ggml-cuda : support stablelm rope #4156
slaren
merged
4 commits into
master
from
cuda-stablelm-rope
ggml-cuda : support stablelm rope
58444931
remove unused freq_base kernel parameter
4a3469f2
ggerganov
approved these changes
on 2023-11-21
cebtenzzre
commented
on 2023-11-21
add n_dims parameter to llm_build_k_shift, default to n_rot via overload
84adb541
llama : fix llm_build_k_shift args
5ed3e1a8
Login to write a write a comment.
Login via GitHub