Compute Library
 21.02
InstanceNormalizationLayerKernelInfo Struct Reference

#include <KernelDescriptors.h>

Public Member Functions

 InstanceNormalizationLayerKernelInfo ()
 Default constructor. More...
 
 InstanceNormalizationLayerKernelInfo (float gamma, float beta, float epsilon, bool use_mixed_precision)
 Constructor. More...
 

Data Fields

float gamma
 The scale scalar value applied to the normalized tensor. More...
 
float beta
 The offset scalar value applied to the normalized tensor. More...
 
float epsilon
 Lower bound value for the normalization. More...
 
bool use_mixed_precision
 Use mixed precision in case of FP16 execution. More...
 

Detailed Description

Definition at line 128 of file KernelDescriptors.h.

Constructor & Destructor Documentation

◆ InstanceNormalizationLayerKernelInfo() [1/2]

Default constructor.

Definition at line 131 of file KernelDescriptors.h.

132  : InstanceNormalizationLayerKernelInfo(1.f, 0.f, 1e-12, true)
133  {
134  }

◆ InstanceNormalizationLayerKernelInfo() [2/2]

InstanceNormalizationLayerKernelInfo ( float  gamma,
float  beta,
float  epsilon,
bool  use_mixed_precision 
)
inline

Constructor.

Parameters
[in]gammaThe scale scalar value applied to the normalized tensor.
[in]betaThe offset scalar value applied to the normalized tensor
[in]epsilonLower bound value for the normalization.
[in]use_mixed_precisionUse mixed precision in case of FP16 execution.

Definition at line 142 of file KernelDescriptors.h.

144  {
145  }
bool use_mixed_precision
Use mixed precision in case of FP16 execution.
float epsilon
Lower bound value for the normalization.
float gamma
The scale scalar value applied to the normalized tensor.
float beta
The offset scalar value applied to the normalized tensor.

Field Documentation

◆ beta

float beta

The offset scalar value applied to the normalized tensor.

Defaults to 0.0

Definition at line 148 of file KernelDescriptors.h.

Referenced by NEInstanceNormalizationLayerKernel::configure(), CLInstanceNormalizationLayerKernel::configure(), and NEInstanceNormalizationLayerKernel::validate().

◆ epsilon

float epsilon

Lower bound value for the normalization.

Defaults to 1e-12

Definition at line 149 of file KernelDescriptors.h.

Referenced by NEInstanceNormalizationLayerKernel::configure(), CLInstanceNormalizationLayerKernel::configure(), and NEInstanceNormalizationLayerKernel::validate().

◆ gamma

float gamma

The scale scalar value applied to the normalized tensor.

Defaults to 1.0

Definition at line 147 of file KernelDescriptors.h.

Referenced by NEInstanceNormalizationLayerKernel::configure(), CLInstanceNormalizationLayerKernel::configure(), and NEInstanceNormalizationLayerKernel::validate().

◆ use_mixed_precision

bool use_mixed_precision

Use mixed precision in case of FP16 execution.

Defaults to true

Definition at line 150 of file KernelDescriptors.h.

Referenced by NEInstanceNormalizationLayerKernel::configure(), and CLInstanceNormalizationLayerKernel::configure().


The documentation for this struct was generated from the following file: