Package apple.metalperformanceshaders

Class MPSCNNConvolutionTranspose

  • All Implemented Interfaces:
    NSCoding, NSCopying, NSSecureCoding, NSObject
    public class MPSCNNConvolutionTranspose
extends MPSCNNKernel
    MPSCNNConvolutionTranspose [@dependency] This depends on Metal.framework The MPSCNNConvolutionTranspose specifies a transposed convolution. The MPSCNNConvolutionTranspose convolves the input image with a set of filters, each producing one feature map in the output image. Some third-party frameworks may rotate the weights spatially by 180 degrees for Convolution Transpose. MPS uses the weights specified by the developer as-is and does not perform any rotation. The developer may need to rotate the weights appropriately in case this rotation is needed before the convolution transpose is applied. When the stride in any dimension is greater than 1, the convolution transpose puts (stride - 1) zeroes in-between the source image pixels to create an expanded image. Then a convolution is done over the expanded image to generate the output of the convolution transpose. Intermediate image size = (srcSize - 1) * Stride + 1 Examples: [@code] So in case of sride == 2 (this behaves same in both dimensions) Source image: _______________ | | | | | | 1 | 2 | 3 | 4 | | | | | | --------------- Intermediate Image: ___________________________ | | | | | | | | | 1 | 0 | 2 | 0 | 3 | 0 | 4 | | | | | | | | | --------------------------- NOTE on Offset: There are 2 types of offsets defined: 1) The Offset defined in MPSCNNKernel from which MPSCNNConvolutionTranspose inherits. This offset is applied to from where the kernel will be applied on the source. 2) The kernelOffsetX and kernelOffsetY which is the offset applied to the kernel when it is finally applied on the intermediate image. So totalOffset = Offset * stride + kernelOffset The offset defined by user refers to the coordinate frame of the expanded image (we are showing only 1 dimension X it can be extended to Y dimension as well) : X indicates where the convolution transpose begins: Intermediate Image: Offset = 0, kernelOffset = 0 ___________________________ | | | | | | | | | 1 | 0 | 2 | 0 | 3 | 0 | 4 | | X | | | | | | | --------------------------- X indicates where the convolution transpose begins: Intermediate Image: Offset = 0, kernelOffset = 1 ___________________________ | | | | | | | | | 1 | 0 | 2 | 0 | 3 | 0 | 4 | | | X | | | | | | --------------------------- X indicates where the convolution transpose begins: Intermediate Image: Offset = 0, kernelOffset = -1 ___________________________ | | | | | | | | X | 1 | 0 | 2 | 0 | 3 | 0 | 4 | | | | | | | | | --------------------------- So if the user wanted to apply an offset of 2 on the source image of convolution transpose: Source image: _______________ | | | | | | 1 | 2 | 3 | 4 | | | | X | | --------------- offset = 2, kernelOffset = 0 Intermediate Image: ___________________________ | | | | | | | | | 1 | 0 | 2 | 0 | 3 | 0 | 4 | | | | | | X | | | --------------------------- [@endcode] Note that if your application is not using MPSCNNConvolutionGradientState to configure the convolution transpose with respect to convolution, your application may do this using padding policy. In such case if convolution uses valid padding policy, than convolution transpose should use full padding policy and vice vera. Full padding remains full.

    • Constructor Detail

      • MPSCNNConvolutionTranspose

        protected MPSCNNConvolutionTranspose​(org.moe.natj.general.Pointer peer)

    • Method Detail

      • accessInstanceVariablesDirectly

        public static boolean accessInstanceVariablesDirectly()

      • allocWithZone

        public static java.lang.Object allocWithZone​(org.moe.natj.general.ptr.VoidPtr zone)

      • automaticallyNotifiesObserversForKey

        public static boolean automaticallyNotifiesObserversForKey​(java.lang.String key)

