JacopoMangiavacchi

import Foundation
import CoreML

func compileCoreML(path: String) -> (MLModel, URL) {
    let modelUrl = URL(fileURLWithPath: path)
    let compiledUrl = try! MLModel.compileModel(at: modelUrl)
    
    print("Compiled Model Path: \(compiledUrl)")
    return try! (MLModel(contentsOf: compiledUrl), compiledUrl)
}

JacopoMangiavacchi

import Foundation
import CoreML

func generateData(sampleSize: Int = 100) -> ([Float], [Float]) {
    let a: Float = 2.0
    let b: Float = 1.5
    
    var X = [Float]()
    var Y = [Float]()

JacopoMangiavacchi

import Foundation

struct OrderedDict<K: Hashable, V> {
    var _dict = [K : V]()
    var _ordered = [K]()
    
    var keys: [K] {
        _ordered
    }
    

JacopoMangiavacchi

func prepareBatchProvider() -> MLBatchProvider {
    var featureProviders = [MLFeatureProvider]()

    var count = 0
    errno = 0
    let trainFilePath = Bundle.main.url(forResource: "mnist_train", withExtension: "csv")!
    if freopen(trainFilePath.path, "r", stdin) == nil {
        print("error opening file")
    }
    while let line = readLine()?.split(separator: ",") {

JacopoMangiavacchi

public func prepareModel() {
    let coremlModel = Model(version: 4,
                            shortDescription: "MNIST-Trainable",
                            author: "Jacopo Mangiavacchi",
                            license: "MIT",
                            userDefined: ["SwiftCoremltoolsVersion" : "0.0.12"]) {
        Input(name: "image", shape: [1, 28, 28])
        Output(name: "output", shape: [10], featureType: .float)
        TrainingInput(name: "image", shape: [1, 28, 28])
        TrainingInput(name: "output_true", shape: [1], featureType: .int)

JacopoMangiavacchi

import Foundation
import MLCompute
import PlaygroundSupport

let page = PlaygroundPage.current
page.needsIndefiniteExecution = true

let tensor1 = MLCTensor(descriptor: MLCTensorDescriptor(shape: [6, 1], dataType: .float32)!)
let tensor2 = MLCTensor(descriptor: MLCTensorDescriptor(shape: [6, 1], dataType: .float32)!)
let tensor3 = MLCTensor(descriptor: MLCTensorDescriptor(shape: [6, 1], dataType: .float32)!)

JacopoMangiavacchi

private func readDataSet(fileName: String, updateStatus: @escaping (Int) -> Void) -> ([Float], [Float]) { //}(MLCTensor, MLCTensor) {
    guard let filePath = Bundle.main.path(forResource: fileName, ofType: "csv") else {
        fatalError("CSV file not found")
    }

    let serialQueue = DispatchQueue(label: "MNIST.serial.queue.\(fileName)")

    var count = 0
    var X = [Float]()
    var Y = [Float]()

JacopoMangiavacchi

// Load in memory and split is not performant
private func getFileLine(filePath: String, process: (String) -> Void) {
    guard let filePointer:UnsafeMutablePointer<FILE> = fopen(filePath,"r") else {
        preconditionFailure("Could not open file at \(filePath)")
    }

    defer {
        fclose(filePointer)
    }

JacopoMangiavacchi

private func initializeTensors() {
    device = MLCDevice(type: .cpu)!

    inputTensor = MLCTensor(descriptor: MLCTensorDescriptor(shape: [batchSize, imageSize, 1, 1], dataType: .float32)!)

    dense1WeightsTensor = MLCTensor(descriptor: MLCTensorDescriptor(shape: [1, imageSize*dense1LayerOutputSize, 1, 1], dataType: .float32)!,
                                        randomInitializerType: .glorotUniform)
    dense1BiasesTensor = MLCTensor(descriptor: MLCTensorDescriptor(shape: [1, dense1LayerOutputSize, 1, 1], dataType: .float32)!,
                                       randomInitializerType: .glorotUniform)
    dense2WeightsTensor = MLCTensor(descriptor: MLCTensorDescriptor(shape: [1, dense1LayerOutputSize*numberOfClasses, 1, 1], dataType: .float32)!,

JacopoMangiavacchi

private func buildTrainingGraph() {
    trainingGraph = MLCTrainingGraph(graphObjects: [graph],
                                         lossLayer: MLCLossLayer(descriptor: MLCLossDescriptor(type: .softmaxCrossEntropy,
                                                                                               reductionType: .mean)),
                                         optimizer: MLCAdamOptimizer(descriptor: MLCOptimizerDescriptor(learningRate: 0.001,
                                                                                                       gradientRescale: 1.0,
                                                                                                    regularizationType: .none,
                                                                                                    regularizationScale: 0.0),
                                                                     beta1: 0.9,
                                                                     beta2: 0.999,