llama.cpp

aa79524c - HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (#14945)