Get started with Windows ML

• .NET 6 or greater
• Targeting a Windows 10-specific TFM like net6.0-windows10.0.19041.0 or greater

C++20 or later.

Python versions 3.10 to 3.13, on x64 and ARM64 devices.

See use the Windows App SDK in an existing project for how to add the Windows App SDK to your project, or if you're already using Windows App SDK, update your packages.

• Make sure you install the 1.8.0-Experimental4 version, because the release versions don't contain Windows ML yet.
• Make sure you use the framework-dependent deployment option, since Windows ML currently doesn't support self-contained deployment.

See use the Windows App SDK in an existing project for how to add the Windows App SDK to your project, or if you're already using Windows App SDK, update your packages.

• Make sure you install the 1.8.0-Experimental4 version, because the release versions don't contain Windows ML yet.
• Make sure you use the framework-dependent deployment option, since Windows ML currently doesn't support self-contained deployment.

The Python binding leverages the pywinrt project for the Windows App SDK projection. For the 1.8 experimental 4 release, these packages are hosted in a private index and will upstream to pywinrt and publish to PyPI in future versions.

You'll also need to install a special onnxruntime package with the auto-ep feature enabled (this feature will be available in a future onnxruntime release).

Create a requirements.txt file with the following content:

--index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple
--extra-index-url https://pypi.org/simple
onnxruntime-winml==1.22.0.post2
winrt-runtime==3.2.1
winrt-Windows.Foundation==3.2.1
winrt-Windows.Foundation.Collections==3.2.1
winui3-Microsoft.Windows.AI.MachineLearning==1!1.8.250702007.dev4
winui3-Microsoft.Windows.ApplicationModel.DynamicDependency.Bootstrap==1!1.8.250702007.dev4

Then install using pip:

pip install -r requirements.txt

using Microsoft.ML.OnnxRuntime;
using Microsoft.Windows.AI.MachineLearning;

// First we create a new instance of EnvironmentCreationOptions
EnvironmentCreationOptions envOptions = new()
{
    logId = "WinMLDemo", // Use an ID of your own choice
    logLevel = OrtLoggingLevel.ORT_LOGGING_LEVEL_ERROR
};

// And then use that to create the ORT environment
using var ortEnv = OrtEnv.CreateInstanceWithOptions(ref envOptions);

// Get the default ExecutionProviderCatalog
var catalog = ExecutionProviderCatalog.GetDefault();

// Ensure and register all compatible execution providers with ONNX Runtime
// This downloads any necessary components and registers them
await catalog.EnsureAndRegisterAllAsync();

#include <winrt/Microsoft.Windows.AI.MachineLearning.h>
#include <win_onnxruntime_cxx_api.h>

// First we need to create an ORT environment
Ort::Env env(ORT_LOGGING_LEVEL_ERROR, "WinMLDemo"); // Use an ID of your own choice

// Get the default ExecutionProviderCatalog
winrt::Microsoft::Windows::AI::MachineLearning::ExecutionProviderCatalog catalog =
    winrt::Microsoft::Windows::AI::MachineLearning::ExecutionProviderCatalog::GetDefault();

// Ensure and register all compatible execution providers with ONNX Runtime
catalog.EnsureAndRegisterAllAsync().get();

# Known issue: import winrt.runtime will cause the TensorRTRTX execution provider to fail registration.
# As a workaround, please run pywinrt related code in a separate thread.

# winml.py
import json

def _get_ep_paths() -> dict[str, str]:
    from winui3.microsoft.windows.applicationmodel.dynamicdependency.bootstrap import (
        InitializeOptions,
        initialize
    )
    import winui3.microsoft.windows.ai.machinelearning as winml
    eps = {}
    with initialize(options = InitializeOptions.ON_NO_MATCH_SHOW_UI):
        catalog = winml.ExecutionProviderCatalog.get_default()
        providers = catalog.find_all_providers()
        for provider in providers:
            provider.ensure_ready_async().get()
            eps[provider.name] = provider.library_path
            # DO NOT call provider.try_register in python. That will register to the native env.
    return eps

if __name__ == "__main__":
    eps = _get_ep_paths()
    print(json.dumps(eps))

# In your application code
import subprocess
import json
import sys
from pathlib import Path
import onnxruntime as ort

def register_execution_providers():
    worker_script = str(Path(__file__).parent / 'winml.py')
    result = subprocess.check_output([sys.executable, worker_script], text=True)
    paths = json.loads(result)
    for item in paths.items():
        ort.register_execution_provider_library(item[0], item[1])
    _ep_registered = True

register_execution_providers()