The hessian matrix at the "mode" is not positive definite!

Dynare help (legacy posts)
2

Search the forum. See e.g. [Usual Bayesian estimation problem). Do you already use mode_compute=6? If nothing else works, use mcmc_jumping_covariance