llama.cpp

Add AWQ (Activation-aware Weight Quantization) for llama, llama2, mpt, and mistral models

#4593

Merged