User Profile

Hello, I have been following the excellent/informative "Metal for Machine Learning" from WWDC19 to learn how to do on device training (I have a specific use case for this) and it is all working really well using the MPSNNGraph. However, I would like to call my own metal compute/render function/pipeline to transform the inference result before calculating the loss, does anyone know if this possible and what would this look like in code? Please see my current code below, at the comment I need to call an intermediate compute/render function to transform the inference result image before passing to the MPSNNForwardLossNode. let rgbImageNode = MPSNNImageNode(handle: nil) let inferGraph = makeInferenceGraph() let reshape = MPSNNReshapeNode(source: inferGraph.resultImage, resultWidth: 64, resultHeight: 64, resultFeatureChannels: 4) //Need to call render or compute pipeline to post process in the inference result image let rgbLoss = MPSNNForwardLossNode(source:reshape.resultImage, labels:rgbImageNode, lossDescriptor:lossDescriptor) let initGrad = MPSNNInitialGradientNode(source:rgbLoss.resultImage) let gradNodes = initGrad.trainingGraph(withSourceGradient:nil, nodeHandler:nil) guard let trainGraph = MPSNNGraph(device: device, resultImage: gradNodes![0].resultImage, resultImageIsNeeded: true) else{ fatalError("Unable to get training graph.") } Thanks

Hi, In SceneKit I pass custom parameters to a metal shader using a SCNAnimation, for example: let revealAnimation = CABasicAnimation(keyPath: "revealage") revealAnimation.duration =  duration revealAnimation.toValue = toValue let scnRevealAnimation = SCNAnimation(caAnimation: revealAnimation) material.addAnimation(scnRevealAnimation, forKey: "Reveal") How would I do similar to a metal shader in RealityKit? I saw in the Octopus example: //int(params.uniforms().custom_parameter()[0]) But it's commented out and there is no example how to set the custom variable and animate it? (unless I missed it) Great session BTW Thanks