CodeLlama-70B

#5
by eramax - opened

Could you please quantize the CodeLlama-70B-instruct & Python versions using the SOTA approach.

Alas, eramax, I have only a small CPU (i7-6700k), and so I'm crunching only my favorite models, or at least good ones.
And CodeLlama 70b is an absolutely grotesque release of Meta in my opinion. I tested it, and deleted it on the spot.
Nevertheless, a lot of people might use it professionnally, so there's an audience and maybe Kooten or Artefact2 can help you, because they do these IQ quants also.

Alright, thanks

eramax changed discussion status to closed

Sign up or log in to comment