David Widmann
David Widmann
This has come up quite a few times but fortunately the solution is easy: Use NaN-safe mode in ForwardDiff (by default, it does return incorrect result for infinite values with...
Initially I assumed it would, but based on the tests I ran it didn't, which is why I opted for removing this special branch completely. My hypothesis was that it...
Thank you, I'll try to have a look at and fix the remaining issues this week.
The latest commit added a way to mark some Enzyme tests as broken (and I also introduced a bug in the test setup :smile:) but it also reveals that there...
@wsmoses Is this an enzyme bug or did I call autodiff incorrectly? https://github.com/TuringLang/DistributionsAD.jl/actions/runs/5872312798/job/15923610267?pr=254#step:6:303
No, this is a major change, so it requires a bit more discussion, feedback from other maintainers, and should in the end be approved by another maintainer. I'm also not...
I'm merging this PR, based on the tests it seems to be a clear improvement. We can iterate further if someone notices additional problems.
Seems every PR requires approval before it can be merged.
Maybe we could even think about removing the `RFunctions` submodule completely. With #191 interfacing with Rmath will not require complex `ccall`s anymore but we could instead directly define `nbetapdf(...) =...
ChainRules tests are generally a bit sensitive with respect to the (random) inputs, I've seen random test failures in StatsFuns but in other repos as well. We could easily avoid...