Llama.cpp server question

#3
by Ransom - opened

Currently, Llama.cpp server endpoint does not support the rope-scale parameter, but it does support:
printf(" --rope-freq-base N RoPE base frequency (default: loaded from model)\n");
printf(" --rope-freq-scale N RoPE frequency scaling factor (default: loaded from model)\n");

Could you provide insight into what the best parameters for this model would be?

Thanks so much!

Sign up or log in to comment