      • cancelPreviousPerformRequestsWithTarget

        public static void cancelPreviousPerformRequestsWithTarget​(java.lang.Object aTarget)

      • cancelPreviousPerformRequestsWithTargetSelectorObject

        public static void cancelPreviousPerformRequestsWithTargetSelectorObject​(java.lang.Object aTarget,
                                                                         org.moe.natj.objc.SEL aSelector,
                                                                         java.lang.Object anArgument)

      • classFallbacksForKeyedArchiver

        public static NSArray<java.lang.String> classFallbacksForKeyedArchiver()

      • classForKeyedUnarchiver

        public static org.moe.natj.objc.Class classForKeyedUnarchiver()

      • debugDescription_static

        public static java.lang.String debugDescription_static()

      • description_static

        public static java.lang.String description_static()

      • groups

        public long groups()
        [@property] groups Number of groups input and output channels are divided into.

      • hash_static

        public static long hash_static()

      • initWithCoderDevice

        public MPSCNNConvolutionTranspose initWithCoderDevice​(NSCoder aDecoder,
                                                      java.lang.Object device)
        support
        Overrides:
        initWithCoderDevice in class MPSCNNKernel
        Parameters:
        aDecoder - The NSCoder subclass with your serialized MPSKernel
        device - The MTLDevice on which to make the MPSKernel
        Returns:
        A new MPSKernel object, or nil if failure.

      • initWithDevice

        public MPSCNNConvolutionTranspose initWithDevice​(java.lang.Object device)
        Description copied from class: MPSCNNKernel
        Standard init with default properties per filter type
        Overrides:
        initWithDevice in class MPSCNNKernel
        Parameters:
        device - The device that the filter will be used on. May not be NULL.
        Returns:
        A pointer to the newly initialized object. This will fail, returning nil if the device is not supported. Devices must be MTLFeatureSet_iOS_GPUFamily2_v1 or later.

      • initWithDeviceWeights

        public MPSCNNConvolutionTranspose initWithDeviceWeights​(MTLDevice device,
                                                        MPSCNNConvolutionDataSource weights)
        Initializes a convolution transpose kernel
        Parameters:
        device - The MTLDevice on which this MPSCNNConvolutionTranspose filter will be used
        weights - A pointer to a object that conforms to the MPSCNNConvolutionDataSource protocol. The MPSCNNConvolutionDataSource protocol declares the methods that an instance of MPSCNNConvolutionTranspose uses to obtain the weights and bias terms for the CNN convolutionTranspose filter. Currently we support only Float32 weights.
        Returns:
        A valid MPSCNNConvolutionTranspose object.

      • inputFeatureChannels

        public long inputFeatureChannels()
        [@property] inputFeatureChannels The number of feature channels per pixel in the input image.

      • instanceMethodSignatureForSelector

        public static NSMethodSignature instanceMethodSignatureForSelector​(org.moe.natj.objc.SEL aSelector)

      • instancesRespondToSelector

        public static boolean instancesRespondToSelector​(org.moe.natj.objc.SEL aSelector)

      • isSubclassOfClass

        public static boolean isSubclassOfClass​(org.moe.natj.objc.Class aClass)

      • kernelOffsetX

        public long kernelOffsetX()
        [@property] kernelOffsetX Offset in X from which the kernel starts sliding

      • kernelOffsetY

        public long kernelOffsetY()
        [@property] kernelOffsetY Offset in Y from which the kernel starts sliding

      • keyPathsForValuesAffectingValueForKey

        public static NSSet<java.lang.String> keyPathsForValuesAffectingValueForKey​(java.lang.String key)

      • new_objc

        public static java.lang.Object new_objc()

      • outputFeatureChannels

        public long outputFeatureChannels()
        [@property] outputFeatureChannels The number of feature channels per pixel in the output image.

