Any plan to support GGUF?

[thipokKub]

GGUF is one of the prominent quantization method, which vastly reduced the model size, and inference time. However, it seems like Draw Things did not support this model mode, are there any plans to support this? Or it is recommended to convert everything to *.safetensors first?

[ianscrivener]

You absolutely can import models including main mjodels LoRAs and ControlNet models from safetensor files.

DrawThings is heavily optimised to deliver significantly faster inference than what is possible on a Mac using GGUF. Adding GGUF support would be a big backward step in regards to inference speed... and probably no body would bother using it. I can't see that it would be supported by either the users of DrawThings or the developer.