We are planning for a POC on a Linux bare metal node with L40s. We are running Kubernetes on top of it . Given that L40s doesn’t support MiG, we were looking at other options for GPU partitioning.
We wanted to know what the best option is to configure VGPU on such a set up .
What are the hardware specific requirements and what are the software requirements that needs to be installed on the Linux node and Kubernetes and in what order .
Is a hypervisor sufficient or VMs need to be configured to use VGPUs.
I understand these could be basic questions but appreciate any help in this regard