try *cd llama.cpp && python convert-pth-to-ggml.py models/7B/ 1*