      • resolveClassMethod

        public static boolean resolveClassMethod​(org.moe.natj.objc.SEL sel)

      • resolveInstanceMethod

        public static boolean resolveInstanceMethod​(org.moe.natj.objc.SEL sel)

      • setKernelOffsetX

        public void setKernelOffsetX​(long value)
        [@property] kernelOffsetX Offset in X from which the kernel starts sliding

      • setKernelOffsetY

        public void setKernelOffsetY​(long value)
        [@property] kernelOffsetY Offset in Y from which the kernel starts sliding

      • setVersion_static

        public static void setVersion_static​(long aVersion)

      • superclass_static

        public static org.moe.natj.objc.Class superclass_static()

      • supportsSecureCoding

        public static boolean supportsSecureCoding()

      • _supportsSecureCoding

        public boolean _supportsSecureCoding()
        Description copied from interface: NSSecureCoding
        This property must return YES on all classes that allow secure coding. Subclasses of classes that adopt NSSecureCoding and override initWithCoder: must also override this method and return YES. The Secure Coding Guide should be consulted when writing methods that decode data.
        Specified by:
        _supportsSecureCoding in interface NSSecureCoding
        Overrides:
        _supportsSecureCoding in class MPSCNNKernel

      • version_static

        public static long version_static()

      • accumulatorPrecisionOption

        public long accumulatorPrecisionOption()
        Precision of accumulator used in convolution. See MPSNeuralNetworkTypes.h for discussion. Default is MPSNNConvolutionAccumulatorPrecisionOptionFloat.

      • dataSource

        public MPSCNNConvolutionDataSource dataSource()
        [@property] dataSource dataSource with which convolution transpose object was created

      • encodeToCommandBufferSourceImageConvolutionGradientState

        public MPSImage encodeToCommandBufferSourceImageConvolutionGradientState​(MTLCommandBuffer commandBuffer,
                                                                         MPSImage sourceImage,
                                                                         MPSCNNConvolutionGradientState convolutionGradientState)
        Encode a MPSCNNKernel into a command Buffer. Create a texture to hold the result and return it. In the first iteration on this method, encodeToCommandBuffer:sourceImage:destinationImage: some work was left for the developer to do in the form of correctly setting the offset property and sizing the result buffer. With the introduction of the padding policy (see padding property) the filter can do this work itself. If you would like to have some input into what sort of MPSImage (e.g. temporary vs. regular) or what size it is or where it is allocated, you may set the destinationImageAllocator to allocate the image yourself. This method uses the MPSNNPadding padding property to figure out how to size the result image and to set the offset property. See discussion in MPSNeuralNetworkTypes.h. Note: the regular encodeToCommandBuffer:sourceImage: method may be used when no state is needed, such as when the convolution transpose operation is not balanced by a matching convolution object upstream. These encode methods are for auto encoders where each convolution in inference pass is coupled with convolution transpose. In order for convolution transpose to correctly undo the convolution downsampling, MPSCNNConvolutionGradientState produced by convolution is needed by convolution transpose to correctly size destination image. These methods are only useful for inference only network. For training, use encode methods that take MPSCNNConvolutionTransposeGradientState below.
        Parameters:
        commandBuffer - The command buffer
        sourceImage - A MPSImage to use as the source images for the filter.
        convolutionGradientState - A valid MPSCNNConvolutionGradientState from the MPSCNNConvoluton counterpart to this MPSCNNConvolutionTranspose. If there is no forward convolution counterpart, pass NULL here. This state affects the sizing the result.
        Returns:
        A MPSImage or MPSTemporaryImage allocated per the destinationImageAllocator containing the output of the graph. The offset property will be adjusted to reflect the offset used during the encode. The returned image will be automatically released when the command buffer completes. If you want to keep it around for longer, retain the image. (ARC will do this for you if you use it later.)

