Namespaces
namespace	Internal

namespace	SOFIE

Classes
class	Classification

class	ClassificationResult

class	RBDT

class	RReader
	A replacement for the TMVA::Reader legacy interface. More...

class	RSofieReader
	TMVA::RSofieReader class for reading external Machine Learning models in ONNX files, Keras .h5 files or PyTorch .pt files and performing the inference using SOFIE It is reccomended to use ONNX if possible since there is a larger support for model operators. More...

class	RStandardScaler

class	RTensor
	RTensor is a container with contiguous memory and shape information. More...

class	SofieFunctorHelper
	Helper class used by SOFIEFunctor to wrap the infer signature interface to RDataFrame. More...

class	SofieFunctorHelper< std::index_sequence< N... >, Session_t, T >

Enumerations
enum class	MemoryLayout : uint8_t { RowMajor = 0x01 , ColumnMajor = 0x02 , RowMajor = 0x01 , ColumnMajor = 0x02 }
	Memory layout type (copy from RTensor.hxx) More...

enum class	MemoryLayout : uint8_t { RowMajor = 0x01 , ColumnMajor = 0x02 , RowMajor = 0x01 , ColumnMajor = 0x02 }
	Memory layout type. More...

Functions
template<typename T , typename U >
RTensor< T >	AsTensor (U &dataframe, std::vector< std::string > columns={}, MemoryLayout layout=MemoryLayout::RowMajor)
	Convert the content of an RDataFrame to an RTensor.

template<std::size_t N, typename T , typename F >
auto	Compute (F &&f) -> Internal::ComputeHelper< std::make_index_sequence< N >, T, F >
	Helper to pass TMVA model to RDataFrame.Define nodes.

template<typename T >
std::ostream &	operator<< (std::ostream &os, RTensor< T > &x)
	Pretty printing.

template<std::size_t N, typename Session_t >
auto	SofieFunctor (unsigned int nslots=0, const std::string &weightsFile="") -> SofieFunctorHelper< std::make_index_sequence< N >, Session_t, float >
	SofieFunctor : used to wrap the infer function of the generated model by SOFIE in a RDF compatible signature.

Enumeration Type Documentation

◆ MemoryLayout [1/2]

enum class TMVA::Experimental::MemoryLayout : uint8_t

strong

Memory layout type (copy from RTensor.hxx)

Enumerator
RowMajor
ColumnMajor
RowMajor
ColumnMajor

Definition at line 47 of file CudaTensor.h.

◆ MemoryLayout [2/2]

enum class TMVA::Experimental::MemoryLayout : uint8_t

strong

Memory layout type.

Enumerator
RowMajor
ColumnMajor
RowMajor
ColumnMajor

Definition at line 17 of file RTensor.hxx.

Function Documentation

◆ AsTensor()

template<typename T , typename U >

RTensor< T > TMVA::Experimental::AsTensor	(	U &	dataframe,
		std::vector< std::string >	columns = `{}`,
		MemoryLayout	layout = `MemoryLayout::RowMajor`
	)

Convert the content of an RDataFrame to an RTensor.

Parameters

[in]	dataframe	RDataFrame node
[in]	columns	Vector of column names
[in]	layout	Memory layout

Returns: RTensor with content from selected columns

Definition at line 21 of file RTensorUtils.hxx.

◆ Compute()

template<std::size_t N, typename T , typename F >

auto TMVA::Experimental::Compute ( F && f ) -> Internal::ComputeHelper<std::make_index_sequence<N>, T, F>

Helper to pass TMVA model to RDataFrame.Define nodes.

Definition at line 32 of file RInferenceUtils.hxx.

◆ operator<<()

template<typename T >

std::ostream & TMVA::Experimental::operator<<	(	std::ostream &	os,
		RTensor< T > &	x
	)

Pretty printing.

Parameters

[in]	os	Output stream
[in]	x	RTensor

Returns: Modified output stream

Definition at line 582 of file RTensor.hxx.

◆ SofieFunctor()

template<std::size_t N, typename Session_t >

auto TMVA::Experimental::SofieFunctor	(	unsigned int	nslots = `0`,
		const std::string &	weightsFile = `""`
	)		-> SofieFunctorHelper<std::make_index_sequence<N>, Session_t, float>

SofieFunctor : used to wrap the infer function of the generated model by SOFIE in a RDF compatible signature.

The number of slots is an optional parameter used to create multiple SOFIE Sessions, which can be run in a parallel model evaluation. One shouild use as number of slots the number of slots used by RDataFrame. By default, in case of nslots=0, only a single Session will be created and the Functor cannot be run in parallel. Examples of using the SofieFunctor are the C++ tutorial TMVA_SOFIE_RDataFrame.C and the Python tutorial TMVA_SOFIE_RDataFrame.py which makes use of the ROOT JIT to compile on the fly the generated SOFIE model.

Definition at line 64 of file SOFIEHelpers.hxx.

Namespaces

Classes

Enumerations

Functions