How to use the increased context with llama.cpp?

#1
by avion23 - opened

Is it enough to just increase the context? like this?
./main --color -c 16384 --temp 0.2 --repeat_penalty 1.15 -n -1 -i -ins --multiline-input -t 8 -ngl 128 -m ../models/longchat-13b-16k.ggmlv3.q6_K.bin

It's not supported with llama.cpp at this time - please see the README for details of how to use with KoboldCpp, the only client that currently supports it

Oh thanks.
Btw I really appreciate the work you are doing by uploading all these LLM

avion23 changed discussion status to closed

Sign up or log in to comment