Class GPUComputeBackend

Represents a GPUCompute backend ops.

Inheritance

object

GPUComputeBackend

Implements

IBackend

IDisposable

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Namespace: Unity.Sentis

Assembly: Unity.Sentis.dll

Syntax

public class GPUComputeBackend : IBackend, IDisposable

Constructors

GPUComputeBackend()

Initializes and returns an instance of GPUComputeOps.

Declaration

public GPUComputeBackend()

Properties

backendType

Returns the BackendType for the ops.

Declaration

public BackendType backendType { get; }

Property Value

Type	Description
BackendType

Methods

Abs(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Abs math function: f(x) = f(x) = |x|.

Declaration

public void Abs(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Abs(TensorInt, TensorInt)

Computes an output tensor by applying the element-wise Abs math function: f(x) = f(x) = |x|.

Declaration

public void Abs(TensorInt X, TensorInt O)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

Acos(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Acos trigonometric function: f(x) = acos(x).

Declaration

public void Acos(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Acosh(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Acosh trigonometric function: f(x) = acosh(x).

Declaration

public void Acosh(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Add(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Add math operation: f(a, b) = a + b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Add(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Add(TensorInt, TensorInt, TensorInt)

Performs an element-wise Add math operation: f(a, b) = a + b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Add(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

And(TensorInt, TensorInt, TensorInt)

Performs an element-wise And logical operation: f(a, b) = a & b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void And(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

ArgMax(TensorFloat, TensorInt, int, bool)

Computes the indices of the maximum elements of the input tensor along a given axis.

Declaration

public void ArgMax(TensorFloat X, TensorInt O, int axis, bool selectLastIndex)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	axis	The axis along which to reduce.
bool	selectLastIndex	Whether to perform the operation from the back of the axis.

ArgMax(TensorInt, TensorInt, int, bool)

Computes the indices of the maximum elements of the input tensor along a given axis.

Declaration

public void ArgMax(TensorInt X, TensorInt O, int axis, bool selectLastIndex)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	axis	The axis along which to reduce.
bool	selectLastIndex	Whether to perform the operation from the back of the axis.

ArgMin(TensorFloat, TensorInt, int, bool)

Computes the indices of the minimum elements of the input tensor along a given axis.

Declaration

public void ArgMin(TensorFloat X, TensorInt O, int axis, bool selectLastIndex)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	axis	The axis along which to reduce.
bool	selectLastIndex	Whether to perform the operation from the back of the axis.

ArgMin(TensorInt, TensorInt, int, bool)

Computes the indices of the minimum elements of the input tensor along a given axis.

Declaration

public void ArgMin(TensorInt X, TensorInt O, int axis, bool selectLastIndex)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	axis	The axis along which to reduce.
bool	selectLastIndex	Whether to perform the operation from the back of the axis.

Asin(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Asin trigonometric function: f(x) = asin(x).

Declaration

public void Asin(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Asinh(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Asinh trigonometric function: f(x) = asinh(x).

Declaration

public void Asinh(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Atan(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Atan trigonometric function: f(x) = atan(x).

Declaration

public void Atan(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Atanh(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Atanh trigonometric function: f(x) = atanh(x).

Declaration

public void Atanh(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

AveragePool(TensorFloat, TensorFloat, int[], int[], int[])

Calculates an output tensor by pooling the mean values of the input tensor across its spatial dimensions according to the given pool and stride values.

Declaration

public void AveragePool(TensorFloat X, TensorFloat O, int[] kernelShape, int[] strides, int[] pads)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int[]	kernelShape	The size of the kernel along each spatial axis.
int[]	strides	The stride along each spatial axis.
int[]	pads	The lower and upper padding values for each spatial dimension. For example, [pad_left, pad_right] for 1D, or [pad_top, pad_bottom, pad_left, pad_right] for 2D.

BatchNormalization(TensorFloat, TensorFloat, TensorFloat, TensorFloat, TensorFloat, TensorFloat, float)

Computes the mean variance on the last dimension of the input tensor and normalizes it according to scale and bias tensors.

Declaration

public void BatchNormalization(TensorFloat X, TensorFloat S, TensorFloat B, TensorFloat mean, TensorFloat variance, TensorFloat O, float epsilon)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	S	The scale tensor.
TensorFloat	B	The bias tensor.
TensorFloat	mean	The mean tensor.
TensorFloat	variance	The variance tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	epsilon	The epsilon value the layer uses to avoid division by zero.

Bernoulli(TensorFloat, Tensor, int?)

Generates an output tensor with values 0 or 1 from a Bernoulli distribution. The input tensor contains the probabilities to use for generating the output values.

Declaration

public void Bernoulli(TensorFloat X, Tensor O, int? seed)

Parameters

Type	Name	Description
TensorFloat	X	The probabilities input tensor.
Tensor	O	The output tensor to be computed and filled.
int?	seed	The optional seed to use for the random number generation. If this is null the operation generates a seed using System.Random().

Cast(TensorFloat, TensorInt)

Computes the output tensor using an element-wise Cast function: f(x) = (int)x.

Declaration

public void Cast(TensorFloat X, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

Cast(TensorInt, TensorFloat)

Computes the output tensor using an element-wise Cast function: f(x) = (float)x.

Declaration

public void Cast(TensorInt X, TensorFloat O)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Cast(TensorShort, TensorFloat)

Computes the output tensor using an element-wise Cast function: f(x) = (float)x.

Declaration

public void Cast(TensorShort X, TensorFloat O)

Parameters

Type	Name	Description
TensorShort	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Ceil(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Ceil math function: f(x) = ceil(x).

Declaration

public void Ceil(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Celu(TensorFloat, TensorFloat, float)

Computes an output tensor by applying the element-wise Celu activation function: f(x) = max(0, x) + min(0, alpha * (exp(x / alpha) - 1)).

Declaration

public void Celu(TensorFloat X, TensorFloat O, float alpha)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	alpha	The alpha value to use for the Celu activation function.

Clip(TensorFloat, TensorFloat, float, float)

Computes an output tensor by applying the element-wise Clip math function: f(x) = clamp(x, min, max).

Declaration

public void Clip(TensorFloat X, TensorFloat O, float min, float max)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	min	The lower clip value.
float	max	The upper clip value.

Clip(TensorInt, TensorInt, int, int)

Computes an output tensor by applying the element-wise Clip math function: f(x) = clamp(x, min, max).

Declaration

public void Clip(TensorInt X, TensorInt O, int min, int max)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	min	The lower clip value.
int	max	The upper clip value.

CompressWithIndices(Tensor, TensorInt, Tensor, int, int)

Computes the output tensor by selecting slices from an input tensor according to the 'indices' tensor along an 'axis'.

Declaration

public void CompressWithIndices(Tensor X, TensorInt indices, Tensor O, int numIndices, int axis)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
TensorInt	indices	The indices tensor.
Tensor	O	The output tensor to be computed and filled.
int	numIndices	The number of indices.
int	axis	The axis along which to compress.

