mpscnnconvolutionweightsandbiasesstate(3) [mojave man page]
MPSCNNConvolutionWeightsAndBiasesState(3) MetalPerformanceShaders.framework MPSCNNConvolutionWeightsAndBiasesState(3) NAME
MPSCNNConvolutionWeightsAndBiasesState SYNOPSIS
#import <MPSCNNConvolution.h> Inherits MPSState. Instance Methods (nonnull instancetype) - initWithWeights:biases: (nonnull instancetype) - initWithDevice:cnnConvolutionDescriptor: Class Methods (nonnull instancetype) + temporaryCNNConvolutionWeightsAndBiasesStateWithCommandBuffer:cnnConvolutionDescriptor: Properties __nonnull id< MTLBuffer > weights __nullable id< MTLBuffer > biases Detailed Description The MPSCNNConvolutionWeightsAndBiasesState is returned by exportWeightsAndBiasesWithCommandBuffer: method on MPSCNNConvolution object. This is mainly used for GPU side weights/biases update process. During training, application can keep a copy of weights, velocity, momentum MTLBuffers in its data source, update the weights (in-place or out of place) with gradients obtained from MPSCNNConvolutionGradientState and call [MPSCNNConvolution reloadWeightsAndBiasesWithCommandBuffer] with resulting updated MTLBuffer. If application does not want to keep a copy of weights/biases, it can call [MPSCNNConvolution exportWeightsAndBiasesWithCommandBuffer:] to get the current weights from convolution itself, do the updated and call reloadWithCommandBuffer. Method Documentation - (nonnull instancetype) initWithDevice: (__nonnull id< MTLDevice >) device(MPSCNNConvolutionDescriptor *__nonnull) descriptor - (nonnull instancetype) initWithWeights: (__nonnull id< MTLBuffer >) weights(__nullable id< MTLBuffer >) biases + (nonnull instancetype) temporaryCNNConvolutionWeightsAndBiasesStateWithCommandBuffer: (__nonnull id< MTLCommandBuffer >) commandBuffer(MPSCNNConvolutionDescriptor *__nonnull) descriptor Property Documentation - biases [read], [nonatomic], [assign] A buffer that contains the biases. Each value is float and there are ouputFeatureChannels values. - weights [read], [nonatomic], [assign] A buffer that contains the weights. Each value in the buffer is a float. The layout of the weights with respect to the weights is the same as the weights layout provided by data source i.e. it can be interpreted as 4D array weights[outputFeatureChannels][kernelHeight][kernelWidth][inputFeatureChannels/groups] Author Generated automatically by Doxygen for MetalPerformanceShaders.framework from the source code. Version MetalPerformanceShaders-100 Thu Feb 8 2018 MPSCNNConvolutionWeightsAndBiasesState(3)
Check Out this Related Man Page
MPSCNNConvolutionTransposeNode(3) MetalPerformanceShaders.framework MPSCNNConvolutionTransposeNode(3) NAME
MPSCNNConvolutionTransposeNode SYNOPSIS
#import <MPSNNGraphNodes.h> Inherits MPSCNNConvolutionNode. Instance Methods (nonnull instancetype) - initWithSource:convolutionGradientState:weights: Class Methods (nonnull instancetype) + nodeWithSource:convolutionGradientState:weights: Properties MPSCNNConvolutionGradientStateNode * convolutionGradientState Detailed Description A MPSNNFilterNode representing a MPSCNNConvolutionTranspose kernel Method Documentation - (nonnull instancetype) initWithSource: (MPSNNImageNode *__nonnull) sourceNode(MPSCNNConvolutionGradientStateNode *__nullable) convolutionGradientState(nonnull id< MPSCNNConvolutionDataSource >) weights Init a node representing a MPSCNNConvolutionTransposeNode kernel Parameters: sourceNode The MPSNNImageNode representing the source MPSImage for the filter convolutionGradientState When the convolution transpose is used to 'undo' an earlier convolution in the graph, it is generally desired that the output image be the same size as the input image to the earlier convolution. You may optionally specify this size identity by passing in the MPSCNNConvolutionGradientState node here. weights A pointer to a valid object conforming to the MPSCNNConvolutionDataSource protocol. This object is provided by you to encapsulate storage for convolution weights and biases. Returns: A new MPSNNFilter node for a MPSCNNConvolutionTransposeNode kernel. + (nonnull instancetype) nodeWithSource: (MPSNNImageNode *__nonnull) sourceNode(MPSCNNConvolutionGradientStateNode *__nullable) convolutionGradientState(nonnull id< MPSCNNConvolutionDataSource >) weights Init an autoreleased not representing a MPSCNNConvolutionTransposeNode kernel Parameters: sourceNode The MPSNNImageNode representing the source MPSImage for the filter convolutionGradientState When the convolution transpose is used to 'undo' an earlier convolution in the graph, it is generally desired that the output image be the same size as the input image to the earlier convolution. You may optionally specify this size identity by passing in the MPSNNConvolutionGradientStateNode created by the convolution node here. weights A pointer to a valid object conforming to the MPSCNNConvolutionDataSource protocol. This object is provided by you to encapsulate storage for convolution weights and biases. Returns: A new MPSNNFilter node for a MPSCNNConvolutionTransposeNode kernel. Property Documentation - (MPSCNNConvolutionGradientStateNode*) convolutionGradientState [read], [nonatomic], [assign] unavailable Author Generated automatically by Doxygen for MetalPerformanceShaders.framework from the source code. Version MetalPerformanceShaders-100 Thu Feb 8 2018 MPSCNNConvolutionTransposeNode(3)