Interface IOps

Interfaces for backend implementers see ModelBuilder.cs for detail on layers.

Namespace: Unity.Barracuda

Syntax

public interface IOps

Methods

Abs(Tensor)

Abs

Declaration

Tensor Abs(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Acos(Tensor)

Acos

Declaration

Tensor Acos(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Acosh(Tensor)

Acosh

Declaration

Tensor Acosh(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Add(Tensor[])

Add tensors together

Declaration

Tensor Add(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

ArgMax(Tensor, Int32)

ArgMax

Declaration

Tensor ArgMax(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

ArgMin(Tensor, Int32)

ArgMax

Declaration

Tensor ArgMin(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

Asin(Tensor)

Asin

Declaration

Tensor Asin(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Asinh(Tensor)

Asinh

Declaration

Tensor Asinh(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Atan(Tensor)

Atan

Declaration

Tensor Atan(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Atanh(Tensor)

Atanh

Declaration

Tensor Atanh(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

AvgPool2D(Tensor, Int32[], Int32[], Int32[])

2D average pooling

Declaration

Tensor AvgPool2D(Tensor x, int[] pool, int[] stride, int[] pad)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pool	pooling
Int32[]	stride	stride
Int32[]	pad	padding

Returns

Type	Description
Tensor	output Tensor

Border2D(Tensor, Int32[], Single)

2D border padding

Declaration

Tensor Border2D(Tensor x, int[] pad, float borderValue)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pad	padding
Single	borderValue	border value

Returns

Type	Description
Tensor	output Tensor

Border3D(Tensor, Int32[], Single)

3D border padding

Declaration

Tensor Border3D(Tensor x, int[] pad, float borderValue)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pad	padding
Single	borderValue	border value

Returns

Type	Description
Tensor	output Tensor

Ceil(Tensor)

Ceil

Declaration

Tensor Ceil(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Clip(Tensor, Single, Single)

Clip

Declaration

Tensor Clip(Tensor x, float min, float max)

Parameters

Type	Name	Description
Tensor	x	input
Single	min	min value
Single	max	max value

Returns

Type	Description
Tensor	output Tensor

Concat(Tensor[], Int32)

Concatenate tensors across axis

Declaration

Tensor Concat(Tensor[] tensors, int axis)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

ConstantOfShape(TensorShape, Single)

Creates a constant of shape input

Declaration

Tensor ConstantOfShape(TensorShape X, float value = 0F)

Parameters

Type	Name	Description
TensorShape	X	input shape
Single	value	value

Returns

Type	Description
Tensor	output Tensor

Conv2D(Tensor, Tensor, Tensor, Int32[], Int32[], Layer.FusedActivation)

2D convolution

Declaration

Tensor Conv2D(Tensor x, Tensor k, Tensor b, int[] stride, int[] pad, Layer.FusedActivation fusedActivation)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	k	kernel
Tensor	b	bias
Int32[]	stride	stride
Int32[]	pad	padding
Layer.FusedActivation	fusedActivation	fused activation type

Returns

Type	Description
Tensor	output Tensor

Conv2DTrans(Tensor, Tensor, Tensor, Int32[], Int32[], Int32[], Layer.FusedActivation)

Transpose 2D convolution

Declaration

Tensor Conv2DTrans(Tensor x, Tensor k, Tensor b, int[] stride, int[] pad, int[] outputAdjustment, Layer.FusedActivation fusedActivation)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	k	kernel
Tensor	b	bias
Int32[]	stride	stride
Int32[]	pad	padding
Int32[]	outputAdjustment	output adjustments
Layer.FusedActivation	fusedActivation	fused activation type

Returns

Type	Description
Tensor	output Tensor

Conv3D(Tensor, Tensor, Tensor, Int32[], Int32[], Layer.FusedActivation)

