Class ND4JBaseTensorBackend

java.lang.Object

org.episteme.nativ.mathematics.linearalgebra.tensors.backends.ND4JBaseTensorBackend

All Implemented Interfaces:

TensorBackend, TensorProvider, AlgorithmProvider, Backend, ComputeBackend

Direct Known Subclasses:

NativeND4JCUDATensorBackend, NativeND4JTensorBackend

public abstract class ND4JBaseTensorBackend
extends Object
implements TensorBackend

Abstract base for ND4J-backed Tensor backends.

Sub-classes target specific ND4J device backends (native CPU, CUDA GPU).

Field Summary

Fields

Modifier and Type	Field	Description
protected final boolean	isAvailable

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	ND4JBaseTensorBackend()

Method Summary

Modifier and Type	Method	Description
protected abstract boolean	checkAvailability()
protected boolean	checkCommonClasses()
protected long[]	convertShape(int[] shape)
<T> Tensor<T>	create(T[] data, int... shape)	Creates a tensor from a flat array and shape.
ExecutionContext	createContext()	Creates an execution context for running operations.
protected abstract ExecutionContext	doGetContext()
Tensor<Real>	fromINDArray(org.nd4j.linalg.api.ndarray.INDArray indArray)
boolean	isAvailable()	Checks if this backend is currently available (libraries loaded, etc.).
<T> Tensor<T>	ones(Class<T> elementType, int... shape)	Creates a tensor with the given shape, filled with ones.
abstract boolean	supportsGPU()	Returns true if this provider supports GPU acceleration.
boolean	supportsParallelOps()	Checks if this backend supports parallel operations.
org.nd4j.linalg.api.ndarray.INDArray	toINDArray(Tensor<Real> tensor)
<T> Tensor<T>	zeros(Class<T> elementType, int... shape)	Creates a tensor with the given shape, filled with zeros.

Methods inherited from class Object

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

Methods inherited from interface AlgorithmProvider

description, getMetadata, score

Methods inherited from interface Backend

getAlgorithmProviders, getStatusMessage, isExplicitlyDisabled

Methods inherited from interface ComputeBackend

createBackend, getAcceleratorType, getDescription, getId, getType, supportsComplexNumbers, supportsFloatingPoint

Methods inherited from interface TensorBackend

getName, getPriority, shutdown

Methods inherited from interface TensorProvider

getAlgorithmType

Field Details

isAvailable

protected final boolean isAvailable

Constructor Details

ND4JBaseTensorBackend

protected ND4JBaseTensorBackend()

Method Details

checkAvailability

protected abstract boolean checkAvailability()

supportsGPU

public abstract boolean supportsGPU()

Description copied from interface: TensorBackend

Returns true if this provider supports GPU acceleration.

Specified by:

supportsGPU in interface TensorBackend

Returns:

true if GPU is available

isAvailable

public boolean isAvailable()

Description copied from interface: Backend

Checks if this backend is currently available (libraries loaded, etc.).

Default implementation checks Backend.isExplicitlyDisabled() and returns false if it is.

Specified by:

isAvailable in interface AlgorithmProvider

Specified by:

isAvailable in interface Backend

Specified by:

isAvailable in interface TensorBackend

supportsParallelOps

public boolean supportsParallelOps()

Description copied from interface: ComputeBackend

Checks if this backend supports parallel operations.

Specified by:

supportsParallelOps in interface ComputeBackend

Returns:

true if parallel execution is supported

checkCommonClasses

protected boolean checkCommonClasses()

doGetContext

protected abstract ExecutionContext doGetContext()

createContext

public ExecutionContext createContext()

Description copied from interface: ComputeBackend

Creates an execution context for running operations.

The context should be used within a try-with-resources block to ensure proper cleanup of resources.

Specified by:

createContext in interface ComputeBackend

Returns:

new execution context

zeros

public <T> Tensor<T> zeros(Class<T> elementType,
 int... shape)

Description copied from interface: TensorBackend

Creates a tensor with the given shape, filled with zeros.

Specified by:

zeros in interface TensorBackend

Specified by:

zeros in interface TensorProvider

Type Parameters:

T - the element type

Parameters:

elementType - the class of field elements

shape - the tensor shape

Returns:

zero-initialized tensor

ones

public <T> Tensor<T> ones(Class<T> elementType,
 int... shape)

Description copied from interface: TensorBackend

Creates a tensor with the given shape, filled with ones.

Specified by:

ones in interface TensorBackend

Specified by:

ones in interface TensorProvider

Type Parameters:

T - the element type

Parameters:

elementType - the class of field elements

shape - the tensor shape

Returns:

one-initialized tensor

create

public <T> Tensor<T> create(T[] data,
 int... shape)

Description copied from interface: TensorBackend

Creates a tensor from a flat array and shape.

Specified by:

create in interface TensorBackend

Specified by:

create in interface TensorProvider

Type Parameters:

T - the element type

Parameters:

data - the flat data array

shape - the tensor shape

Returns:

tensor with given data

toINDArray

public org.nd4j.linalg.api.ndarray.INDArray toINDArray(Tensor<Real> tensor)

fromINDArray

public Tensor<Real> fromINDArray(org.nd4j.linalg.api.ndarray.INDArray indArray)

convertShape

protected long[] convertShape(int[] shape)