Penelope Yong
Penelope Yong
Fixed now
This is finally fixed on main: ```julia julia> Bijectors.logabsdetjac(b, θ) 225.96796253912498 ``` I've triggered a new release, so it should be available soon.
closing for now, will try to write docs on it soon :)
I agree that the fields of the different `Transition`s don't define a common interface (yet) — the only usefulness of defining an abstract supertype would be to fix the method...
Hi @SamuelBrand1, thanks for the issue! Could you provide a MWE that shows what you're trying to do with Turing that causes this crash? My initial impression is that this...
Closing in favour of upstream issue
Thinking about this and #2338 a little, it's not super easy to dive into the internals of `sample()` to check that the gradient is correct, but what I'm thinking we...
Ok, so it seems we want to construct a list of `LogDensityFunctions`, then. (Or a list of models to construct them from.) And I suspect these should go into DynamicPPL.TestUtils.
I'll start playing around with this in a new package, we can decide whether to put it in DPPL proper at a later point in time.
Ooh. I like that. But I don't think LogDensityProblemsAD uses DI for all its backends just yet, right? In which case, I _think_ we'd still need to test `logdensity_and_gradient` itself?...