llama.cpp
hexagon: support for HMX flash-attention head_dim not multiple of 64
#26539
Merged

Loading