By the way you don’t need to wrap everything in deterministics, only if you need those values afterwards. It will save on memory/computation.
You might benefit from changing how you compute gammaMu_c, but as you commented it might not be trivial.
Otherwise it looks fine, but the fact you get divergences could mean bad priors/ challenging model. I would focus on fixing that before worrying about speedup. As you would still be getting invalid samples, just faster.