When an object detection model created with ML.net Model builder is run on Maui (iOS), the model fails to load.

[ysanada12]

Description: I have created a machine learning model for object detection using ML.net Model Builder (MLModel1.zip). The model was successfully tested using the automatically generated project by ML.net Model Builder.

I then added the model as an embedded resource in a .NET MAUI project and implemented the model loading. Below is the source code:

        public static readonly Lazy<PredictionEngine<ModelInput, ModelOutput>> PredictEngine = new Lazy<PredictionEngine<ModelInput, ModelOutput>>(() => CreatePredictEngine(), true);

        private static PredictionEngine<ModelInput, ModelOutput>CreatePredictEngine()
        {
            var mlContext = new MLContext();
            ITransformer mlModel = LoadModel();
            return mlContext.Model.CreatePredictionEngine<ModelInput, ModelOutput>(mlModel);
        }
        public static ITransformer LoadModel()
        {
            var assembly = typeof(MLModel1).GetTypeInfo().Assembly;
            System.IO.Stream stream = assembly.GetManifestResourceStream("MauiProject.MLModel1.zip");

            if (stream == null)
            {
                throw new FileNotFoundException("Resource not found");
            }

            using (var fileStream = new MemoryStream())
            {
                stream.CopyTo(fileStream);
                fileStream.Position = 0;

                var mlContext = new MLContext();
                return mlContext.Model.Load(fileStream, out _);
            }
        }

If you build this code in the Ios simulator, you will get an error that libtorch-cpu-any does not exist when you reach mlContext.Model.Load.
libtorch-cpuのパッケージを確認したところ、この名称のdllが存在しないようです。
https://www.nuget.org/packages?q=libtorch-cpu

Error Message

This application or script uses TorchSharp but doesn't contain a reference to libtorch-cpu-any, Version=2.2.1.1.

Consider referencing one of the combination packages TorchSharp-cpu, TorchSharp-cuda-linux, TorchSharp-cuda-windows or call System.Runtime.InteropServices.NativeLibrary.Load(path-to-libLibTorchSharp.so) explicitly for a Python install of pytorch. See https://github.com/dotnet/TorchSharp/issues/169.".

For CUDA, you may need to call 'TorchSharp.torch.InitializeDeviceType(TorchSharp.DeviceType.CUDA)' before any use of TorchSharp CUDA packages from scripts or notebooks.

Trace from LoadNativeBackend:

TorchSharp: LoadNativeBackend: Initialising native backend, useCudaBackend = False

Step 1 - First try regular load of native libtorch binaries.

    Trying to load native component torch_cpu relative to /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers/Bundle/Application/74E75385-2642-4E2C-8105-E9C1107CECCF/calouselTest.app/TorchSharp.dll
    Failed to load native component torch_cpu relative to /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers/Bundle/Application/74E75385-2642-4E2C-8105-E9C1107CECCF/calouselTest.app/TorchSharp.dll
    Trying to load native component LibTorchSharp relative to /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers/Bundle/Application/74E75385-2642-4E2C-8105-E9C1107CECCF/calouselTest.app/TorchSharp.dll
    Failed to load native component LibTorchSharp relative to /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers/Bundle/Application/74E75385-2642-4E2C-8105-E9C1107CECCF/calouselTest.app/TorchSharp.dll
    Result from regular native load of LibTorchSharp is False

Step 3 - Alternative load from consolidated directory of native binaries from nuget packages

    torchsharpLoc = /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers/Bundle/Application/74E75385-2642-4E2C-8105-E9C1107CECCF/calouselTest.app
    packagesDir = /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers
    torchsharpHome = /Users/devadmin/Library/Developer/CoreSimulator/Devices/2027A329-FB06-4053-AB75-3F1DB97EE874/data/Containers/Bundle/Application
    Giving up, TorchSharp.dll does not appear to have been loaded from package directories

Could anyone help me resolve this issue?

[feiyun0112]

Required NuGet in addition to Microsoft.ML | Microsoft.ML.TorchSharp and libtorch-cpu or libtorch-cuda-11.3 or any of the OS specific variants.