The device plugin should support Windows 2019 and Windows 2022 containers (Windows Containers on Windows).
2. Steps to reproduce the issue
This feature is implemented when the daemonset distributing the k8s-device-plugin installs NVIDIA drivers using Windows hostprocess containers headlessly. In my experience, the following line is sufficient once the drivers have been extracted from the executables provided by NVIDIA:
1. Issue or feature description
The device plugin should support Windows 2019 and Windows 2022 containers (Windows Containers on Windows).
2. Steps to reproduce the issue
This feature is implemented when the daemonset distributing the k8s-device-plugin installs NVIDIA drivers using Windows hostprocess containers headlessly. In my experience, the following line is sufficient once the drivers have been extracted from the executables provided by NVIDIA:
3. Additional Information
You should probably use in the work from https://tensorworks.com.au/blog/bringing-full-gpu-support-to-windows-containers-in-kubernetes/ to expose DirectX APIs. In my experience, before the release of the prior software, using https://unrealcontainers.com/blog/enabling-vendor-specific-graphics-apis-in-windows-containers/, CUDA works correctly.