Running tensorflow/etc. inside vms…? Is it workable for performance?

This feels like a profoundly stupid question, and maybe that’s why I’m not finding any answers to it… am new to machine learning.

I’m used to doing development inside VMs, but as I want to benefit from the GPU that’s not really an option here, right? I was thinking maybe I could do it in a Docker container instead (am on Windows) but not sure that’s viable, either. Would either a VM or Docker work for Windows and doing ML? Thanks.