David Widmann

I haven't checked the tutorial (BTW there exist already some on the Turing webpage which could be updated and moved here) but I think it might be better to add...

> Keeping a simple tutorial inside DynamicPPL help to demystify stuff here, I think. I think it's great to improve documentation (we really have), and I als think it should...

To make my comments more concrete and also more constructive. This is a rough sketch of a simple MH sampler for DynamicPPL: ```julia using DynamicPPL using AbstractMCMC struct MHSampler{P} (m...

It's fully working. The AbstractMCMC tutorial is focused on Turing specifics and largely outdated. The correct, (hopefully) complete and up-to-date docs are in the AbstractMCMC repo: https://turinglang.github.io/AbstractMCMC.jl/dev/

With #381 being merged and released, the MH sampler can be simplified to: ```julia using DynamicPPL using AbstractMCMC using Random struct MHSampler{P} (m = randn(rng) + x.m,)) sample(model, sampler, 100)...

Couldn't we just save and print the model definition of the user as an expression (if you actually intend to display what the user defined and not just a summary)?...

I thought there was an issue about this in Bijectors, for inputs and outputs of different size but also more generally for inputs and outputs of different types, but I...

Or to not transform in-place (this was also discussed at some point) and/or to use a different non-linearized structure (IIRC there are some discussions that were moved to AbstractPPL).

As a temporary workaround you can just fill the transformed values with 0s (or some other value) but I guess that's maybe what you want to avoid here.

Even if that might be a way to solve the problem, I'm not sure that's how we should actually do it (sorry :stuck_out_tongue:). Adding a deterministic distribution and an identity...