org.nd4j.linalg.dataset

Class AsyncMultiDataSetIterator

java.lang.Object

org.nd4j.linalg.dataset.AsyncMultiDataSetIterator

All Implemented Interfaces:

Serializable, Iterator<MultiDataSet>, MultiDataSetIterator

public class AsyncMultiDataSetIterator
extends Object
implements MultiDataSetIterator

Async prefetching iterator wrapper for MultiDataSetIterator implementations This will asynchronously prefetch the specified number of minibatches from the underlying iterator.
Also has the option (enabled by default for most constructors) to use a cyclical workspace to avoid creating INDArrays with off-heap memory that needs to be cleaned up by the JVM garbage collector.
Note that appropriate DL4J fit methods automatically utilize this iterator, so users don't need to manually wrap their iterators when fitting a network

Author:

[email protected]

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected class	AsyncMultiDataSetIterator.AsyncPrefetchThread

Field Summary

Fields

Modifier and Type	Field and Description
protected MultiDataSetIterator	backedIterator
protected BlockingQueue<MultiDataSet>	buffer
protected DataSetCallback	callback
protected Integer	deviceId
protected AtomicBoolean	hasDepleted
protected MultiDataSet	nextElement
protected int	prefetchSize
protected AtomicBoolean	shouldWork
protected MultiDataSet	terminator
protected AsyncMultiDataSetIterator.AsyncPrefetchThread	thread
protected RuntimeException	throwable
protected boolean	useWorkspaces
protected String	workspaceId

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	AsyncMultiDataSetIterator()
	AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator)
	AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator, int queueSize)
	AsyncMultiDataSetIterator(MultiDataSetIterator iterator, int queueSize, BlockingQueue<MultiDataSet> queue)
	AsyncMultiDataSetIterator(MultiDataSetIterator iterator, int queueSize, BlockingQueue<MultiDataSet> queue, boolean useWorkspace)
	AsyncMultiDataSetIterator(MultiDataSetIterator iterator, int queueSize, BlockingQueue<MultiDataSet> queue, boolean useWorkspace, DataSetCallback callback)
	AsyncMultiDataSetIterator(MultiDataSetIterator iterator, int queueSize, BlockingQueue<MultiDataSet> queue, boolean useWorkspace, DataSetCallback callback, Integer deviceId)
	AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator, int queueSize, boolean useWorkspace)
	AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator, int queueSize, boolean useWorkspace, Integer deviceId)

Method Summary

Modifier and Type	Method and Description
boolean	asyncSupported() Does this DataSetIterator support asynchronous prefetching of multiple DataSet objects? Most DataSetIterators do, but in some cases it may not make sense to wrap this iterator in an iterator that does asynchronous prefetching.
protected void	externalCall()
MultiDataSetPreProcessor	getPreProcessor() Get the MultiDataSetPreProcessor, if one has previously been set.
boolean	hasNext() Returns true if the iteration has more elements.
MultiDataSet	next() Returns the next element in the iteration.
MultiDataSet	next(int num) Like the standard next method but allows a customizable number of examples returned
void	remove() Removes from the underlying collection the last element returned by this iterator (optional operation).
void	reset() Resets the iterator back to the beginning
boolean	resetSupported() Is resetting supported by this DataSetIterator? Many DataSetIterators do support resetting, but some don't
void	setPreProcessor(MultiDataSetPreProcessor preProcessor) Set the preprocessor to be applied to each MultiDataSet, before each MultiDataSet is returned.
void	shutdown() This method will terminate background thread AND will destroy attached workspace (if any) PLEASE NOTE: After shutdown() call, this instance can't be used anymore

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface java.util.Iterator

forEachRemaining

Field Detail

backedIterator

protected MultiDataSetIterator backedIterator

terminator

protected MultiDataSet terminator

nextElement

protected MultiDataSet nextElement

buffer

protected BlockingQueue<MultiDataSet> buffer

thread

protected AsyncMultiDataSetIterator.AsyncPrefetchThread thread

shouldWork

protected AtomicBoolean shouldWork

throwable

protected volatile RuntimeException throwable

useWorkspaces

protected boolean useWorkspaces

prefetchSize

protected int prefetchSize

workspaceId

protected String workspaceId

callback

protected DataSetCallback callback

deviceId

protected Integer deviceId

hasDepleted

protected AtomicBoolean hasDepleted

Constructor Detail

AsyncMultiDataSetIterator

protected AsyncMultiDataSetIterator()

