Yet another llm model format to support? AWQ

[dreemur99]

Im sure you already seen it already but theres a another new model format. AWQ. Claims to be "blazing-fast" with much lower vram requirements. Looks like an almost 45% reduction in reqs.

[LostRuins]

Probably won't be adding this on my own, unless llama.cpp also adds it, then koboldcpp will inherit their support.

[dreemur99]

makes sense.