b3389 is already out

#1
by Nexesenex - opened

And might require yet another reconversion.. ^^

https://github.com/ggerganov/llama.cpp/pull/8473

reuploading ^^

Thanks!
Q8_0 benches great (ppl512 <8.8), and is coherent in my KoboldCPP fork with all the Gemma commits up to b3389.

Sign up or log in to comment