model definition by https://github.com/0cc4m/GPTQ-for-LLaMa/commits?author=0cc4m

[Tom-Neverwinter]

adding in awareness as tagging users in GitHub is a pain

https://github.com/0cc4m/GPTQ-for-LLaMa/blob/latestmerge/gptq/mpt.py

thanks to baptistejamin for pointing it out

[sw]

You can tag users by prepending @ to their username, but what is the issue with llama.cpp here? Please describe what you think should be done/fixed/etc.

[Tom-Neverwinter]

You can tag users by prepending @ to their username, but what is the issue with llama.cpp here? Please describe what you think should be done/fixed/etc.

last week in an issue @ggerganov was looking for someone to write the C++ for gptq and 4bit. this appears to be a working implementation.

[sw]

Ok, let's keep the discussion in #9, closing this as duplicate.