Package ai.djl.nn

Class Blocks

java.lang.Object

ai.djl.nn.Blocks

public final class Blocks
extends Object

Utility class that provides some useful blocks.

Method Summary

Modifier and Type	Method	Description
static NDArray	batchFlatten(NDArray array)	Inflates the NDArray provided as input to a 2-D NDArray of shape (batch, size).
static NDArray	batchFlatten(NDArray array, long size)	Inflates the NDArray provided as input to a 2-D NDArray of shape (batch, size).
static Block	batchFlattenBlock()	Creates a Block whose forward function applies the batchFlatten method.
static Block	batchFlattenBlock(long size)	Creates a Block whose forward function applies the batchFlatten method.
static String	describe(Block block, String blockName, int beginAxis)	Returns a string representation of the passed Block describing the input axes, output axes, and the block's children.
static Block	identityBlock()	Creates a LambdaBlock that performs the identity function.
static Block	onesBlock(ai.djl.util.PairList<DataType,Shape> shapes, String[] names)	Creates a LambdaBlock that return all-ones NDList.

Methods inherited from class java.lang.Object

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

Method Details

batchFlatten

public static NDArray batchFlatten(NDArray array)

Inflates the NDArray provided as input to a 2-D NDArray of shape (batch, size).

Parameters:

array - a array to be flattened

Returns:

a NDList that contains the inflated NDArray

batchFlatten

public static NDArray batchFlatten(NDArray array,
 long size)

Inflates the NDArray provided as input to a 2-D NDArray of shape (batch, size).

Parameters:

array - a array to be flattened

size - the input size

Returns:

a NDList that contains the inflated NDArray

Throws:

IndexOutOfBoundsException - if the input NDList has more than one NDArray

batchFlattenBlock

public static Block batchFlattenBlock()

Creates a Block whose forward function applies the batchFlatten method.

Returns:

a Block whose forward function applies the batchFlatten method

batchFlattenBlock

public static Block batchFlattenBlock(long size)

Creates a Block whose forward function applies the batchFlatten method. The size of input to the block returned must be batch_size * size.

Parameters:

size - the expected size of each input

Returns:

a Block whose forward function applies the batchFlatten method

identityBlock

public static Block identityBlock()

Creates a LambdaBlock that performs the identity function.

Returns:

an identity Block

onesBlock

public static Block onesBlock(ai.djl.util.PairList<DataType,Shape> shapes,
 String[] names)

Creates a LambdaBlock that return all-ones NDList.

Returns:

an all-ones Block

describe

public static String describe(Block block,
 String blockName,
 int beginAxis)

Returns a string representation of the passed Block describing the input axes, output axes, and the block's children.

Parameters:

block - the block to describe

blockName - the name to be used for the passed block, or null if its class name is to be used

beginAxis - skips all axes before this axis; use 0 to print all axes and 1 to skip the batch axis.

Returns:

the string representation