@yizhang Excited about your library. I’ve played with it some.
Is it possible to use fazang to compute the band jacobian terms only, and exploit the matrix structure? In other words, is it possible to re-write the finite-difference jacobian function you gave in your example, but using AD instead.