      • encodeToCommandBufferSourceImageConvolutionGradientStateDestinationStateDestinationStateIsTemporary

        public MPSImage encodeToCommandBufferSourceImageConvolutionGradientStateDestinationStateDestinationStateIsTemporary​(MTLCommandBuffer commandBuffer,
                                                                                                                    MPSImage sourceImage,
                                                                                                                    MPSCNNConvolutionGradientState convolutionGradientState,
                                                                                                                    org.moe.natj.general.ptr.Ptr<MPSCNNConvolutionTransposeGradientState> outState,
                                                                                                                    boolean isTemporary)
        These low level encode functions should be used during training. The first two encode functions, which return destination image on left hand side, takes in MPSCNNConvolutionGradientState that was produced by corresponding MPSCNNConvolution when there is one e.g. auto encoders. This state is used to correctly size destination being returned. These encode methods return MPSCNNConvoltionTransposeGradientState object on auto release pool to be consumed by MPSCNNConvolutionTransposeGradient.

      • exportWeightsAndBiasesWithCommandBufferResultStateCanBeTemporary

        public MPSCNNConvolutionWeightsAndBiasesState exportWeightsAndBiasesWithCommandBufferResultStateCanBeTemporary​(MTLCommandBuffer commandBuffer,
                                                                                                               boolean resultStateCanBeTemporary)
        GPU side export. Enqueue a kernel to export current weights and biases stored in MPSCNNConvoltionTranspose's internal buffers into weights and biases MTLBuffer * returned in MPSCNNConvolutionWeightsAndBiasesState. * * @param commandBuffer Metal command buffer on which export kernel is enqueued. * @param resultStateCanBeTemporary If FALSE, state returned will be non-temporary. If TRUE, returned state may or may not be temporary. * @return MPSCNNConvolutionWeightsAndBiasesState containing weights and biases buffer to which weights got exported. This state and be temporary or non-temporary depending on the flag resultStateCanBeTemporary

      • reloadWeightsAndBiasesFromDataSource

        public void reloadWeightsAndBiasesFromDataSource()
        CPU side reload. Reload the updated weights and biases from data provider into internal weights and bias buffers. Weights and biases gradients needed for update are obtained from MPSCNNConvolutionTransposeGradientState object. Data provider passed in init call is used for this purpose.

      • reloadWeightsAndBiasesWithCommandBufferState

        public void reloadWeightsAndBiasesWithCommandBufferState​(MTLCommandBuffer commandBuffer,
                                                         MPSCNNConvolutionWeightsAndBiasesState state)
        GPU side reload. Reload the updated weights and biases from update buffer produced by application enqueued metal kernel into internal weights and biases buffer. Weights and biases gradients needed for update are obtained from MPSCNNConvolutionTransposeGradientState object's gradientForWeights and gradientForBiases metal buffer.
        Parameters:
        commandBuffer - Metal command buffer on which application update kernel was enqueued consuming MPSCNNConvolutionGradientState's gradientForWeights and gradientForBiases buffers and producing updateBuffer metal buffer.
        state - MPSCNNConvolutionWeightsAndBiasesState containing weights and biases buffers which have updated weights produced by application's update kernel. The state readcount will be decremented.

      • resultStateForSourceImageSourceStatesDestinationImage

        public MPSCNNConvolutionTransposeGradientState resultStateForSourceImageSourceStatesDestinationImage​(MPSImage sourceImage,
                                                                                                     NSArray<? extends MPSState> sourceStates,
                                                                                                     MPSImage destinationImage)
        Allocate a MPCNNConvolutionTransposeGradientState to hold the results from a -encodeBatchToCommandBuffer... operation
        Overrides:
        resultStateForSourceImageSourceStatesDestinationImage in class MPSCNNKernel
        Parameters:
        sourceImage - The MPSImage consumed by the associated -encode call.
        sourceStates - The list of MPSCNNConvolutionGradientState consumed by the associated -encode call, for a batch size of 1. In auto encoders, this state is produced by corresponding MPSCNNConvolution.
        destinationImage - The destination image for the encode call
        Returns:
        The list of states produced by the -encode call for batch size of 1. -isResultStateReusedAcrossBatch returns YES for MPSCNNConvolutionTranspose so same state is used across entire batch. State object is not reusasable across batches.

