llama.cpp

ggml-cuda: remove unused params in ggml_cuda_graph

#18579

Merged