Conv(TensorFloat, TensorFloat, TensorFloat, TensorFloat, int, Span<int>, Span<int>, Span<int>, FusableActivation)

Applies a convolution filter to an input tensor.

Declaration

public void Conv(TensorFloat X, TensorFloat K, TensorFloat B, TensorFloat O, int groups, Span<int> strides, Span<int> pads, Span<int> dilations, FusableActivation fusedActivation)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	K	The filter tensor.
TensorFloat	B	The optional bias tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	groups	The number of groups that input channels and output channels are divided into.
Span<int>	strides	The optional stride value for each spatial dimension of the filter.
Span<int>	pads	The optional lower and upper padding values for each spatial dimension of the filter.
Span<int>	dilations	The optional dilation value of each spatial dimension of the filter.
FusableActivation	fusedActivation	The fused activation type to apply after the convolution.

ConvTranspose(TensorFloat, TensorFloat, TensorFloat, TensorFloat, Span<int>, Span<int>, Span<int>, FusableActivation)

Applies a transpose convolution filter to an input tensor.

Declaration

public void ConvTranspose(TensorFloat X, TensorFloat W, TensorFloat B, TensorFloat O, Span<int> strides, Span<int> pads, Span<int> outputPadding, FusableActivation fusedActivation)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	W	The filter tensor.
TensorFloat	B	The optional bias tensor.
TensorFloat	O	The output tensor to be computed and filled.
Span<int>	strides	The optional stride value for each spatial dimension of the filter.
Span<int>	pads	The optional lower and upper padding values for each spatial dimension of the filter.
Span<int>	outputPadding	The output padding value for each spatial dimension in the filter.
FusableActivation	fusedActivation	The fused activation type to apply after the convolution.

Cos(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Cos trigonometric function: f(x) = cos(x).

Declaration

public void Cos(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Cosh(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Cosh trigonometric function: f(x) = cosh(x).

Declaration

public void Cosh(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

CumSum(TensorFloat, TensorFloat, int, bool, bool)

Performs the cumulative sum along a given axis.

Declaration

public void CumSum(TensorFloat X, TensorFloat O, int axis, bool reverse, bool exclusive)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	axis	The axis along which to apply the cumulative sum.
bool	reverse	Whether to perform the cumulative sum from the end of the axis.
bool	exclusive	Whether to include the respective input element in the cumulative sum.

CumSum(TensorInt, TensorInt, int, bool, bool)

Performs the cumulative sum along a given axis.

Declaration

public void CumSum(TensorInt X, TensorInt O, int axis, bool reverse, bool exclusive)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	axis	The axis along which to apply the cumulative sum.
bool	reverse	Whether to perform the cumulative sum from the end of the axis.
bool	exclusive	Whether to include the respective input element in the cumulative sum.

Dense(TensorFloat, TensorFloat, TensorFloat, TensorFloat, FusableActivation)

Performs a matrix multiplication operation: f(x, w, b) = X x W + B.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Dense(TensorFloat X, TensorFloat W, TensorFloat B, TensorFloat O, FusableActivation fusedActivation)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	W	The weights tensor.
TensorFloat	B	The bias tensor.
TensorFloat	O	The output tensor to be computed and filled.
FusableActivation	fusedActivation	The fused activation to apply to the output tensor after the dense operation.

DepthToSpace(TensorFloat, TensorFloat, int, DepthToSpaceMode)

Computes the output tensor by permuting data from depth into blocks of spatial data.

Declaration

public void DepthToSpace(TensorFloat X, TensorFloat O, int blocksize, DepthToSpaceMode mode)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	blocksize	The size of the blocks to move the depth data into.
DepthToSpaceMode	mode	The ordering of the data in the output tensor as a DepthToSpaceMode.

DequantizeLinear(TensorByte, TensorFloat, float, byte)

Computes the output tensor by unpacking four uint8 values from each int value and scaling to floats.

Declaration

public void DequantizeLinear(TensorByte X, TensorFloat O, float scale, byte zeroPoint)

Parameters

Type	Name	Description
TensorByte	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	scale	The scale value to use for dequantization.
byte	zeroPoint	The zero point value to use for dequantization.

Dispose()

Disposes of the ops and any associated memory.

Declaration

public void Dispose()

Div(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Div math operation: f(a, b) = a / b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Div(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Div(TensorInt, TensorInt, TensorInt)

Performs an element-wise Div math operation: f(a, b) = a / b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Div(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Einsum(TensorFloat[], TensorFloat, TensorIndex[], TensorIndex, TensorIndex, TensorShape)

Performs an Einsum math operation. The Einsum operator evaluates algebraic tensor operations on a sequence of tensors, using the Einstein summation convention. The equation string contains a comma-separated sequence of lower case letters. Each term corresponds to an operand tensor, and the characters within the terms correspond to operands dimensions. This sequence may be followed by "->" to separate the left and right hand side of the equation. If the equation contains "->" followed by the right-hand side, the explicit (not classical) form of the Einstein summation is performed, and the right-hand side indices indicate output tensor dimensions. In other cases, output indices are (implicitly) set to the alphabetically sorted sequence of indices appearing exactly once in the equation. When a dimension character is repeated in the left-hand side, it represents summation along the dimension. The equation may contain ellipsis ("...") to enable broadcasting. Ellipsis must indicate a fixed number of dimensions. Specifically, every occurrence of ellipsis in the equation must represent the same number of dimensions. The right-hand side may contain exactly one ellipsis. In implicit mode, the ellipsis dimensions are set to the beginning of the output. The equation string may contain space (U+0020) character.

Declaration

public void Einsum(TensorFloat[] inputTensors, TensorFloat O, TensorIndex[] operandIndices, TensorIndex outputIndices, TensorIndex sumIndices, TensorShape sumShape)

Parameters

Type	Name	Description
TensorFloat[]	inputTensors	The input tensors.
TensorFloat	O	The output tensor to be computed and filled.
TensorIndex[]	operandIndices	The operand indices for each input tensor.
TensorIndex	outputIndices	The output indices for each input tensor.
TensorIndex	sumIndices	The indices along which to sum.
TensorShape	sumShape	The shape along which to sum.

Elu(TensorFloat, TensorFloat, float)

Computes an output tensor by applying the element-wise Elu activation function: f(x) = x if x >= 0, otherwise f(x) = alpha * (e^x - 1).

Declaration

public void Elu(TensorFloat X, TensorFloat O, float alpha)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	alpha	The alpha value to use for the Elu activation function.

Equal(TensorFloat, TensorFloat, TensorInt)

Performs an element-wise Equal logical comparison operation: f(a, b) = 1 if a == b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Equal(TensorFloat A, TensorFloat B, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Equal(TensorInt, TensorInt, TensorInt)

Performs an element-wise Equal logical comparison operation: f(a, b) = 1 if a == b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Equal(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Erf(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Erf activation function: f(x) = erf(x).

