Closed kossikater closed 3 years ago
Hi! I agree, however, this is currently not possible. What you can do instead is to override the Jacobian function in the problem definition and there implement (either and FD or AD scheme) that takes into account the sparsity pattern.
@chria Is it possible for you to create a simple example of how to use AD scheme to provide Jacobian for sundials solver such as IDA or CVODE ? I am using casadi python version to provide symbolic jacobian and try to use it in 'ida_with_jac.py' example but I keep getting errors such as f function return type error or Jacobian function is not callable
It would be great if the solvers could use a Jacobian sparsity pattern indicating which entries are zero for sure, similar to
scipy
'ssolve_ivp
. This could increase speed for calculations on approximating the Jacobian is no analytical one is supplied.