llama.cpp
ggml-cuda: hip: Tune stream_k algo and mmq configs to use stream_k on AMD GCN
#30022
Merged

Loading