Declaration

public void Erf(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Exp(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Exp math function: f(x) = exp(x).

Declaration

public void Exp(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Expand(Tensor, Tensor)

Calculates an output tensor by broadcasting the input tensor into a given shape.

Declaration

public void Expand(Tensor X, Tensor O)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.

FMod(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Mod math operation: f(a, b) = a % b.

The sign of the remainder is the same as the sign of the dividend, as in C#.

This supports numpy-style broadcasting of input tensors.

Declaration

public void FMod(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

FMod(TensorInt, TensorInt, TensorInt)

Performs an element-wise Mod math operation: f(a, b) = a % b.

The sign of the remainder is the same as the sign of the dividend, as in C#.

This supports numpy-style broadcasting of input tensors.

Declaration

public void FMod(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Floor(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Floor math function: f(x) = floor(x).

Declaration

public void Floor(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Gather(Tensor, TensorInt, Tensor, int)

Takes values from the input tensor indexed by the indices tensor along a given axis and concatenates them.

Declaration

public void Gather(Tensor X, TensorInt indices, Tensor O, int axis)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
TensorInt	indices	The indices tensor.
Tensor	O	The output tensor to be computed and filled.
int	axis	The axis along which to gather.

GatherElements(Tensor, TensorInt, Tensor, int)

Takes values from the input tensor indexed by the indices tensor along a given axis and concatenates them.

Declaration

public void GatherElements(Tensor X, TensorInt indices, Tensor O, int axis)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
TensorInt	indices	The indices tensor.
Tensor	O	The output tensor to be computed and filled.
int	axis	The axis along which to gather.

GatherND(Tensor, TensorInt, Tensor, int)

Takes slices of values from the batched input tensor indexed by the indices tensor.

Declaration

public void GatherND(Tensor X, TensorInt indices, Tensor O, int batchDims)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
TensorInt	indices	The indices tensor.
Tensor	O	The output tensor to be computed and filled.
int	batchDims	The number of batch dimensions of the input tensor, the gather begins at the next dimension.

Gelu(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Gelu activation function: f(x) = x / 2 * (1 + erf(x / sqrt(2))).

Declaration

public void Gelu(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

GeluFast(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Gelu aproximate but fast gelu function: f(x) = (x / 2) * (tanh(x + x^3 * 0.04472) * 0.7978) + 1.

Declaration

public void GeluFast(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

GlobalAveragePool(TensorFloat, TensorFloat)

Calculates an output tensor by pooling the mean values of the input tensor across all of its spatial dimensions. The spatial dimensions of the output are size 1.

Declaration

public void GlobalAveragePool(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

GlobalAverageVariancePool(TensorFloat, TensorFloat, int)

Declaration

public void GlobalAverageVariancePool(TensorFloat X, TensorFloat O, int axis)

Parameters

Type	Name	Description
TensorFloat	X
TensorFloat	O
int	axis

GlobalMaxPool(TensorFloat, TensorFloat)

Calculates an output tensor by pooling the maximum values of the input tensor across all of its spatial dimensions. The spatial dimensions of the output are size 1.

Declaration

public void GlobalMaxPool(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Greater(TensorFloat, TensorFloat, TensorInt)

Performs an element-wise Greater logical comparison operation: f(a, b) = 1 if a > b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Greater(TensorFloat A, TensorFloat B, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Greater(TensorInt, TensorInt, TensorInt)

Performs an element-wise Greater logical comparison operation: f(a, b) = 1 if a > b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Greater(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

GreaterOrEqual(TensorFloat, TensorFloat, TensorInt)

Performs an element-wise GreaterOrEqual logical comparison operation: f(a, b) = 1 if a >= b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void GreaterOrEqual(TensorFloat A, TensorFloat B, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

GreaterOrEqual(TensorInt, TensorInt, TensorInt)

Performs an element-wise GreaterOrEqual logical comparison operation: f(a, b) = 1 if a >= b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void GreaterOrEqual(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

GridSample(TensorFloat, TensorFloat, TensorFloat, InterpolationMode, PaddingMode, bool)

Calculates an output tensor by sampling the input tensor by coordinates given by the grid tensor.

Declaration

public void GridSample(TensorFloat X, TensorFloat grid, TensorFloat O, InterpolationMode mode, PaddingMode paddingMode, bool alignCorners)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	grid	The grid tensor containing the spatial coordinates per output pixel.
TensorFloat	O	The output tensor to be computed and filled.
InterpolationMode	mode	The InterpolationMode to use for the operation.
PaddingMode	paddingMode	The PaddingMode to use for the operation.
bool	alignCorners	Whether to map the extreme values in the coordinates 0 and 1 to the centre of the corner pixels rather than the outer corners.

HardSigmoid(TensorFloat, TensorFloat, float, float)

Computes an output tensor by applying the element-wise HardSigmoid activation function: f(x) = clamp(alpha * x + beta, 0, 1).

Declaration

public void HardSigmoid(TensorFloat X, TensorFloat O, float alpha, float beta)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	alpha	The alpha value to use for the HardSigmoid activation function.
float	beta	The beta value to use for the HardSigmoid activation function.

HardSwish(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise HardSwish activation function: f(x) = x * max(0, min(1, 1/6 * x + 0.5)).

Declaration

public void HardSwish(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Hardmax(TensorFloat, TensorFloat, int)

Computes an output tensor by applying the Hardmax activation function along an axis: f(x, axis) = 1 if x is the first maximum value along the specified axis, otherwise f(x) = 0.

Declaration

public void Hardmax(TensorFloat X, TensorFloat O, int axis)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	axis	The axis along which to apply the Hardmax activation function.

InstanceNormalization(TensorFloat, TensorFloat, TensorFloat, TensorFloat, float)

Computes the mean variance on the spatial dimensions of the input tensor and normalizes them according to scale and bias tensors.

Declaration

public void InstanceNormalization(TensorFloat X, TensorFloat S, TensorFloat B, TensorFloat O, float epsilon)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	S	The scale tensor.
TensorFloat	B	The bias tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	epsilon	The epsilon value the layer uses to avoid division by zero.

IsInf(TensorFloat, TensorInt, bool, bool)

Performs an element-wise IsInf logical operation: f(x) = 1 elementwise if x is +Inf and detectPositive is true, or x is -Inf and detectNegative is true. Otherwise f(x) = 0.

Declaration

public void IsInf(TensorFloat X, TensorInt O, bool detectNegative, bool detectPositive)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
bool	detectNegative	Whether to detect negative infinities in the IsInf function.
bool	detectPositive	Whether to detect positive infinities in the IsInf function.

IsNaN(TensorFloat, TensorInt)

Performs an element-wise IsNaN logical operation: f(x) = 1 if x is NaN, otherwise f(x) = 0.

