llama.cpp
CUDA: fix MMQ stream-k for --split-mode row
#8167
Merged

Loading