can you tell me please how to convert model fp8?

Provide more details about this model. Does this need VAE? SD3 Lora will this work?. Provide ComfyUI workflow.

Edit: the medium model is rocksolid, fits well and runs great.

This also won't work with my 8GB v-ram.

Clips always load first, there goes +6GB of v-ram.

Model loads on system memory, it's slow and painful if you have +32GB if 16GB then it's a no go.

and we still don't have loras or ip/controlnet yet which will make things even worst. Right now the only option is GGUF Q4/4_1 or 5 until next week.

Thank you! 3 minutes in my GTX 1650, 8 GB ram, i7, Euler Normal CFG 1 768x960