Internal data structured used by XGBoost during training. More...

#include <data.h>

Collaboration diagram for xgboost::DMatrix:

Public Member Functions
	DMatrix ()=default
	default constructor More...

virtual MetaInfo &	Info ()=0
	meta information of the dataset More...

virtual void	SetInfo (const char key, const void dptr, DataType dtype, size_t num)

virtual void	SetInfo (const char *key, std::string const &interface_str)

virtual const MetaInfo &	Info () const =0
	meta information of the dataset More...

XGBAPIThreadLocalEntry &	GetThreadLocal () const
	Get thread local memory for returning data from DMatrix. More...

template<typename T >
BatchSet< T >	GetBatches (const BatchParam &param={})
	Gets batches. Use range based for loop over BatchSet to access individual batches. More...

template<typename T >
bool	PageExists () const

virtual bool	SingleColBlock () const =0

virtual	~DMatrix ()
	virtual destructor More...

bool	IsDense () const
	Whether the matrix is dense. More...

virtual DMatrix *	Slice (common::Span< int32_t const > ridxs)=0

template<>
BatchSet< SparsePage >	GetBatches (const BatchParam &)

template<>
bool	PageExists () const

template<>
bool	PageExists () const

Static Public Member Functions
static DMatrix *	Load (const std::string &uri, bool silent, bool load_row_split, const std::string &file_format="auto", size_t page_size=kPageSize)
	Load DMatrix from URI. More...

template<typename AdapterT >
static DMatrix *	Create (AdapterT *adapter, float missing, int nthread, const std::string &cache_prefix="", size_t page_size=kPageSize)
	Creates a new DMatrix from an external data adapter. More...

template<typename DataIterHandle , typename DMatrixHandle , typename DataIterResetCallback , typename XGDMatrixCallbackNext >
static DMatrix *	Create (DataIterHandle iter, DMatrixHandle proxy, DataIterResetCallback reset, XGDMatrixCallbackNext next, float missing, int nthread, int max_bin)
	Create a new Quantile based DMatrix used for histogram based algorithm. More...

Static Public Attributes
static const size_t	kPageSize = 32UL << 12UL
	Number of rows per page in external memory. Approximately 100MB per page for dataset with 100 features. More...

Protected Member Functions
virtual BatchSet< SparsePage >	GetRowBatches ()=0

virtual BatchSet< CSCPage >	GetColumnBatches ()=0

virtual BatchSet< SortedCSCPage >	GetSortedColumnBatches ()=0

virtual BatchSet< EllpackPage >	GetEllpackBatches (const BatchParam &param)=0

virtual bool	EllpackExists () const =0

virtual bool	SparsePageExists () const =0

Detailed Description

Internal data structured used by XGBoost during training.

Constructor & Destructor Documentation

◆ DMatrix()

xgboost::DMatrix::DMatrix ( )

default

default constructor

◆ ~DMatrix()

virtual xgboost::DMatrix::~DMatrix ( )

virtual

virtual destructor

Member Function Documentation

◆ Create() [1/2]

template<typename AdapterT >

static DMatrix* xgboost::DMatrix::Create	(	AdapterT *	adapter,
		float	missing,
		int	nthread,
		const std::string &	cache_prefix = `""`,
		size_t	page_size = `kPageSize`
	)

static

Creates a new DMatrix from an external data adapter.

Template Parameters

AdapterT Type of the adapter.

Parameters

[in,out]	adapter	View onto an external data.
	missing	Values to count as missing.
	nthread	Number of threads for construction.
	cache_prefix	(Optional) The cache prefix for external memory.
	page_size	(Optional) Size of the page.

Returns: a Created DMatrix.

◆ Create() [2/2]

template<typename DataIterHandle , typename DMatrixHandle , typename DataIterResetCallback , typename XGDMatrixCallbackNext >

static DMatrix* xgboost::DMatrix::Create	(	DataIterHandle	iter,
		DMatrixHandle	proxy,
		DataIterResetCallback *	reset,
		XGDMatrixCallbackNext *	next,
		float	missing,
		int	nthread,
		int	max_bin
	)

static

Create a new Quantile based DMatrix used for histogram based algorithm.

Template Parameters

DataIterHandle	External iterator type, defined in C API.
DMatrixHandle	DMatrix handle, defined in C API.
DataIterResetCallback	Callback for reset, prototype defined in C API.
XGDMatrixCallbackNext	Callback for next, prototype defined in C API.

Parameters

iter	External data iterator
proxy	A hanlde to ProxyDMatrix
reset	Callback for reset
next	Callback for next
missing	Value that should be treated as missing.
nthread	number of threads used for initialization.
max_bin	Maximum number of bins.

