GPU Passthrough with VMWare on Windows on Laptop - CUDA as a service


I know that VMWare Workstation does not support GPU passthrough from Host to VM Guests. But I am wondering if it is possible to do what I want to do.

My question: Is it possible to set up my laptop (a Windows 10 ThinkPad with RTX A2000 ) as a “CUDA server” that meets the following needs:

  • have my IDEs and tools (VSCode, Jupyter, others) running inside VMWare Workstation Guest VMs
  • VM can submit jobs to the VMWare workstation Host for training and inference
  • Host uses the GPU on the laptop to run the jobs
  • config does not jeopardize the stability of the Host
  • easy enough for a beginner to install and maintain software and config on the Host and the VMs.

I am new to CUDA dev and new to Jupyter. I am hoping to learn by doing. I am about to buy a new laptop and i will spend a little extra to upgrade the graphics to RTX A2000 if it is possible to do what I want above.

Thanks in advance.