3D convolution

Declaration

Tensor Conv3D(Tensor x, Tensor k, Tensor b, int[] stride, int[] pad, Layer.FusedActivation fusedActivation)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	k	kernel
Tensor	b	bias
Int32[]	stride	stride
Int32[]	pad	padding
Layer.FusedActivation	fusedActivation	fused activation type

Returns

Type	Description
Tensor	output Tensor

Copy(Tensor)

Copy

Declaration

Tensor Copy(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Cos(Tensor)

Cos

Declaration

Tensor Cos(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Cosh(Tensor)

Cosh

Declaration

Tensor Cosh(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Dense(Tensor, Tensor, Tensor, Layer.FusedActivation)

Dense layer (matrix multiplication) o = x ⨯ w + b

Declaration

Tensor Dense(Tensor x, Tensor w, Tensor b, Layer.FusedActivation fusedActivation)

Parameters

Type	Name	Description
Tensor	x	x argument
Tensor	w	w argument
Tensor	b	bias argument
Layer.FusedActivation	fusedActivation	fused activation type

Returns

Type	Description
Tensor	output Tensor

Dense3(Tensor, Tensor, Tensor)

rank3 Dense layer (matrix multiplication) o = x ⨯ w + b O: N,,W,C / X: N,,W,C / W:N,,,C / B:N,,,_

Declaration

Tensor Dense3(Tensor x, Tensor w, Tensor b)

Parameters

Type	Name	Description
Tensor	x	x argument (rank3)
Tensor	w	w argument (rank2)
Tensor	b	bias argument (rank1)

Returns

Type	Description
Tensor	output Tensor

DepthToSpace(Tensor, Int32[], Layer.DepthToSpaceMode)

Depth to space

Declaration

Tensor DepthToSpace(Tensor x, int[] scale, Layer.DepthToSpaceMode mode)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	scale	scale
Layer.DepthToSpaceMode	mode	mode

Returns

Type	Description
Tensor	output Tensor

DepthwiseConv2D(Tensor, Tensor, Tensor, Int32[], Int32[], Layer.FusedActivation)

Depthwise 2D convolution

Declaration

Tensor DepthwiseConv2D(Tensor x, Tensor k, Tensor b, int[] stride, int[] pad, Layer.FusedActivation fusedActivation)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	k	kernel
Tensor	b	bias
Int32[]	stride	stride
Int32[]	pad	padding
Layer.FusedActivation	fusedActivation	fused activation type

Returns

Type	Description
Tensor	output Tensor

Div(Tensor[])

Divide tensors o = tensors[0] / tensors[1] / ... / tensors[N-1]

Declaration

Tensor Div(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

Dropout(Tensor, Single)

Dropout

Declaration

Tensor Dropout(Tensor x, float alpha)

Parameters

Type	Name	Description
Tensor	x	input
Single	alpha	alpha

Returns

Type	Description
Tensor	output Tensor

Elu(Tensor, Single)

ELU

Declaration

Tensor Elu(Tensor x, float alpha)

Parameters

Type	Name	Description
Tensor	x	input
Single	alpha	alpha

Returns

Type	Description
Tensor	output Tensor

Equal(Tensor, Tensor)

Equal

Declaration

Tensor Equal(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a == b

Exp(Tensor)

Exponent e^x

Declaration

Tensor Exp(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Expand(Tensor, TensorShape)

Expand

Declaration

Tensor Expand(Tensor x, TensorShape shape)

Parameters

Type	Name	Description
Tensor	x	input
TensorShape	shape	new shape

Returns

Type	Description
Tensor	output Tensor

Flatten(Tensor)

Flatten

Declaration

Tensor Flatten(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Floor(Tensor)

Floor

Declaration

Tensor Floor(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Gather(Tensor[], Int32)

Gather

Declaration

Tensor Gather(Tensor[] tensors, int axis)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

GlobalAvgPool2D(Tensor)

2D global average pooling

Declaration

Tensor GlobalAvgPool2D(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

GlobalAvgVariancePool2D(Tensor)

