Kaan Öcal

Looks like I got a little confused as to where bijectors went in all these packages. I assumed that they were built into AdvancedHMC but I might have been relying...

> To be clear, this is not built into AdvancedMH because it is a non-trivial task. A large part of [DynamicPPL.jl](https://github.com/TuringLang/DynamicPPL.jl), Turing's modelling language, is designed to explicitly handle just...

Constructing an array in general fails with Zygote: ```julia using Zygote using ComponentArrays Zygote.gradient(x -> ComponentArray(a = [5])[1], [0.]) ``` gives ```julia ERROR: Mutating arrays is not supported -- called...

From what I understand the parallel `mcmcsample` implementations call the serial `mcmcsample` in each thread/process, so maybe one could hook each sampler to a callback which updates the corresponding progress...

I see! It might be a good idea to provide a generic threaded/distributed dispatching mechanism (extending the `mcmcsample` implementation) so a user basically only has to write a serial `sample`...

Seems like I misinterpreted your comment then, you're right, the distributed `mcmcsample` implementations call `StatsBase.sample`.

We should probably get this PR out of the way. I did not pursue this much further because I had trouble getting it to work on my sampling problems -...

Have there been any updates on this? I am using some hand-written special functions and the type-unstable code caused by this issue is very slow in comparison to ForwardDiff...

Sure, happy to do that - I've not been very up to date with some of the more recent changes to Catalyst and Symbolics, as you might have noticed...

I appreciate the work on both packages. Unfortunately the invitation link has expired as I had other commitments for the past week or two - if you could please re-send...