Class CompareOps

Compares output of two different implementations of IOps. Useful for debugging purposes

Inheritance
object
CompareOps
Implements
Inherited Members
Namespace: Unity.Barracuda
Assembly: solution.dll
Syntax
public class CompareOps : IOps, IOpsStatistics

Constructors

CompareOps(IOps, IOps, LogLevel, float)

Create CompareOps

Declaration
public CompareOps(IOps ops1, IOps ops2, CompareOpsUtils.LogLevel differenceLogLevel, float epsilon)
Parameters
Type Name Description
IOps ops1

first IOps implementation
IOps ops2

second IOps implementation
CompareOpsUtils.LogLevel differenceLogLevel

difference log level
float epsilon

error threshold

Methods

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

LSTM

Declaration
public 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)]

NonZero(Tensor)

Indices for non zero values

Declaration
public Tensor NonZero(Tensor X)
Parameters
Type Name Description
Tensor X

input
Returns
Type Description
Tensor

output Tensor

PostLayerCleanup()

Called after every layer execution. It allows IOps to run cleanup operations such as clearing temporary buffers only used in the scope of the last layer executed.

Declaration
public virtual void PostLayerCleanup()

PreExecuteLayer(Layer, Tensor[])

Prepare for layer execution

Declaration
public virtual void PreExecuteLayer(Layer layer, Tensor[] inputs)
Parameters
Type Name Description
Layer layer

layer
Tensor[] inputs

inputs

PrepareModel(Model, IDictionary<string, TensorShape>, IVars)

Prepare model for execution, allocating required intermediate tensors

Declaration
public virtual void PrepareModel(Model model, IDictionary<string, TensorShape> inputShapes, IVars vars)
Parameters
Type Name Description
Model model

model
IDictionary<string, TensorShape> inputShapes

input shapes
IVars vars

model variables

TopKValues(Tensor, Tensor, int)

Top K values

Declaration
public Tensor TopKValues(Tensor X, Tensor I, int axis)
Parameters
Type Name Description
Tensor X

input
Tensor I

indices
int axis

axis
Returns
Type Description
Tensor

output Tensor

Implements