Learns an entire epoch of training, for data that wont fit in memory. More...

Public Member Functions
publicapi	OnDemandBatcherv2 (Trainable net, NetAction netAction, GenericLoaderv2 *loader, int N, int fileReadBatches, int batchSize)

virtual publicapi int	getNextFileBatch ()

virtual publicapi int	getNextBatch ()

virtual publicapi float	getLoss ()

virtual publicapi int	getNumRight ()

virtual publicapi bool	getEpochDone ()

virtual publicapi int	getN ()

publicapi void	reset ()

publicapi bool	tick (int epoch)

publicapi EpochResult	run (int epoch)

Detailed Description

Learns an entire epoch of training, for data that wont fit in memory.

This handles an entire epoch of training, loading in data in chunks and then passing it to a 'Batcher' class, to train/test each chunk

If you want to run multiple epochs, you can use a 'NetLearnerOnDemand' class

compared to v1, v2 recevies a GenericLoaderv2 loader object, instead of a filepath so we can handle imagenet manifests etc

Constructor & Destructor Documentation

publicapi OnDemandBatcherv2::OnDemandBatcherv2	(	Trainable *	net,
		NetAction *	netAction,
		GenericLoaderv2 *	loader,
		int	N,
		int	fileReadBatches,
		int	batchSize
	)

publicapi VIRTUAL int OnDemandBatcherv2::getNextFileBatch ( )

virtual

publicapi VIRTUAL int OnDemandBatcherv2::getNextBatch ( )

virtual

publicapi VIRTUAL float OnDemandBatcherv2::getLoss ( )

virtual

publicapi VIRTUAL int OnDemandBatcherv2::getNumRight ( )

virtual

publicapi VIRTUAL bool OnDemandBatcherv2::getEpochDone ( )

virtual

publicapi VIRTUAL int OnDemandBatcherv2::getN ( )

virtual

publicapi void OnDemandBatcherv2::reset ( )

publicapi bool OnDemandBatcherv2::tick ( int epoch )

publicapi EpochResult OnDemandBatcherv2::run ( int epoch )

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