Declaration

public void IsNaN(TensorFloat X, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

LSTM(TensorFloat, TensorFloat, TensorFloat, TensorFloat, TensorInt, TensorFloat, TensorFloat, TensorFloat, TensorFloat, TensorFloat, TensorFloat, RnnDirection, RnnActivation[], float[], float[], bool, float, RnnLayout)

Generates an output tensor by computing a one-layer long short-term memory (LSTM) on an input tensor.

Declaration

public void LSTM(TensorFloat X, TensorFloat W, TensorFloat R, TensorFloat B, TensorInt sequenceLens, TensorFloat initialH, TensorFloat initialC, TensorFloat P, TensorFloat Y, TensorFloat Yh, TensorFloat Yc, RnnDirection direction, RnnActivation[] activations, float[] activationAlpha, float[] activationBeta, bool inputForget, float clip, RnnLayout layout)

Parameters

Type	Name	Description
TensorFloat	X	The input sequences tensor.
TensorFloat	W	The weights tensor for the gates of the LSTM.
TensorFloat	R	The recurrent weights tensor for the gates of the LSTM.
TensorFloat	B	The optional bias tensor for the input gate of the LSTM.
TensorInt	sequenceLens	The optional 1D tensor specifying the lengths of the sequences in a batch.
TensorFloat	initialH	The optional initial values tensor of the hidden neurons of the LSTM. If this is null, the layer uses 0.
TensorFloat	initialC	The optional initial values tensor of the cells of the LSTM. If this is null, the layer uses 0.
TensorFloat	P	The optional weight tensor for the peepholes of the LSTM. If this is null, the layer uses 0.
TensorFloat	Y	The output tensor to be computed and filled with the concatenated intermediate output values of the hidden.
TensorFloat	Yh	The output tensor to be computed and filled with the last output value of the hidden.
TensorFloat	Yc	The output tensor to be computed and filled with the last output value of the cell.
RnnDirection	direction	The direction of the LSTM as an RnnDirection.
RnnActivation[]	activations	The activation functions of the LSTM as an array of RnnActivation.
float[]	activationAlpha	The alpha values of the activation functions of the LSTM.
float[]	activationBeta	The beta values of the activation functions of the LSTM.
bool	inputForget	Whether to forget the input values in the LSTM. If this is false, the layer couples the input and forget gates.
float	clip	The cell clip threshold of the LSTM.
RnnLayout	layout	The layout of the tensors as an RnnLayout.

LayerNormalization(TensorFloat, TensorFloat, TensorFloat, TensorFloat, float)

Computes the mean variance on the last dimension of the input tensor and normalizes it according to scale and bias tensors.

Declaration

public void LayerNormalization(TensorFloat X, TensorFloat S, TensorFloat B, TensorFloat O, float epsilon)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	S	The scale tensor.
TensorFloat	B	The bias tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	epsilon	The epsilon value the layer uses to avoid division by zero.

LeakyRelu(TensorFloat, TensorFloat, float)

Computes an output tensor by applying the element-wise LeakyRelu activation function: f(x) = x if x >= 0, otherwise f(x) = alpha * x.

Declaration

public void LeakyRelu(TensorFloat X, TensorFloat O, float alpha)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	alpha	The alpha value to use for the LeakyRelu activation function.

Less(TensorFloat, TensorFloat, TensorInt)

Performs an element-wise Less logical comparison operation: f(a, b) = 1 if a < b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Less(TensorFloat A, TensorFloat B, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Less(TensorInt, TensorInt, TensorInt)

Performs an element-wise Less logical comparison operation: f(a, b) = 1 if a < b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Less(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

LessOrEqual(TensorFloat, TensorFloat, TensorInt)

Performs an element-wise LessOrEqual logical comparison operation: f(a, b) = 1 if a <= b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void LessOrEqual(TensorFloat A, TensorFloat B, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

LessOrEqual(TensorInt, TensorInt, TensorInt)

Performs an element-wise LessOrEqual logical comparison operation: f(a, b) = 1 if a <= b, otherwise f(x) = 0.

This supports numpy-style broadcasting of input tensors.

Declaration

public void LessOrEqual(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Log(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Log math function: f(x) = log(x).

Declaration

public void Log(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

LogSoftmax(TensorFloat, TensorFloat, int)

Computes an output tensor by applying the LogSoftmax activation function along an axis: f(x, axis) = log(Softmax(x, axis)).

Declaration

public void LogSoftmax(TensorFloat X, TensorFloat O, int axis)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	axis	The axis along which to apply the LogSoftmax activation function.

MatMul(TensorFloat, TensorFloat, TensorFloat)

Performs a multi-dimensional matrix multiplication operation: f(a, b) = a x b.

Declaration

public void MatMul(TensorFloat X, TensorFloat Y, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The first input tensor.
TensorFloat	Y	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

MatMul2D(TensorFloat, TensorFloat, TensorFloat, bool, bool)

Performs a matrix multiplication operation with optional transposes: f(a, b) = a' x b'.

Declaration

public void MatMul2D(TensorFloat X, TensorFloat Y, TensorFloat O, bool xTranspose, bool yTranspose)

Parameters

Type	Name	Description
TensorFloat	X	The first input tensor.
TensorFloat	Y	The second input tensor.
TensorFloat	O	The output tensor.
bool	xTranspose	Whether to transpose the first input tensor before performing the matrix multiplication.
bool	yTranspose	Whether to transpose the second input tensor before performing the matrix multiplication.

Max(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Max math operation: f(a, b) = max(a, b).

This supports numpy-style broadcasting of input tensors.

Declaration

public void Max(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Max(TensorInt, TensorInt, TensorInt)

Performs an element-wise Max math operation: f(a, b) = max(a, b).

This supports numpy-style broadcasting of input tensors.

Declaration

public void Max(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

MaxPool(TensorFloat, TensorFloat, int[], int[], int[])

Calculates an output tensor by pooling the maximum values of the input tensor across its spatial dimensions according to the given pool and stride values.

Declaration

public void MaxPool(TensorFloat X, TensorFloat O, int[] kernelShape, int[] strides, int[] pads)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int[]	kernelShape	The size of the kernel along each spatial axis.
int[]	strides	The stride along each spatial axis.
int[]	pads	The lower and upper padding values for each spatial dimension. For example, [pad_left, pad_right] for 1D, or [pad_top, pad_bottom, pad_left, pad_right] for 2D.

MemClear(Tensor)

Sets the entries of a tensor to 0.

Declaration

public void MemClear(Tensor O)

Parameters

Type	Name	Description
Tensor	O	The output tensor to be computed and filled.

MemCopy(Tensor, Tensor)

Creates a copy of a given input tensor with the same shape and values.

Declaration

public void MemCopy(Tensor X, Tensor O)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.

