Just simple workflow with 2 custom nodes. UnetLoaderGGufAdvancedDisTorchMultiGPU as the GGUF loader.
This can be changed to whatever loader you wanna use. I use this because I read that it could help with speed on my 3060 12gb. It helped with wan2.1, maybe it helps with flux too.

I also use Power Lora Loader with hyper-flux.1-dev-8step lora. Power lora loader (rgthbecause you have the ability to right click on lora and check tags.

I also have upscale model and a skindiffdetail model in the workflow.