Kai Xu

Originally it was just to make the codebase more readable, as the types basically annotate the arguments. It sometimes prevents trivial bugs as well. I actually lean to do what...

Yes you are right. Users just need to set global logger's IO to text. We probably just need to provide a helper function for this.

Related PR: #297

I'm not sure if we want to do this in AHMC. My intension is indeed to control the verbosity using Julia's logging system, as David suggested [there](https://github.com/TuringLang/Turing.jl/issues/1398#issuecomment-687259280). Maybe Turing can...

Can you elaborate the unicode issue for Python? As I believe Python 3 does support unicode quite well (for both function and variable).

Mohamed once had such an attempt (https://github.com/TuringLang/AdvancedHMC.jl/pull/76) but we didn't finally take it because it makes the code harder to read, and at that moment the package was still in...

Just to keep track of things: quick benchmark by @sethaxen (https://github.com/TuringLang/AdvancedHMC.jl/pull/214#discussion_r512947999)

https://github.com/TuringLang/Turing.jl/issues/769

I think for the scope of this package, we should assume the density and gradient function returns that of a mini batch.

Great point. I plan to work on this actually. And yes @StannisZhou - we need to extend the current design as it currently suppose every dimension is in `AbstractVector{T} where...