llama.cpp

007a8f6f - Support all LLaMA models + change Q4_0 quantization storage