2D global average variance pooling

Declaration

Tensor GlobalAvgVariancePool2D(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

GlobalMaxPool2D(Tensor)

2D global max pooling

Declaration

Tensor GlobalMaxPool2D(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Greater(Tensor, Tensor)

Greater

Declaration

Tensor Greater(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a > b

GreaterEqual(Tensor, Tensor)

Greater or equal

Declaration

Tensor GreaterEqual(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a >= b

LeakyRelu(Tensor, Single)

Leaky ReLU

Declaration

Tensor LeakyRelu(Tensor x, float alpha)

Parameters

Type	Name	Description
Tensor	x	input
Single	alpha	alpha

Returns

Type	Description
Tensor	output Tensor

Less(Tensor, Tensor)

Less

Declaration

Tensor Less(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a < b

LessEqual(Tensor, Tensor)

Less or equal

Declaration

Tensor LessEqual(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a < b

Log(Tensor)

Log

Declaration

Tensor Log(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

LogicalAnd(Tensor, Tensor)

And

Declaration

Tensor LogicalAnd(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a && b

LogicalNot(Tensor)

Not

Declaration

Tensor LogicalNot(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	Tensor with !x values

LogicalOr(Tensor, Tensor)

Declaration

Tensor LogicalOr(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a \|\| b

LogicalXor(Tensor, Tensor)

Xor

Declaration

Tensor LogicalXor(Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	a	left Tensor
Tensor	b	right Tensor

Returns

Type	Description
Tensor	Tensor with `true` where a xor b

LogSoftmax(Tensor)

LogSoftmax

Declaration

Tensor LogSoftmax(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

LRN(Tensor, Single, Single, Single, Int32)

LRN (Local Response Normalization)

Declaration

Tensor LRN(Tensor x, float alpha, float beta, float bias, int size)

Parameters

Type	Name	Description
Tensor	x	input
Single	alpha	alpha
Single	beta	beta
Single	bias	bias
Int32	size	size

Returns

Type	Description
Tensor	output Tensor

LSTM(Tensor, Tensor[], Tensor[], Tensor[], Tensor[], Tensor, Tensor)

LSTM

Declaration

Tensor[] LSTM(Tensor X, Tensor[] W, Tensor[] R, Tensor[] Wb, Tensor[] Rb, Tensor hidden, Tensor cell)

Parameters

Type	Name	Description
Tensor	X	The input sequences packed into one 3-D tensor.
Tensor[]	W	W parameter weight matrix for input, output, forget, and cell gates - W[iofc]
Tensor[]	R	R recurrence weight matrix for input, output, forget, and cell gates - R[iofc]
Tensor[]	Wb	W bias vectors for input, output, forget, and cell gates - Wb[iofc]
Tensor[]	Rb	R bias vectors for input, output, forget, and cell gates - Rb[iofc]
Tensor	hidden	Initial value of the hidden
Tensor	cell	Initial value of the cell

Returns

Type	Description
Tensor[]	[Y (concatenated intermediate values of the hidden), Y_h (final hidden), Y_c (final cell)]

MatMul(Tensor, Boolean, Tensor, Boolean)

Matrix multiplication o = x ⨯ y

Declaration

Tensor MatMul(Tensor x, bool xTranspose, Tensor y, bool yTranspose)

Parameters

Type	Name	Description
Tensor	x	left Tensor
Boolean	xTranspose	transposed `x` flag
Tensor	y	right Tensor
Boolean	yTranspose	transposed `y` flag

Returns

Type	Description
Tensor	output Tensor

MatMul(Tensor, Int32, Tensor, Int32)

Multidimensional Matrix multiplication o = x ⨯ y

Declaration

Tensor MatMul(Tensor x, int rankX, Tensor y, int rankY)

Parameters

Type	Name	Description
Tensor	x	left Tensor
Int32	rankX	rank of `x`
Tensor	y	right Tensor
Int32	rankY	rank of `y`