MemCopyStride(Tensor, Tensor, int, int, int, int, int, int)

Copy blocks of values from X to O, we copy 'count' blocks each of length 'length' values with initial offsets given by 'offsetX', 'offsetO' and with strides given by 'strideX', 'strideO'

Declaration

public void MemCopyStride(Tensor X, Tensor O, int strideX, int strideO, int length, int count, int offsetX, int offsetO)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
int	strideX	The stride of the blocks in the input tensor.
int	strideO	The stride of the blocks in the output tensor.
int	length	The number of elements in each block.
int	count	The number of blocks to copy.
int	offsetX	The first index to copy from in the input tensor.
int	offsetO	The first index to copy to in the output tensor.

MemSet(TensorFloat, float)

Sets the entries of a tensor to a given fill value.

Declaration

public void MemSet(TensorFloat O, float value)

Parameters

Type	Name	Description
TensorFloat	O	The output tensor to be computed and filled.
float	value	The fill value.

MemSet(TensorInt, int)

Sets the entries of a tensor to a given fill value.

Declaration

public void MemSet(TensorInt O, int value)

Parameters

Type	Name	Description
TensorInt	O	The output tensor to be computed and filled.
int	value	The fill value.

Min(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Min math operation: f(a, b) = min(a, b).

This supports numpy-style broadcasting of input tensors.

Declaration

public void Min(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Min(TensorInt, TensorInt, TensorInt)

Performs an element-wise Min math operation: f(a, b) = min(a, b).

This supports numpy-style broadcasting of input tensors.

Declaration

public void Min(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Mod(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Mod math operation: f(a, b) = a % b.

The sign of the remainder is the same as the sign of the divisor, as in Python.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Mod(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Mod(TensorInt, TensorInt, TensorInt)

Performs an element-wise Mod math operation: f(a, b) = a % b.

The sign of the remainder is the same as the sign of the divisor, as in Python.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Mod(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Mul(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Mul math operation: f(a, b) = a * b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Mul(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Mul(TensorInt, TensorInt, TensorInt)

Performs an element-wise Mul math operation: f(a, b) = a * b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Mul(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Neg(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Neg math function: f(x) = -x.

Declaration

public void Neg(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Neg(TensorInt, TensorInt)

Computes an output tensor by applying the element-wise Neg math function: f(x) = -x.

Declaration

public void Neg(TensorInt X, TensorInt O)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

NonMaxSuppression(TensorFloat, TensorFloat, TensorInt, int, float, float, CenterPointBox)

Performs NonMaxSuppression on boxes with scores.

Declaration

public void NonMaxSuppression(TensorFloat boxes, TensorFloat scores, TensorInt O, int maxOutputBoxesPerClass, float iouThreshold, float scoreThreshold, CenterPointBox centerPointBox)

Parameters

Type	Name	Description
TensorFloat	boxes	The boxes input tensor containing the position and dimensions of the boxes for each batch.
TensorFloat	scores	The scores input tensor containing the score for each box per class per batch.
TensorInt	O	The output tensor to be computed and filled (and truncated) with the batch, class and index of the boxes in decreasing order of score.
int	maxOutputBoxesPerClass	The maximum number of output boxes per class.
float	iouThreshold	Boxes with intersect-over-union with a selected box above this threshold are discarded.
float	scoreThreshold	Boxes with a score below this threshold are discarded.
CenterPointBox	centerPointBox	The types of the box coordinates, either [x1, y1, x2, y2] or [x, y, w, h].

Not(TensorInt, TensorInt)

Performs an element-wise Not logical operation: f(x) = ~x.

Declaration

public void Not(TensorInt X, TensorInt O)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

OneHot(TensorInt, TensorFloat, int, int, float, float)

Generates a one-hot tensor with a given depth, indices and on and off values.

Declaration

public void OneHot(TensorInt X, TensorFloat O, int axis, int depth, float offValue, float onValue)

Parameters

Type	Name	Description
TensorInt	X
TensorFloat	O	The output tensor to be computed and filled.
int	axis	The axis along which the operation adds the one-hot representation.
int	depth	The depth of the one-hot tensor.
float	offValue	The value to use for an off element.
float	onValue	The value to use for an on element.

OneHot(TensorInt, TensorInt, int, int, int, int)

Generates a one-hot tensor with a given depth, indices and on and off values.

Declaration

public void OneHot(TensorInt X, TensorInt O, int axis, int depth, int offValue, int onValue)

Parameters

Type	Name	Description
TensorInt	X
TensorInt	O	The output tensor to be computed and filled.
int	axis	The axis along which the operation adds the one-hot representation.
int	depth	The depth of the one-hot tensor.
int	offValue	The value to use for an off element.
int	onValue	The value to use for an on element.

Or(TensorInt, TensorInt, TensorInt)

Performs an element-wise Or logical operation: f(a, b) = a | b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Or(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

PRelu(TensorFloat, TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise PRelu activation function: f(x) = x if x >= 0, otherwise f(x) = slope * x.

Declaration

public void PRelu(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A
TensorFloat	B
TensorFloat	O	The output tensor to be computed and filled.

Pad(TensorFloat, TensorFloat, ReadOnlySpan<int>, PadMode, float)

Calculates the output tensor by adding padding to the input tensor according to the given padding values, mode and constant value.

Declaration

public void Pad(TensorFloat X, TensorFloat O, ReadOnlySpan<int> pad, PadMode padMode, float constant)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	pad	The lower and upper padding values for each dimension.
PadMode	padMode	The PadMode to use when padding.
float	constant	The constant value to fill with.

Pad(TensorInt, TensorInt, ReadOnlySpan<int>, PadMode, int)

Calculates the output tensor by adding padding to the input tensor according to the given padding values, mode and constant value.

Declaration

public void Pad(TensorInt X, TensorInt O, ReadOnlySpan<int> pad, PadMode padMode, int constant)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	pad	The lower and upper padding values for each dimension.
PadMode	padMode	The PadMode to use when padding.
int	constant	The constant value to fill with.

PinToDevice(Tensor, bool)

Pins and returns a tensor using this backend.

Declaration

public Tensor PinToDevice(Tensor X, bool clearOnInit = false)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
bool	clearOnInit	Whether to initialize the backend data. The default value is true.

Returns

Type	Description
Tensor	The pinned input tensor.

Pow(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Pow math operation: f(a, b) = pow(a, b).

This supports numpy-style broadcasting of input tensors.

Declaration

public void Pow(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Pow(TensorFloat, TensorInt, TensorFloat)

Performs an element-wise Pow math operation: f(a, b) = pow(a, b).

This supports numpy-style broadcasting of input tensors.

Declaration

public void Pow(TensorFloat A, TensorInt B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

RandomNormal(TensorFloat, float, float, int?)

Generates an output tensor of a given shape with random values in a normal distribution with given mean and scale, and an optional seed value.

Declaration

public void RandomNormal(TensorFloat O, float mean, float scale, int? seed)

Parameters

Type	Name	Description
TensorFloat	O	The output tensor to be computed and filled.
float	mean	The mean of the normal distribution to use to generate the output.
float	scale	The standard deviation of the normal distribution to use to generate the output.
int?	seed	The optional seed to use for the random number generation. If this is null the operation generates a seed using System.Random().

