llama.cpp

7f3955a0 - model : make falcon-h1 use shared mamba2 layer builder