Hello,
In my scenario I have access to the covariance matrix at the MAP (I’m using Levenberg Marquardt optimization). I should be able to provide the covariance matrix as the inv mass matrix directly to the nuts kernel and tell it to use that. I was able to find that the kernel has the attribute “_adapter.mass_matrix_adapter.inverse_mass_matrix” with a setter function, however when I try to use this it just ignores my mass matrix. I also tried setting: “_adapter.mass_matrix_adapter.inverse_mass_matrix” which also just ignored my matrix. Is there a way to set the mass matrix, or should I open an issue? Thanks!