MPSCNNConvolutionDescriptor
A description of the attributes of a convolution kernel.
Declaration
class MPSCNNConvolutionDescriptorOverview
You use an MPSCNNConvolutionDescriptor object to describe the properties of an MPSCNNConvolution kernel such as its size, pixel format and CPU cache mode.
Topics
Type Methods
init(coder:)init(kernelWidth:kernelHeight:inputFeatureChannels:outputFeatureChannels:)init(kernelWidth:kernelHeight:inputFeatureChannels:outputFeatureChannels:neuronFilter:)
Instance Properties
groupsinputFeatureChannelskernelHeightkernelWidthoutputFeatureChannelsstrideInPixelsXstrideInPixelsYneurondilationRateXdilationRateYneuronParameterA()neuronParameterB()neuronType()fusedNeuronDescriptor
Instance Methods
encode(with:)setBatchNormalizationParametersForInferenceWithMean(_:variance:gamma:beta:epsilon:)setNeuronToPReLUWithParametersA(_:)setNeuronType(_:parameterA:parameterB:)