xgboost::SparsePage Class Reference

In-memory storage unit of sparse batch, stored in CSR format. More...

#include <data.h>

Inheritance diagram for xgboost::SparsePage:

Public Types
using	Inst = common::Span< Entry const >
	an instance of sparse vector in the batch

Public Member Functions
HostSparsePageView	GetView () const
	SparsePage ()
	constructor
	SparsePage (SparsePage const &that)=delete
	SparsePage (SparsePage &&that)=default
SparsePage &	operator= (SparsePage const &that)=delete
SparsePage &	operator= (SparsePage &&that)=default
size_t	Size () const
size_t	MemCostBytes () const
void	Clear ()
	clear the page
void	SetBaseRowId (size_t row_id)
	Set the base row id for this page.
SparsePage	GetTranspose (int num_columns, int32_t n_threads) const
void	SortIndices (int32_t n_threads)
	Sort the column index.
bool	IsIndicesSorted (int32_t n_threads) const
	Check wether the column index is sorted.
void	Reindex (uint64_t feature_offset, int32_t n_threads)
	Reindex the column index with an offset.
void	SortRows (int32_t n_threads)
template<typename AdapterBatchT >
uint64_t	Push (const AdapterBatchT &batch, float missing, int nthread)
	Pushes external data batch onto this page.
void	Push (const SparsePage &batch)
	Push a sparse page.
void	PushCSC (const SparsePage &batch)
	Push a SparsePage stored in CSC format.

Data Fields
HostDeviceVector< bst_row_t >	offset
HostDeviceVector< Entry >	data
	the data of the segments
size_t	base_rowid {0}

Detailed Description

In-memory storage unit of sparse batch, stored in CSR format.

Member Function Documentation

MemCostBytes()

size_t xgboost::SparsePage::MemCostBytes ( ) const

inline

Returns

estimation of memory cost of this page

Push() [1/2]

template<typename AdapterBatchT >

uint64_t xgboost::SparsePage::Push	(	const AdapterBatchT &	batch,
		float	missing,
		int	nthread
	)

Pushes external data batch onto this page.

Template Parameters

AdapterBatchT

Parameters

batch
missing
nthread

Returns

The maximum number of columns encountered in this input batch. Useful when pushing many adapter batches to work out the total number of columns.

Push() [2/2]

void xgboost::SparsePage::Push

(

const SparsePage &

batch

)

Push a sparse page.

Parameters

batch

the row page

PushCSC()

void xgboost::SparsePage::PushCSC

(

const SparsePage &

batch

)

Push a SparsePage stored in CSC format.

Parameters

batch

The row batch to be pushed

Size()

size_t xgboost::SparsePage::Size ( ) const

inline

Returns

Number of instances in the page.

---

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

• External/xgboost/include/xgboost/data.h
• External/xgboost/src/data/data.cc