llama.cpp
vulkan: preprocess mul_mat_id experts and discard workgroups more quickly
#18352
Merged

Loading