      • setAccumulatorPrecisionOption

        public void setAccumulatorPrecisionOption​(long value)
        Precision of accumulator used in convolution. See MPSNeuralNetworkTypes.h for discussion. Default is MPSNNConvolutionAccumulatorPrecisionOptionFloat.

      • temporaryResultStateForCommandBufferSourceImageSourceStatesDestinationImage

        public MPSCNNConvolutionTransposeGradientState temporaryResultStateForCommandBufferSourceImageSourceStatesDestinationImage​(MTLCommandBuffer commandBuffer,
                                                                                                                           MPSImage sourceImage,
                                                                                                                           NSArray<? extends MPSState> sourceStates,
                                                                                                                           MPSImage destinationImage)
        Description copied from class: MPSCNNKernel
        Allocate a temporary MPSState (subclass) to hold the results from a -encodeBatchToCommandBuffer... operation A graph may need to allocate storage up front before executing. This may be necessary to avoid using too much memory and to manage large batches. The function should allocate any MPSState objects that will be produced by an -encode call with the indicated sourceImages and sourceStates inputs. Though the states can be further adjusted in the ensuing -encode call, the states should be initialized with all important data and all MTLResource storage allocated. The data stored in the MTLResource need not be initialized, unless the ensuing -encode call expects it to be. The MTLDevice used by the result is derived from the command buffer. The padding policy will be applied to the filter before this is called to give it the chance to configure any properties like MPSCNNKernel.offset. CAUTION: The kernel must have all properties set to values that will ultimately be passed to the -encode call that writes to the state, before -resultStateForSourceImages:sourceStates:destinationImage: is called or behavior is undefined. Please note that -destinationImageDescriptorForSourceImages:sourceStates:destinationImage: will alter some of these properties automatically based on the padding policy. If you intend to call that to make the destination image, then you should call that before -resultStateForSourceImages:sourceStates:destinationImage:. This will ensure the properties used in the encode call and in the destination image creation match those used to configure the state. The following order is recommended: // Configure MPSCNNKernel properties first kernel.edgeMode = MPSImageEdgeModeZero; kernel.destinationFeatureChannelOffset = 128; // concatenation without the copy ... // ALERT: will change MPSCNNKernel properties MPSImageDescriptor * d = [kernel destinationImageDescriptorForSourceImage: source sourceStates: states]; MPSTemporaryImage * dest = [MPSTemporaryImage temporaryImageWithCommandBuffer: cmdBuf imageDescriptor: d]; // Now that all properties are configured properly, we can make the result state // and call encode. MPSState * __nullable destState = [kernel temporaryResultStateForCommandBuffer: cmdBuf sourceImage: source sourceStates: states]; // This form of -encode will be declared by the MPSCNNKernel subclass [kernel encodeToCommandBuffer: cmdBuf sourceImage: source destinationState: destState destinationImage: dest ]; Default: returns nil
        Overrides:
        temporaryResultStateForCommandBufferSourceImageSourceStatesDestinationImage in class MPSCNNKernel
        Parameters:
        commandBuffer - The command buffer to allocate the temporary storage against The state will only be valid on this command buffer.
        sourceImage - The MPSImage consumed by the associated -encode call.
        sourceStates - The list of MPSStates consumed by the associated -encode call, for a batch size of 1.
        destinationImage - The destination image for the encode call
        Returns:
        The list of states produced by the -encode call for batch size of 1. When the batch size is not 1, this function will be called repeatedly unless -isResultStateReusedAcrossBatch returns YES. If -isResultStateReusedAcrossBatch returns YES, then it will be called once per batch and the MPSStateBatch array will contain MPSStateBatch.length references to the same object.