David Widmann

This seems to be a bug, I should have updated the internal calls to `pairwise!` and `colwise!` only in a breaking release and left them unchanged until the deprecations are...

Yes, it provides everything in that function (and some additional features). For instance, the example in the scipy docstring could be written as (functionality for exact computations was extracted to...

Seems one could exploit though that $$ \begin{split} DFT(A)[k] &= \sum^{length(A)}_{n=1} \exp{\bigg(\frac{- 2\pi i (k - 1) (n - 1)}{length(A)}\bigg)} A[n] \\ &= \begin{dcases} N A[1] & \text{if } k...

Yep, I just noticed that I could have simplified it even more 😅

It's still there, just using `AbstractFFT` instead of `FFTW` (which I assume one would also want to use in an extension rather than `FFTW`): https://github.com/FluxML/Zygote.jl/blob/e2d7839a476fc1cda407500d9a2ca125bd6e2aa5/src/lib/array.jl#L672-L677

IMO it's related to #835 but not a duplicate (I guess one could view it differently though): #835 is about reverse mode rules whereas this issue is just about blatant...

That was our point above :slightly_smiling_face:

The problem can be fixed by ```diff M src/lib/broadcast.jl @@ -295,9 +295,14 @@ end y = broadcast(x -> value(x), out) function bc_fwd_back(ȳ) dargs = ntuple(valN) do i - unbroadcast(args[i], broadcast((y1,...

Seems the Zygote test failures only point to a few tests that have to be updated since they rely on the current non-collapsing behaviour.

DynamicPPL can be ignored - AbstractMCMC 4.5 contains a breaking change which was accidentally released in a non-breaking release (https://github.com/TuringLang/AbstractMCMC.jl/pull/132#issue-1959740311). Hopefully it will be yanked soon (https://github.com/JuliaRegistries/General/pull/94031).