Returns: A created quantile based DMatrix.

◆ EllpackExists()

virtual bool xgboost::DMatrix::EllpackExists ( ) const

protectedpure virtual

◆ GetBatches() [1/2]

template<>

BatchSet< EllpackPage > xgboost::DMatrix::GetBatches ( const BatchParam & param )

inline

◆ GetBatches() [2/2]

template<typename T >

BatchSet<T> xgboost::DMatrix::GetBatches ( const BatchParam & param = {} )

Gets batches. Use range based for loop over BatchSet to access individual batches.

◆ GetColumnBatches()

virtual BatchSet<CSCPage> xgboost::DMatrix::GetColumnBatches ( )

protectedpure virtual

◆ GetEllpackBatches()

virtual BatchSet<EllpackPage> xgboost::DMatrix::GetEllpackBatches ( const BatchParam & param )

protectedpure virtual

◆ GetRowBatches()

virtual BatchSet<SparsePage> xgboost::DMatrix::GetRowBatches ( )

protectedpure virtual

◆ GetSortedColumnBatches()

virtual BatchSet<SortedCSCPage> xgboost::DMatrix::GetSortedColumnBatches ( )

protectedpure virtual

◆ GetThreadLocal()

XGBAPIThreadLocalEntry& xgboost::DMatrix::GetThreadLocal ( ) const

Get thread local memory for returning data from DMatrix.

◆ Info() [1/2]

virtual const MetaInfo& xgboost::DMatrix::Info ( ) const

pure virtual

meta information of the dataset

◆ Info() [2/2]

virtual MetaInfo& xgboost::DMatrix::Info ( )

pure virtual

meta information of the dataset

◆ IsDense()

bool xgboost::DMatrix::IsDense ( ) const

inline

Whether the matrix is dense.

◆ Load()

static DMatrix* xgboost::DMatrix::Load	(	const std::string &	uri,
		bool	silent,
		bool	load_row_split,
		const std::string &	file_format = `"auto"`,
		size_t	page_size = `kPageSize`
	)

static

Load DMatrix from URI.

Parameters

uri	The URI of input.
silent	Whether print information during loading.
load_row_split	Flag to read in part of rows, divided among the workers in distributed mode.
file_format	The format type of the file, used for dmlc::Parser::Create. By default "auto" will be able to load in both local binary file.
page_size	Page size for external memory.

Returns: The created DMatrix.

◆ PageExists() [1/3]

template<typename T >

bool xgboost::DMatrix::PageExists ( ) const

◆ PageExists() [2/3]

template<>

bool xgboost::DMatrix::PageExists ( ) const

inline

◆ PageExists() [3/3]

template<>

bool xgboost::DMatrix::PageExists ( ) const

inline

◆ SetInfo() [1/2]

virtual void xgboost::DMatrix::SetInfo	(	const char *	key,
		const void *	dptr,
		DataType	dtype,
		size_t	num
	)

inlinevirtual

◆ SetInfo() [2/2]

virtual void xgboost::DMatrix::SetInfo	(	const char *	key,
		std::string const &	interface_str
	)

inlinevirtual

◆ SingleColBlock()

virtual bool xgboost::DMatrix::SingleColBlock ( ) const

pure virtual

Returns: Whether the data columns single column block.

◆ Slice()

virtual DMatrix* xgboost::DMatrix::Slice ( common::Span< int32_t const > ridxs )

pure virtual

◆ SparsePageExists()

virtual bool xgboost::DMatrix::SparsePageExists ( ) const

protectedpure virtual

Member Data Documentation

◆ kPageSize

const size_t xgboost::DMatrix::kPageSize = 32UL << 12UL

static

Number of rows per page in external memory. Approximately 100MB per page for dataset with 100 features.

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

/home/runner/work/xgboost/xgboost/include/xgboost/data.h

Public Member Functions

Static Public Member Functions

Static Public Attributes

Protected Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ DMatrix()

◆ ~DMatrix()

Member Function Documentation

◆ Create() [1/2]

◆ Create() [2/2]

◆ EllpackExists()

◆ GetBatches() [1/2]

◆ GetBatches() [2/2]

◆ GetColumnBatches()

◆ GetEllpackBatches()

◆ GetRowBatches()

◆ GetSortedColumnBatches()

◆ GetThreadLocal()

◆ Info() [1/2]

◆ Info() [2/2]

◆ IsDense()

◆ Load()

◆ PageExists() [1/3]

◆ PageExists() [2/3]

◆ PageExists() [3/3]

◆ SetInfo() [1/2]

◆ SetInfo() [2/2]

◆ SingleColBlock()

◆ Slice()

◆ SparsePageExists()

Member Data Documentation

◆ kPageSize