llama.cpp

models : appropriately flag noscan ssm_a tensors

#28121

Merged