BilevelJuMP: Best way to initialize primal and dual problem

Dear Community,

I have a relatively large bilevel optimization problem that I’m trying to solve with BilevelJuMP; however it is very slow in reaching a feasible solution.
I am using the FortuntyAmat mode, aiming to fasten the computational time.
The initialization of only the primal variables is not enough and I need to initialize the entire model, including the binary constrary constraints of the formulation.

My idea is to solve the model a first time using simplified bounds, then initialize the start values using this simplified solution to help the solver find a feasible initial solution.
What is the best to do so?

I believe after solving the first model I can retrive the solution of the primal and dual variables, then with set_dual_start and set_value_start, I should be able to initialize the primal and dual variables, however I cannot initialize the binary constraints.
Is that enough?
Any other suggested approache?

Thank you very much

There are a number of open issues about this on BilevelJuMP: I suggest you keep the conversation there. BilevelJuMP is still in development.