Load WinML runtime DLL for SDK EP discovery

[prathikr]

Summary

• Load Microsoft.Windows.AI.MachineLearning.dll from the Core native directory for the Python SDK when present.
• Preload the same WinML runtime DLL for Rust winml builds from the native runtime directory.
• Preload the WinML runtime DLL for C# WinML builds alongside ORT and ORT GenAI.
• No JS SDK code change needed; its existing loader already handles the native runtime directory.

Validation

• Rust WinML EP smoke without a caller PATH override discovered CUDA, OpenVINO, and NvTensorRTRTX, then registered OpenVINO.
• C# WinML source smoke without a caller PATH override discovered CUDA, OpenVINO, and NvTensorRTRTX, then registered OpenVINO.
• JS EP smoke without a caller PATH override discovered CUDA, OpenVINO, and NvTensorRTRTX, then registered OpenVINO.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
foundry-local	Ready	Preview, Comment	May 27, 2026 11:05pm

[Copilot]

Pull request overview

This PR updates the Python SDK’s native library initialization on Windows to optionally load the WinML runtime DLL (Microsoft.Windows.AI.MachineLearning.dll) when it is present, enabling WinML execution providers in the WinML variant of the package.

Changes:

• Add a cached _winml_library handle to CoreInterop to keep the WinML DLL loaded for the process lifetime.
• During Windows initialization, detect and load Microsoft.Windows.AI.MachineLearning.dll from the Core binary directory when present.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.