RandomUniform(TensorFloat, float, float, int?)

Generates an output tensor of a given shape with random values in a uniform distribution between a given low and high, and an optional seed value.

Declaration

public void RandomUniform(TensorFloat O, float low, float high, int? seed)

Parameters

Type	Name	Description
TensorFloat	O	The output tensor to be computed and filled.
float	low	The lower end of the interval of the uniform distribution to use to generate the output.
float	high	The upper end of the interval of the uniform distribution to use to generate the output.
int?	seed	The optional seed to use for the random number generation. If this is null the operation generates a seed using System.Random().

Range(TensorFloat, float, float)

Generates a 1D output tensor where the values form an arithmetic progression defined by the start and delta values.

Declaration

public void Range(TensorFloat O, float start, float delta)

Parameters

Type	Name	Description
TensorFloat	O	The output tensor to be computed and filled.
float	start	The first value in the range.
float	delta	The delta between subsequent values in the range.

Range(TensorInt, int, int)

Generates a 1D output tensor where the values form an arithmetic progression defined by the start and delta values.

Declaration

public void Range(TensorInt O, int start, int delta)

Parameters

Type	Name	Description
TensorInt	O	The output tensor to be computed and filled.
int	start	The first value in the range.
int	delta	The delta between subsequent values in the range.

Reciprocal(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Reciprocal math function: f(x) = 1 / x.

Declaration

public void Reciprocal(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

ReduceL1(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceL1 operation: f(x1, x2 ... xn) = |x1| + |x2| + ... + |xn|.

Declaration

public void ReduceL1(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceL1(TensorInt, TensorInt, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceL1 operation: f(x1, x2 ... xn) = |x1| + |x2| + ... + |xn|.

Declaration

public void ReduceL1(TensorInt X, TensorInt O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceL2(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceL2 operation: f(x1, x2 ... xn) = sqrt(x1² + x2² + ... + xn²).

Declaration

public void ReduceL2(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceLogSum(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceLogSum operation: f(x1, x2 ... xn) = log(x1 + x2 + ... + xn).

Declaration

public void ReduceLogSum(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceLogSumExp(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceLogSumExp operation: f(x1, x2 ... xn) = log(e^x1 + e^x2 + ... + e^xn).

Declaration

public void ReduceLogSumExp(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceMax(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceMax operation: f(x1, x2 ... xn) = max(x1, x2, ... , xn).

Declaration

public void ReduceMax(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceMax(TensorInt, TensorInt, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceMean operation: f(x1, x2 ... xn) = max(x1, x2, ... , xn).

Declaration

public void ReduceMax(TensorInt X, TensorInt O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceMean(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceMean operation: f(x1, x2 ... xn) = (x1 + x2 + ... + xn) / n.

Declaration

public void ReduceMean(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceMin(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceMin operation: f(x1, x2 ... xn) = min(x1, x2, ... , xn).

Declaration

public void ReduceMin(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceMin(TensorInt, TensorInt, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceMin operation: f(x1, x2 ... xn) = min(x1, x2, ... , xn).

Declaration

public void ReduceMin(TensorInt X, TensorInt O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceProd(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceProd operation: f(x1, x2 ... xn) = x1 * x2 * ... * xn.

Declaration

public void ReduceProd(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceProd(TensorInt, TensorInt, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceProd operation: f(x1, x2 ... xn) = x1 * x2 * ... * xn.

Declaration

public void ReduceProd(TensorInt X, TensorInt O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceSum(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceSum operation: f(x1, x2 ... xn) = x1 + x2 + ... + xn.

Declaration

public void ReduceSum(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceSum(TensorInt, TensorInt, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceSum operation: f(x1, x2 ... xn) = x1 + x2 + ... + xn.

Declaration

public void ReduceSum(TensorInt X, TensorInt O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceSumExp(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Declaration

public void ReduceSumExp(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X
TensorFloat	O
ReadOnlySpan<int>	axes

ReduceSumSquare(TensorFloat, TensorFloat, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceSumSquare operation: f(x1, x2 ... xn) = x1² + x2² + ... + xn².

Declaration

public void ReduceSumSquare(TensorFloat X, TensorFloat O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

ReduceSumSquare(TensorInt, TensorInt, ReadOnlySpan<int>)

Reduces an input tensor along the given axes using the ReduceSumSquare operation: f(x1, x2 ... xn) = x1² + x2² + ... + xn².

Declaration

public void ReduceSumSquare(TensorInt X, TensorInt O, ReadOnlySpan<int> axes)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	axes	The axes along which to reduce.

Relu(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Relu activation function: f(x) = max(0, x).

Declaration

public void Relu(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Relu6(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Relu6 activation function: f(x) = clamp(x, 0, 6).

Declaration

public void Relu6(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Reshape(Tensor, Tensor)

Calculates an output tensor by copying the data from the input tensor and using a given shape. The data from the input tensor is unchanged.

Declaration

public void Reshape(Tensor X, Tensor O)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.

Resize(TensorFloat, TensorFloat, ReadOnlySpan<float>, InterpolationMode, NearestMode, CoordTransformMode)

Calculates an output tensor by resampling the input tensor along the spatial dimensions with given scales.

Declaration

public void Resize(TensorFloat X, TensorFloat O, ReadOnlySpan<float> scale, InterpolationMode interpolationMode, NearestMode nearestMode, CoordTransformMode coordTransformMode)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
ReadOnlySpan<float>	scale	The factor to scale each dimension by.
InterpolationMode	interpolationMode	The InterpolationMode to use for the operation.
NearestMode	nearestMode	The NearestMode to use for the operation when using InterpolationMode.NearestMode.
CoordTransformMode	coordTransformMode	The CoordTransformMode to use for the operation.

RoiAlign(TensorFloat, TensorFloat, TensorInt, TensorFloat, RoiPoolingMode, int, int, int, float)

Calculates an output tensor by pooling the input tensor across each region of interest given by the rois tensor.

Declaration

public void RoiAlign(TensorFloat X, TensorFloat rois, TensorInt indices, TensorFloat O, RoiPoolingMode mode, int outputHeight, int outputWidth, int samplingRatio, float spatialScale)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	rois	The region of interest input tensor.
TensorInt	indices	The indices input tensor.
TensorFloat	O	The output tensor to be computed and filled.
RoiPoolingMode	mode	The pooling mode of the operation as an RoiPoolingMode.
int	outputHeight	The height of the output tensor.
int	outputWidth	The width of the output tensor.
int	samplingRatio	The number of sampling points in the interpolation grid used to compute the output value of each pooled output bin.
float	spatialScale	The multiplicative spatial scale factor used to translate coordinates from their input spatial scale to the scale used when pooling.

Round(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Round math function: f(x) = round(x).

If the fractional part is equal to 0.5, rounds to the nearest even integer.