Returns

Type	Description
Tensor	output Tensor

Max(Tensor[])

Max

Declaration

Tensor Max(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

MaxPool2D(Tensor, Int32[], Int32[], Int32[])

2D max pooling

Declaration

Tensor MaxPool2D(Tensor x, int[] pool, int[] stride, int[] pad)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pool	pooling
Int32[]	stride	stride
Int32[]	pad	padding

Returns

Type	Description
Tensor	output Tensor

Mean(Tensor[])

Mean

Declaration

Tensor Mean(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

Min(Tensor[])

Min

Declaration

Tensor Min(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

Mul(Tensor[])

Multiply tensors together

Declaration

Tensor Mul(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

Multinomial(Tensor, Int32, Int32)

Multinomial random distribution

Declaration

Tensor Multinomial(Tensor x, int count, int seed)

Parameters

Type	Name	Description
Tensor	x	input
Int32	count	count
Int32	seed	seed

Returns

Type	Description
Tensor	output Tensor

Neg(Tensor)

Neg

Declaration

Tensor Neg(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

NonMaxSuppression(Tensor[], Int32, Single, Single, Int32)

Non max suppression tensors[0] - boxes, tensors[1] - scores

Declaration

Tensor NonMaxSuppression(Tensor[] tensors, int maxOutputBoxesPerClass, float iouThreshold, float scoreThreshold, int centerPointBox)

Parameters

Type	Name	Description
Tensor[]	tensors
Int32	maxOutputBoxesPerClass	max output boxes per class
Single	iouThreshold	IOU (Intersection Over Union) threshold
Single	scoreThreshold	score threshold
Int32	centerPointBox	center point box

Returns

Type	Description
Tensor	output Tensor

NonZero(Tensor)

Indices for non zero values

Declaration

Tensor NonZero(Tensor X)

Parameters

Type	Name	Description
Tensor	X	input

Returns

Type	Description
Tensor	output Tensor

Normalization(Tensor, Tensor, Tensor, Int32, Int32, Single, Layer.FusedActivation)

Normalization

Declaration

Tensor Normalization(Tensor x, Tensor s, Tensor b, int pool, int axis, float epsilon, Layer.FusedActivation fusedActivation)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	s	scale
Tensor	b	bias
Int32	pool	pooling
Int32	axis	axis
Single	epsilon	threshold
Layer.FusedActivation	fusedActivation	fused activation type

Returns

Type	Description
Tensor	output Tensor

OneHot(Tensor, Int32, Single, Single)

One hot

Declaration

Tensor OneHot(Tensor x, int depth, float onValue, float offValue)

Parameters

Type	Name	Description
Tensor	x	input
Int32	depth	output depth
Single	onValue	on value
Single	offValue	off value

Returns

Type	Description
Tensor	output Tensor

Pad2DEdge(Tensor, Int32[])

Edge padding

Declaration

Tensor Pad2DEdge(Tensor x, int[] pad)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pad	padding

Returns

Type	Description
Tensor	output Tensor

Pad2DReflect(Tensor, Int32[])

Reflection padding

Declaration

Tensor Pad2DReflect(Tensor x, int[] pad)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pad	padding

Returns

Type	Description
Tensor	output Tensor

Pad2DSymmetric(Tensor, Int32[])

Symmetric padding

Declaration

Tensor Pad2DSymmetric(Tensor x, int[] pad)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	pad	padding

Returns

Type	Description
Tensor	output Tensor

Pow(Tensor, Single)

Power

Declaration

Tensor Pow(Tensor x, float alpha)

Parameters

Type	Name	Description
Tensor	x	input
Single	alpha	alpha

Returns

Type	Description
Tensor	output Tensor

Pow(Tensor[])

Raise tensors to the power o =tensors[0] ^ tensors[1] ^ ... ^ tensors[N-1]

