::RClusterLoader::fSplitDiscovered {false}

private

Definition at line 104 of file RClusterLoader.hxx.

fSumVecSizes

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fSumVecSizes

private

Definition at line 92 of file RClusterLoader.hxx.

fTotalEntries

std::size_t ROOT::Experimental::Internal::ML::RClusterLoader::fTotalEntries {0}

private

Definition at line 99 of file RClusterLoader.hxx.

fTrainingClusters

std::vector<RClusterRange> ROOT::Experimental::Internal::ML::RClusterLoader::fTrainingClusters

private

Definition at line 96 of file RClusterLoader.hxx.

fValidationClusters

std::vector<RClusterRange> ROOT::Experimental::Internal::ML::RClusterLoader::fValidationClusters

private

Definition at line 97 of file RClusterLoader.hxx.

fValidationSplit

float ROOT::Experimental::Internal::ML::RClusterLoader::fValidationSplit

private

Definition at line 87 of file RClusterLoader.hxx.

fVecPadding

float ROOT::Experimental::Internal::ML::RClusterLoader::fVecPadding

private

Definition at line 86 of file RClusterLoader.hxx.

fVecSizes

std::vector<std::size_t> ROOT::Experimental::Internal::ML::RClusterLoader::fVecSizes

private

Definition at line 85 of file RClusterLoader.hxx.