Hi,
just wanted to let you all know, that Lemonade SDK now also provides image generation via stable diffusion cpp, since v9.2. Models List - Lemonade Server Documentation
I have been using lemonade most of the time on my framework desktop. There is high interest in the developer team to support Strix Halo. It is super convenient. I often also just add some unsloth GGUF and they normally work. Nowadays you can also load multiple models.
Image editing is not yet supported, but might be in the future. For those cases I would recommend the toolboxes of kyuz0 Llama.cpp/vLLM Toolboxes for LLM inference on Strix Halo - #37 by Kathir_Vel
SDXL-Turbo does not seem to know framework desktop though. or my prompt was too unspecific 
- took around 10-15s with those settings.
