llama.cpp

[SYCL] Fixed minor bug when enabling FP16 for non intel targets

#6464

Merged