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator iterator,
                                 int queueSize,
                                 BlockingQueue<MultiDataSet> queue)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator,
                                 int queueSize)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator,
                                 int queueSize,
                                 boolean useWorkspace)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator baseIterator,
                                 int queueSize,
                                 boolean useWorkspace,
                                 Integer deviceId)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator iterator,
                                 int queueSize,
                                 BlockingQueue<MultiDataSet> queue,
                                 boolean useWorkspace)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator iterator,
                                 int queueSize,
                                 BlockingQueue<MultiDataSet> queue,
                                 boolean useWorkspace,
                                 DataSetCallback callback)

AsyncMultiDataSetIterator

public AsyncMultiDataSetIterator(MultiDataSetIterator iterator,
                                 int queueSize,
                                 BlockingQueue<MultiDataSet> queue,
                                 boolean useWorkspace,
                                 DataSetCallback callback,
                                 Integer deviceId)

Method Detail

next

public MultiDataSet next(int num)

Like the standard next method but allows a customizable number of examples returned

Specified by:

next in interface MultiDataSetIterator

Parameters:

num - the number of examples

Returns:

the next data applyTransformToDestination

setPreProcessor

public void setPreProcessor(MultiDataSetPreProcessor preProcessor)

Set the preprocessor to be applied to each MultiDataSet, before each MultiDataSet is returned.

Specified by:

setPreProcessor in interface MultiDataSetIterator

Parameters:

preProcessor - MultiDataSetPreProcessor. May be null.

getPreProcessor

public MultiDataSetPreProcessor getPreProcessor()

Description copied from interface: MultiDataSetIterator

Get the MultiDataSetPreProcessor, if one has previously been set. Returns null if no preprocessor has been set

Specified by:

getPreProcessor in interface MultiDataSetIterator

Returns:

Preprocessor

resetSupported

public boolean resetSupported()

Is resetting supported by this DataSetIterator? Many DataSetIterators do support resetting, but some don't

Specified by:

resetSupported in interface MultiDataSetIterator

Returns:

true if reset method is supported; false otherwise

asyncSupported

public boolean asyncSupported()

Does this DataSetIterator support asynchronous prefetching of multiple DataSet objects? Most DataSetIterators do, but in some cases it may not make sense to wrap this iterator in an iterator that does asynchronous prefetching. For example, it would not make sense to use asynchronous prefetching for the following types of iterators: (a) Iterators that store their full contents in memory already (b) Iterators that re-use features/labels arrays (as future next() calls will overwrite past contents) (c) Iterators that already implement some level of asynchronous prefetching (d) Iterators that may return different data depending on when the next() method is called

Specified by:

asyncSupported in interface MultiDataSetIterator

Returns:

true if asynchronous prefetching from this iterator is OK; false if asynchronous prefetching should not be used with this iterator

reset

public void reset()

Resets the iterator back to the beginning

Specified by:

reset in interface MultiDataSetIterator

shutdown

public void shutdown()

This method will terminate background thread AND will destroy attached workspace (if any) PLEASE NOTE: After shutdown() call, this instance can't be used anymore

hasNext

public boolean hasNext()

Returns true if the iteration has more elements. (In other words, returns true if next(int) would return an element rather than throwing an exception.)

Specified by:

hasNext in interface Iterator<MultiDataSet>

Returns:

true if the iteration has more elements

next

public MultiDataSet next()

Returns the next element in the iteration.

Specified by:

next in interface Iterator<MultiDataSet>

Returns:

the next element in the iteration

remove

public void remove()

Removes from the underlying collection the last element returned by this iterator (optional operation). This method can be called only once per call to next(int). The behavior of an iterator is unspecified if the underlying collection is modified while the iteration is in progress in any way other than by calling this method.

Specified by:

remove in interface Iterator<MultiDataSet>

Throws:

UnsupportedOperationException - if the remove operation is not supported by this iterator

IllegalStateException - if the next method has not yet been called, or the remove method has already been called after the last call to the next method

externalCall

protected void externalCall()