Penelope Yong
Penelope Yong
It's curious, but the ForwardDiff tests that were failing previously seem to have been fixed on Julia 1.11. These were the previous failures: https://github.com/TuringLang/Turing.jl/actions/runs/11072371830/job/30766467971 This was pre-1.11 so the CI...
The remaining test failure is #2371. I could rerun CI to get the green tick, but don't see much of a point.
CI passed on the main branch after we merged 😄 🎉
Sure, happy to. Feel free to close this and I'll open a new PR against main with the same contents.
Sorry, I should probably have checked here before opening https://github.com/TuringLang/Turing.jl/issues/2613. Anyway, the TL;DR is that I benchmarked it and `arraydist` and `filldist` can largely be removed, sans a specialisation for...
Reproducing my comment on that Turing issue. ---------------------------------- My hypothesis is that we can completely replace `filldist` and `arraydist` with the following: ```julia # Current definitions filldist = DistributionsAD.filldist arraydist...
@yebai You'd like to actually just remove both of them and make users use `product_distribution` directly? I was thinking that I wouldn't mind keeping them around indefinitely since it'd only...
Would you assume that you don’t observe it? ;)
That thought has struck me, but I think it's impossible for an automated tool to figure out which of the PRs are related and should be aggregated. Although grouping multiple...
Hmm, I clearly didn't dig back far enough when searching issues. Agreed on the importance of type stability, especially given the context that this isn't an issue with Distributions itself....