As ForwardDiff is still the default autograd backend for Turing.jl, one may wish for it to work outside the model. ``` using Turing @model demo() = x ~ Normal() sample(demo(), NUTS(), 2000) ``` This would motivate calling ForwardDiff via its public API.
As ForwardDiff is still the default autograd backend for Turing.jl, one may wish for it to work outside the model.
This would motivate calling ForwardDiff via its public API.