Hi, I installed Nobara on the weekend and I keep running into OOM errors, if I enable a single LoRA in a Flux workflow it will trigger it, remove the LoRA and it works fine. I’m using Atomix Flux model which is 12GB and Force/Set Clip to move it to the CPU, so I should have enough VRAM for the task.
I have a 3090 with 24GB of VRAM and 64GB of system RAM. I do also have Windows on the same PC and Comfy runs fine without OOM errors on the same workflows with multiple LoRAs enabled, so I figure it’s something with the OS, but I’m not that familiar with Linux.
I’ve tried adding --highvram to my launch arguments, but that hasn’t helped. Anyone with experience with Linux have any ideas on how to resove this?