PyTorch Container & Windows

I was interested in using the nVIDIA PyTorch container with Docker Desktop for Windows and the WSL2 back end enabled. Since I am a novice, I wanted to know if this is possible, and if so, some guidance on how to proceed with pointers if possible. For example, do I also need to install the nVIDIA Container Toolkit in order to make use of my nVIDIA GPUs? Any assistance would be much appreciated…

This user guide indicates how to set up WSL2 for NGC container usage.

Dear Robert:

Thanks for the quick and helpful response. I will study the guide and then come back with further questions as appropriate. My intention is to try and use Docker Desktop for Windows to run a containerized Python code that calls your PyTorch container in order to make use of nVIDIA Quadro RTX5000 GPU cores in my workstation. I will let you know what happens…
Best Regards,
Christopher

I’m not really familiar with docker desktop for windows. It’s not necessary to use that to launch a NGC container in a WSL2 partition. If you decide to use it anyway, I don’t know what issues you may run into, and this wouldn’t be the place to ask questions about docker desktop for windows.

If you follow the guide I linked, you should be able to launch docker containers in WSL2 using a methodology identical to how you would do it on linux. That is the only functionality/modality supported by NVIDIA.

Thanks for that information — you are already helping me much since I am frankly overwhelmed by all the possibilities and associated documentation. Again, I will now concentrate on the guide you recommended and see what happens.