Passthorugh multiple gpu to one vm for llm purpose

I’m searching for ways to pass through two nvidia p40 gpu into specific vm.
Since VRAM size is important while running large language models, i think i need to passthrough multiple gpus at once.

In order to pass thorugh, i followed the guide(Create a Gaming HVM - #48 by rookie).

However, when typing sudo lspci -vvn, i find out either one or another, not both gpus have [Kernel driver in use : pciback] in the description. I tried several method, but since i didn’t know much about syntax using rd.qubes.hide_pci , i can’t manage out.

I searched for several ways to deal with such problems, i find this article:

However, it seems there are no documentation exist in any of those which deals with passthroughing multiple gpu to one vm at once.

Is this possible, pass-through two gpu?

What did you try?
The syntax to hide multiple PCI devices is like this, e.g. to hide 0a:00.0 and 0b:00.0 devices:

GRUB_CMDLINE_LINUX="... rd.qubes.hide_pci=0a:00.0,0b:00.0 "