Installation support for Windows / WSL users (CUDA, DGL, cu117 issues)

[8MaCaVeL8]

Hi,
I’ve been trying to install RFdiffusion on my Windows machine using WSL (Ubuntu), but I keep running into problems with dependencies — especially with CUDA, cuDNN, PyTorch (cu117), and DGL.

Even after following all available guides and trying different versions of these frameworks, installation either fails or the model won’t run properly due to GPU or CUDA compatibility issues.

It would be very helpful if the team (or community) could provide:

1. A step-by-step guide for installation on Windows/WSL.

2. Recommended CUDA and PyTorch versions that actually work under WSL.

3. Any known limitations or workarounds for Windows users.

Since many researchers and developers use Windows as their main system, having an official installation guide or Docker image for WSL would make RFdiffusion much more accessible.

Thank you for your amazing work on this project — it’s an incredible tool, and I’d love to get it running locally!

Best regards,
Kamyar Entezari

[rclune]

Out of curiosity, does the Docker image work on your system?

Also, do you know what version of CUDA your system supports? That will determine the versions of PyTorch and DGL that you want to look at.