llama.cpp

py : use cpu-only torch in requirements.txt

#8335

Merged