llama.cpp

dc1f0510 - convert-llama-7b-pth-to-gguf.py : rope scale and added tokens