The neural network is constructed from user configurations in NetProto. More...

#include <neuralnet.h>

Public Member Functions
	NeuralNet (NetProto netproto, int npartitions=1)
	construct the net structure from protocol buffer. More...

std::string	ToAdjacency ()
	To display the adjacency layers.

void	ShareParamsFrom (shared_ptr< NeuralNet > other)
	Share memory of parameter values from other neuralnet.

const std::vector< Layer * > &	layers ()

const std::vector< ParserLayer * > &	parserlayers () const

const std::vector< LossLayer * > &	losslayers () const

const std::vector< DataLayer * > &	datalayers () const

const std::vector< Param * > &	params () const

Layer *	name2layer (string name) const

Param *	paramid2param (int id) const

Static Public Member Functions
static shared_ptr< NeuralNet >	Create (const NetProto &np, Phase phase, int num)
	Create the neural network for training, test or validation. More...

Protected Member Functions
Graph *	CreateGraph (const NetProto &netproto, int npartitions)
	Create a neural net graph, one node for each layer. More...

void	CreateNetFromGraph (Graph *graph, int npartitions)
	Create neural net from graph, one layer per node.

void	PrepareDataStructures ()
	prepare data structures, e.g., params_, layers_, etc.

Protected Attributes
vector< Layer * >	layers_

vector< ParserLayer * >	parserlayers_

vector< LossLayer * >	losslayers_

vector< DataLayer * >	datalayers_

vector< Param * >	params_

map< string, Layer * >	name2layer_

map< int, Param * >	paramid2param_

Detailed Description

The neural network is constructed from user configurations in NetProto.

Some layers, e.g., SplitLayer and BridgeSrcLayer/BridgeDstLayer will be added implicitly to partition the neural network. TODO create wrappers for popular models, e.g., MLP, CNN.

Constructor & Destructor Documentation

singa::NeuralNet::NeuralNet	(	NetProto	netproto,
		int	npartitions = `1`
	)

explicit

construct the net structure from protocol buffer.

Parameters

netproto	neural net config
npartitions	num of partitions. 1 for no partitioning.

Member Function Documentation

static shared_ptr<NeuralNet> singa::NeuralNet::Create	(	const NetProto &	np,
		Phase	phase,
		int	num
	)

static

Create the neural network for training, test or validation.

Parameters for test/validation net can share those from training after setup (done outside of this funcion).

Parameters

np	proto for the neural network
phase	test/training/validation
num	num of partitions, do partitioning if num > 1

Returns: shared pointer to a neural net

Graph* singa::NeuralNet::CreateGraph	(	const NetProto &	netproto,
		int	npartitions
	)

protected

Create a neural net graph, one node for each layer.

Partition the graph if npartitions > 1, each layer is sliced according to its own partition setting.

Parameters

netproto

Returns: neural net graph

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

/home/wangwei/program/asf/release-0.1/apache-singa-incubating-0.1.0-RC1/include/neuralnet/neuralnet.h

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation