hi there! since the start of the year I have been playing around with harvester using two gpu servers of my department and while everything of the ordinary works, as soon as I start trying to get the GPUs (H100) to pass through to the VM, there are only troubles. • passing only one h100 to a vm works fine, it appears in the guest using

lspci

(I am using an ubuntu 22 server image), but no matter which nvidia driver I select, nvidia throws the usual error (cant detect the nvidia driver, etc) • passing multiple h100 gpus to the guest simply results into crash loops upon start up for the VM I have tried it with the latest stable release (1.2.1?) and the latest 1.3 pre-release, but I cant seem to get it to work. has anyone encountered a similar issue and knows how to resolve it? IOMMU in bios is set to auto (i.e. enabled by default), along wiht any other potential options for virtualization in the BIOS.