Declaration

Tensor Pow(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

PRelu(Tensor, Tensor)

PReLU

Declaration

Tensor PRelu(Tensor x, Tensor alpha)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	alpha	alpha

Returns

Type	Description
Tensor	output Tensor

Prepare(Tensor)

Prepares tensor for use

Declaration

Tensor Prepare(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	Tensor

RandomNormal(TensorShape, Single, Single, Int32)

Normal random distribution

Declaration

Tensor RandomNormal(TensorShape s, float mean, float scale, int seed)

Parameters

Type	Name	Description
TensorShape	s	shape
Single	mean	mean
Single	scale	scale
Int32	seed	seed

Returns

Type	Description
Tensor	output Tensor

RandomUniform(TensorShape, Single, Single, Int32)

Uniform random distribution

Declaration

Tensor RandomUniform(TensorShape s, float mean, float scale, int seed)

Parameters

Type	Name	Description
TensorShape	s	shape
Single	mean	mean
Single	scale	scale
Int32	seed	seed

Returns

Type	Description
Tensor	output Tensor

Reciprocal(Tensor)

Reciprocal (1/x)

Declaration

Tensor Reciprocal(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

ReduceMax(Tensor, Int32)

Reduce with max

Declaration

Tensor ReduceMax(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

ReduceMean(Tensor, Int32)

Reduce with mean

Declaration

Tensor ReduceMean(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

ReduceMin(Tensor, Int32)

Reduce with min

Declaration

Tensor ReduceMin(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

ReduceProd(Tensor, Int32)

Reduce with product

Declaration

Tensor ReduceProd(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

ReduceSum(Tensor, Int32)

Reduce with sum

Declaration

Tensor ReduceSum(Tensor x, int axis)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

Relu(Tensor)

ReLU

Declaration

Tensor Relu(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Relu6(Tensor)

ReLU capped to 6

Declaration

Tensor Relu6(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Resample2D(Tensor, Int32[], Boolean)

Resample 2D

Declaration

Tensor Resample2D(Tensor x, int[] size, bool bilinear)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	size	size
Boolean	bilinear	bilinear flag

Returns

Type	Description
Tensor	output Tensor

ResetAllocator(Boolean)

Reset internal allocator

Declaration

void ResetAllocator(bool keepCachedMemory = true)

Parameters

Type	Name	Description
Boolean	keepCachedMemory	keep cached memory flag

Reshape(Tensor, TensorShape)

Reshape

Declaration

Tensor Reshape(Tensor x, TensorShape shape)

Parameters

Type	Name	Description
Tensor	x	input
TensorShape	shape	new shape

Returns

Type	Description
Tensor	output Tensor

Round(Tensor)

Round to nearest integer. In case of halfs, round to nearest even integer

Declaration

Tensor Round(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

ScaleBias(Tensor, Tensor, Tensor)

Scale bias o = s * x + b, element wise

Declaration

Tensor ScaleBias(Tensor x, Tensor s, Tensor b)

Parameters

Type	Name	Description
Tensor	x	input
Tensor	s	scale
Tensor	b	bias

Returns

Type	Description
Tensor	output Tensor

Selu(Tensor, Single, Single)

SELU

Declaration

Tensor Selu(Tensor x, float alpha, float gamma)

Parameters

Type	Name	Description
Tensor	x	input
Single	alpha	alpha
Single	gamma	gamma

Returns

Type	Description
Tensor	output Tensor

Shape(Tensor, Int32)

Shape of the input

Declaration

Tensor Shape(Tensor X, int axis = null)

Parameters

Type	Name	Description
Tensor	X	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

Sigmoid(Tensor)

Sigmoid

Declaration

Tensor Sigmoid(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Sign(Tensor)

Sign

Declaration

Tensor Sign(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	Tensor with 1 if x > 0 -1 if < 0 and 0 if == 0 values

Sin(Tensor)

Sin

Declaration

Tensor Sin(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Sinh(Tensor)

