CUDA: tune FA for Gemma 4 on Ampere or newer - #29152

CUDA: tune FA for Gemma 4 on Ampere or newer#29152

IMbackK merged 1 commit into

ggml-org:masterggml-org/llama.cpp:masterfrom

JohannesGaessler:cuda-fa-tune-2JohannesGaessler/llama.cpp:cuda-fa-tune-2Copy head branch name to clipboard