Declaration

public void Round(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

ScalarMad(TensorFloat, TensorFloat, float, float)

Performs an element-wise Mad math operation: multiplies and adds bias to a tensor: f(T, s, b) = s * T + b.

Declaration

public void ScalarMad(TensorFloat X, TensorFloat O, float s, float b)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	s	Input scalar for multiplication.
float	b	Input bias for addition.

ScalarMad(TensorInt, TensorInt, int, int)

Performs an element-wise Mad math operation: multiplies and adds bias to a tensor: f(T, s, b) = s * T + b.

Declaration

public void ScalarMad(TensorInt X, TensorInt O, int s, int b)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.
int	s	Input scalar for multiplication.
int	b	Input bias for addition.

ScaleBias(TensorFloat, TensorFloat, TensorFloat, TensorFloat)

Computes the output tensor with an element-wise ScaleBias function: f(x, s, b) = x * s + b.

Declaration

public void ScaleBias(TensorFloat X, TensorFloat S, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	S	The scale tensor.
TensorFloat	B	The bias tensor.
TensorFloat	O	The output tensor to be computed and filled.

ScatterElements(Tensor, TensorInt, Tensor, Tensor, int, ScatterReductionMode)

Copies the input tensor and updates values at indexes specified by the indices tensor with values specified by the updates tensor along a given axis.

ScatterElements updates the values depending on the reduction mode used.

Declaration

public void ScatterElements(Tensor X, TensorInt indices, Tensor updates, Tensor O, int axis, ScatterReductionMode reduction)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
TensorInt	indices	The indices tensor.
Tensor	updates	The updates tensor.
Tensor	O	The output tensor to be computed and filled.
int	axis	The axis on which to perform the scatter.
ScatterReductionMode	reduction	The reduction mode used to update the values as a ScatterReductionMode.

ScatterND(TensorFloat, TensorInt, TensorFloat, TensorFloat, ScatterReductionMode)

Copies the input tensor and updates values at indexes specified by the indices tensor with values specified by the updates tensor.

ScatterND updates the values depending on the reduction mode used.

Declaration

public void ScatterND(TensorFloat X, TensorInt indices, TensorFloat updates, TensorFloat O, ScatterReductionMode reduction)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorInt	indices	The indices tensor.
TensorFloat	updates	The updates tensor.
TensorFloat	O	The output tensor to be computed and filled.
ScatterReductionMode	reduction	The reduction mode used to update the values as a ScatterReductionMode.

ScatterND(TensorInt, TensorInt, TensorInt, TensorInt, ScatterReductionMode)

Copies the input tensor and updates values at indexes specified by the indices tensor with values specified by the updates tensor.

ScatterND updates the values depending on the reduction mode used.

Declaration

public void ScatterND(TensorInt X, TensorInt indices, TensorInt updates, TensorInt O, ScatterReductionMode reduction)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	indices	The indices tensor.
TensorInt	updates	The updates tensor.
TensorInt	O	The output tensor to be computed and filled.
ScatterReductionMode	reduction	The reduction mode used to update the values as a ScatterReductionMode.

Selu(TensorFloat, TensorFloat, float, float)

Computes an output tensor by applying the element-wise Selu activation function: f(x) = gamma * x if x >= 0, otherwise f(x) = (alpha * e^x - alpha).

Declaration

public void Selu(TensorFloat X, TensorFloat O, float alpha, float gamma)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	alpha	The alpha value to use for the Selu activation function.
float	gamma	The alpha value to use for the Selu activation function.

Shrink(TensorFloat, TensorFloat, float, float)

Computes an output tensor by applying the element-wise Shrink activation function: f(x) = x + bias if x < lambd. f(x) = x - bias if x > lambd. Otherwise f(x) = 0.

Declaration

public void Shrink(TensorFloat X, TensorFloat O, float bias, float lambd)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	bias	The bias value to use for the Shrink activation function.
float	lambd	The lambda value to use for the Shrink activation function.

Sigmoid(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Sigmoid activation function: f(x) = 1/(1 + e^(-x)).

Declaration

public void Sigmoid(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Sign(TensorFloat, TensorFloat)

Performs an element-wise Sign math operation: f(x) = 1 if x > 0. f(x) = -1 if x < 0. Otherwise f(x) = 0.

Declaration

public void Sign(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Sign(TensorInt, TensorInt)

Performs an element-wise Sign math operation: f(x) = 1 if x > 0. f(x) = -1 if x < 0. Otherwise f(x) = 0.

Declaration

public void Sign(TensorInt X, TensorInt O)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

Sin(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Sin trigonometric function: f(x) = sin(x).

Declaration

public void Sin(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

SinglePassLSTM(TensorFloat, TensorFloat, TensorFloat, TensorFloat, TensorInt, TensorFloat, TensorFloat, TensorFloat, TensorFloat, RnnActivation[], float[], float[], bool, float, bool, int, RnnLayout)

Declaration

protected void SinglePassLSTM(TensorFloat X, TensorFloat W, TensorFloat R, TensorFloat B, TensorInt sequenceLens, TensorFloat P, TensorFloat Y, TensorFloat Y_h, TensorFloat Y_c, RnnActivation[] activations, float[] activationAlpha, float[] activationBeta, bool inputForget, float clip, bool isReverse, int dirIndex, RnnLayout layout)

Parameters

Type	Name	Description
TensorFloat	X
TensorFloat	W
TensorFloat	R
TensorFloat	B
TensorInt	sequenceLens
TensorFloat	P
TensorFloat	Y
TensorFloat	Y_h
TensorFloat	Y_c
RnnActivation[]	activations
float[]	activationAlpha
float[]	activationBeta
bool	inputForget
float	clip
bool	isReverse
int	dirIndex
RnnLayout	layout

Sinh(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Sinh trigonometric function: f(x) = sinh(x).

Declaration

public void Sinh(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Slice(Tensor, Tensor, ReadOnlySpan<int>, ReadOnlySpan<int>, ReadOnlySpan<int>)

Calculates an output tensor by slicing the input tensor along given axes with given starts, ends, and steps.

Declaration

public void Slice(Tensor X, Tensor O, ReadOnlySpan<int> starts, ReadOnlySpan<int> axes, ReadOnlySpan<int> steps)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	starts	The start index along each axis.
ReadOnlySpan<int>	axes	The axes along which to slice. If this is null, the layer slices all axes.
ReadOnlySpan<int>	steps	The step values for slicing. If this is null, the layer uses step size 1 throughout.

SliceSet(Tensor, Tensor, int, int, int)

Updates values at indexes specified by the slices defined by axis, start and step.

Declaration

public void SliceSet(Tensor X, Tensor O, int axis, int start, int step)

Parameters

Type	Name	Description
Tensor	X
Tensor	O	The output tensor to be computed and filled.
int	axis	The axes along which to slice.
int	start	The start index along the axis.
int	step	The step value for slicing.