Sinh

Declaration

Tensor Sinh(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Softmax(Tensor, Int32)

Softmax

Declaration

Tensor Softmax(Tensor x, int axis = 1)

Parameters

Type	Name	Description
Tensor	x	input
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

Softplus(Tensor)

Softplus

Declaration

Tensor Softplus(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

SpaceToDepth(Tensor, Int32[])

Space to depth

Declaration

Tensor SpaceToDepth(Tensor x, int[] scale)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	scale	scale

Returns

Type	Description
Tensor	output Tensor

Sqrt(Tensor)

Sqrt

Declaration

Tensor Sqrt(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

StridedSlice(Tensor, Int32[], Int32[], Int32[])

Strided slice

Declaration

Tensor StridedSlice(Tensor x, int[] starts4Dor8D, int[] ends4Dor8D, int[] strides4Dor8D)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	starts4Dor8D
Int32[]	ends4Dor8D
Int32[]	strides4Dor8D	stride

Returns

Type	Description
Tensor	output Tensor

Sub(Tensor[])

Subtract tensors o = tensors[0] - tensors[1] - ... - tensors[N-1]

Declaration

Tensor Sub(Tensor[] tensors)

Parameters

Type	Name	Description
Tensor[]	tensors	input tensors

Returns

Type	Description
Tensor	output Tensor

Swish(Tensor)

Swish

Declaration

Tensor Swish(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Tan(Tensor)

Tan

Declaration

Tensor Tan(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Tanh(Tensor)

Tanh

Declaration

Tensor Tanh(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Tile(Tensor, Int32[])

Tile

Declaration

Tensor Tile(Tensor x, int[] repeats)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	repeats	repetition counts

Returns

Type	Description
Tensor	output Tensor

TopKIndices(Tensor, Int32, Int32, Boolean, Boolean)

Top K indices

Declaration

Tensor TopKIndices(Tensor x, int k, int axis, bool largest, bool sorted)

Parameters

Type	Name	Description
Tensor	x	input
Int32	k	k
Int32	axis	axis
Boolean	largest	largest flag
Boolean	sorted	sorted flag

Returns

Type	Description
Tensor	output Tensor

TopKValues(Tensor, Tensor, Int32)

Top K values

Declaration

Tensor TopKValues(Tensor X, Tensor I, int axis)

Parameters

Type	Name	Description
Tensor	X	input
Tensor	I	indices
Int32	axis	axis

Returns

Type	Description
Tensor	output Tensor

Transpose(Tensor)

Transpose matrix

Declaration

Tensor Transpose(Tensor x)

Parameters

Type	Name	Description
Tensor	x	input

Returns

Type	Description
Tensor	output Tensor

Transpose(Tensor, Int32[])

Transpose according to permutations

Declaration

Tensor Transpose(Tensor x, int[] permutations)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	permutations	new axis order

Returns

Type	Description
Tensor	output Tensor

Upsample2D(Tensor, Int32[], Boolean)

Upsample 2D

Declaration

Tensor Upsample2D(Tensor x, int[] scale, bool bilinear)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	scale	scale
Boolean	bilinear	bilinear flag

Returns

Type	Description
Tensor	output Tensor

Upsample3D(Tensor, Int32[], Boolean)

Upsample 3D

Declaration

Tensor Upsample3D(Tensor x, int[] scale, bool trilinear)

Parameters

Type	Name	Description
Tensor	x	input
Int32[]	scale	scale
Boolean	trilinear	trilinear flag

Returns

Type	Description
Tensor	output Tensor

Where(Tensor, Tensor, Tensor)

Where

Declaration

Tensor Where(Tensor c, Tensor a, Tensor b)

Parameters

Type	Name	Description
Tensor	c	Tensor c
Tensor	a	Tensor a
Tensor	b	Tensor b

Returns

Type	Description
Tensor	Tensor with values `c` ? `a` : `b`