Package ai.djl.nn

Class Blocks

java.lang.Object

ai.djl.nn.Blocks

public final class Blocks
extends java.lang.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 Block	identityBlock()	Creates a LambdaBlock that performs the identity function.

Methods inherited from class java.lang.Object

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

Method Detail

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:

java.lang.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