llama.cpp

ggml-cuda: Adding support for unified memory

#8035

Merged