llama.cpp just added GQA and full support for 70B LLaMA-2
#16
by
igzbar
- opened
time to make a GGML version? ;)
I will try. But it's not full support yet. The convert.py script still doesn't work with HF repos. So I can't do any fine tune conversions for sure. I will try with the base llama 2 70B which is available as PTH which apparently does work
Thank you! :)