DiffEqBayes.jl is a library that lets people use Bayesian inference on ODEs without knowledge of those libraries.
https://github.com/JuliaDiffEq/DiffEqBayes.jl
What we’ve found so far is that Turing’s NUTS is the most stable, even more stable than the auto-generated Stan kernels. DynamicHMC diverges for us sometimes. Turing is much faster than it used to be. These tests show that it’s the thing to use, at least for now.