org.nd4j.arrow

Class ArrowSerde

java.lang.Object

org.nd4j.arrow.ArrowSerde

public class ArrowSerde
extends Object

Conversion to and from arrow Tensor and INDArray

Author:

Adam Gibson

Constructor Summary

Constructors

Constructor and Description
ArrowSerde()

Method Summary

Modifier and Type	Method and Description
static int	addDataForArr(com.google.flatbuffers.FlatBufferBuilder bufferBuilder, INDArray arr) Create a Buffer representing the location metadata of the actual data contents for the ndarrays' DataBuffer
static void	addTypeTypeRelativeToNDArray(com.google.flatbuffers.FlatBufferBuilder bufferBuilder, INDArray arr) Convert the given INDArray data type to the proper data type for the tensor.
static int	createDims(com.google.flatbuffers.FlatBufferBuilder bufferBuilder, INDArray arr) Create the dimensions for the flatbuffer builder
static INDArray	fromTensor(org.apache.arrow.flatbuf.Tensor tensor) Convert a Tensor to an INDArray
static long[]	getArrowStrides(INDArray arr) Get the strides of this INDArray multiplieed by the element size.
static org.apache.arrow.flatbuf.Tensor	toTensor(INDArray arr) Convert an INDArray to an arrow Tensor
static org.nd4j.linalg.api.buffer.DataType	typeFromTensorType(byte type, int elementSize) Create thee databuffer type frm the given type, relative to the bytes in arrow in class: Type

Methods inherited from class java.lang.Object

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

Constructor Detail

ArrowSerde

public ArrowSerde()

Method Detail

fromTensor

public static INDArray fromTensor(org.apache.arrow.flatbuf.Tensor tensor)

Convert a Tensor to an INDArray

Parameters:

tensor - the input tensor

Returns:

the equivalent INDArray

toTensor

public static org.apache.arrow.flatbuf.Tensor toTensor(INDArray arr)

Convert an INDArray to an arrow Tensor

Parameters:

arr - the array to convert

Returns:

the equivalent Tensor

addDataForArr

public static int addDataForArr(com.google.flatbuffers.FlatBufferBuilder bufferBuilder,
                                INDArray arr)

Create a Buffer representing the location metadata of the actual data contents for the ndarrays' DataBuffer

Parameters:

bufferBuilder - the buffer builder in use

arr - the array to add the underlying data for

Returns:

the offset added

addTypeTypeRelativeToNDArray

public static void addTypeTypeRelativeToNDArray(com.google.flatbuffers.FlatBufferBuilder bufferBuilder,
                                                INDArray arr)

Convert the given INDArray data type to the proper data type for the tensor.

Parameters:

bufferBuilder - the buffer builder in use

arr - the array to conver tthe data type for

createDims

public static int createDims(com.google.flatbuffers.FlatBufferBuilder bufferBuilder,
                             INDArray arr)

Create the dimensions for the flatbuffer builder

Parameters:

bufferBuilder - the buffer builder to use

arr - the input array

Returns:

getArrowStrides

public static long[] getArrowStrides(INDArray arr)

Get the strides of this INDArray multiplieed by the element size. This is the Tensor and numpy format

Parameters:

arr - the array to convert

Returns:

typeFromTensorType

public static org.nd4j.linalg.api.buffer.DataType typeFromTensorType(byte type,
                                                                     int elementSize)

Create thee databuffer type frm the given type, relative to the bytes in arrow in class: Type

Parameters:

type - the type to create the nd4j DataType from

elementSize - the element size

Returns:

the data buffer type