SliceSet(Tensor, Tensor, Tensor, ReadOnlySpan<int>, ReadOnlySpan<int>, ReadOnlySpan<int>)

Copies the input tensor and updates values at indexes specified by the slices defined by axes, starts, ends, and steps.

Declaration

public void SliceSet(Tensor X, Tensor values, Tensor O, ReadOnlySpan<int> starts, ReadOnlySpan<int> axes, ReadOnlySpan<int> steps)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	values	The values tensor.
Tensor	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	starts	The start index along each axis.
ReadOnlySpan<int>	axes	The axes along which to slice. If this is null, the layer slices all axes.
ReadOnlySpan<int>	steps	The step values for slicing. If this is null, the layer uses step size 1 throughout.

Softmax(TensorFloat, TensorFloat, int)

Computes an output tensor by applying the Softmax activation function along an axis: f(x, axis) = exp(X) / ReduceSum(exp(X), axis).

Declaration

public void Softmax(TensorFloat X, TensorFloat O, int axis)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	axis	The axis along which to apply the Softmax activation function.

Softplus(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Softplus activation function: f(x) = ln(e^x + 1).

Declaration

public void Softplus(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Softsign(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Softsign activation function: f(x) = x/(|x| + 1).

Declaration

public void Softsign(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

SpaceToDepth(TensorFloat, TensorFloat, int)

Computes the output tensor by permuting data from blocks of spatial data into depth.

Declaration

public void SpaceToDepth(TensorFloat X, TensorFloat O, int blocksize)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
int	blocksize	The size of the blocks to move the depth data into.

Split(Tensor, Tensor, int, int)

Calculates an output tensor by splitting the input tensor along a given axis between start and end.

Declaration

public void Split(Tensor X, Tensor O, int axis, int start)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
int	axis	The axis along which to split the input tensor.
int	start	The inclusive start value for the split.

Sqrt(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Sqrt math function: f(x) = sqrt(x).

Declaration

public void Sqrt(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Square(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Square math function: f(x) = x * x.

Declaration

public void Square(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Square(TensorInt, TensorInt)

Computes an output tensor by applying the element-wise Square math function: f(x) = x * x.

Declaration

public void Square(TensorInt X, TensorInt O)

Parameters

Type	Name	Description
TensorInt	X	The input tensor.
TensorInt	O	The output tensor to be computed and filled.

Sub(TensorFloat, TensorFloat, TensorFloat)

Performs an element-wise Sub math operation: f(a, b) = a - b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Sub(TensorFloat A, TensorFloat B, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	A	The first input tensor.
TensorFloat	B	The second input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Sub(TensorInt, TensorInt, TensorInt)

Performs an element-wise Sub math operation: f(a, b) = a - b.

This supports numpy-style broadcasting of input tensors.

Declaration

public void Sub(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Swish(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Swish activation function: f(x) = sigmoid(x) * x = x / (1 + e^{-x}).

Declaration

public void Swish(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Tan(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Tan trigonometric function: f(x) = tan(x).

Declaration

public void Tan(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

Tanh(TensorFloat, TensorFloat)

Computes an output tensor by applying the element-wise Tanh activation function: f(x) = tanh(x).

Declaration

public void Tanh(TensorFloat X, TensorFloat O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.

ThresholdedRelu(TensorFloat, TensorFloat, float)

Computes an output tensor by applying the element-wise ThresholdedRelu activation function: f(x) = x if x > alpha, otherwise f(x) = 0.

Declaration

public void ThresholdedRelu(TensorFloat X, TensorFloat O, float alpha)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	O	The output tensor to be computed and filled.
float	alpha	The alpha value to use for the ThresholdedRelu activation function.

Tile(Tensor, Tensor, ReadOnlySpan<int>)

Calculates an output tensor by repeating the input layer a given number of times along each axis.

Declaration

public void Tile(Tensor X, Tensor O, ReadOnlySpan<int> repeats)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	repeats	The number of times to tile the input tensor along each axis.

TopK(TensorFloat, TensorFloat, TensorInt, int, int, bool)

Calculates the top-K largest or smallest elements of an input tensor along a given axis.

Declaration

public void TopK(TensorFloat X, TensorFloat values, TensorInt indices, int k, int axis, bool largest)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	values	The output tensor to be computed and filled with the top K values from the input tensor.
TensorInt	indices	The output tensor to be computed and filled with the corresponding input tensor indices for the top K values from the input tensor.
int	k	The number of elements to calculate.
int	axis	The axis along which to perform the top-K operation.
bool	largest	Whether to calculate the top-K largest elements. If this is false, the layer calculates the top-K smallest elements.

TopP(TensorFloat, TensorFloat, TensorInt)

Computes the index of the element which cumulative sum until said element is >= than a random value

Declaration

public void TopP(TensorFloat X, TensorFloat random, TensorInt O)

Parameters

Type	Name	Description
TensorFloat	X	The input tensor.
TensorFloat	random	The probability values used for the exit criteria.
TensorInt	O	The output tensor to be computed and filled.

Transpose(Tensor, Tensor)

Calculates an output tensor by reversing the dimensions of the input tensor.

Declaration

public void Transpose(Tensor X, Tensor O)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.

Transpose(Tensor, Tensor, ReadOnlySpan<int>)

Calculates an output tensor by permuting the axes and data of the input tensor according to the given permutations.

Declaration

public void Transpose(Tensor X, Tensor O, ReadOnlySpan<int> permutations)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
ReadOnlySpan<int>	permutations	The axes to sample the output tensor from in the input tensor.

Tril(Tensor, Tensor, int)

Computes the output tensor by retaining the lower triangular values from an input matrix or matrix batch and setting the other values to zero.

Declaration

public void Tril(Tensor X, Tensor O, int k)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
int	k	The offset from the diagonal to keep.

Triu(Tensor, Tensor, int)

Computes the output tensor by retaining the upper triangular values from an input matrix or matrix batch and setting the other values to zero.

Declaration

public void Triu(Tensor X, Tensor O, int k)

Parameters

Type	Name	Description
Tensor	X	The input tensor.
Tensor	O	The output tensor to be computed and filled.
int	k	The offset from the diagonal to exclude.

Where(TensorInt, Tensor, Tensor, Tensor)

Performs an element-wise Where logical operation: f(condition, a, b) = a if condition is true, otherwise f(condition, a, b) = b.

Declaration

public void Where(TensorInt C, Tensor A, Tensor B, Tensor O)

Parameters

Type	Name	Description
TensorInt	C	The condition tensor.
Tensor	A	The first input tensor.
Tensor	B	The second input tensor.
Tensor	O	The output tensor to be computed and filled.

Xor(TensorInt, TensorInt, TensorInt)

Performs an element-wise Xor logical operation: f(a) = a ^ b.

Declaration

public void Xor(TensorInt A, TensorInt B, TensorInt O)

Parameters

Type	Name	Description
TensorInt	A	The first input tensor.
TensorInt	B	The second input tensor.
TensorInt	O	The output tensor to be computed and filled.

Implements

IBackend

IDisposable