6d ago · llama.cpp releases →
moe-cache : compute part of the misses on the CPU --moe-cache-cpu-share F computes a share F of the MoE cache misses on the CPU from the host experts instead…
Includes: b12188: server: visit parent slots before child slots (#30779), b12189: metal : fix GROUP_NORM for ne03 > 1 and non-contiguous inputs (#31852), llama.cpp b12190, b12191: llama_moe_cache: run missed experts on CPU, add tuning script (#30966)