maartenvd

> Seems to be failing only for real matrices, where comparison with rrule for eigen does not give equal results. is the problem the sorting, or is there something mathematically...

That's great news, though I'm not sure in what sense zygote doesn't support this - their buffer type for example mutates in-place https://github.com/FluxML/Zygote.jl/blob/84bf62ea18330389c64d0d918c91d7b897e1a5d8/src/lib/buffer.jl

And in zygote, what happens when I naively define this pullback for a mutating function that does return the argument it modifies (and the underlying code always uses the return...

The way I used to implement mps+mps (apparantly I removed this method, I don't know why), was to explicitly construct the embedding maps (new_space

I would not be against making the TensorKitAD package public but unregistered?

https://github.com/quantumghent/TensorKitAD.jl should be publicly visible now? Feel free to open a bug report with any new errors you encounter, when I'm back from holiday I will look at them.

I have added a possible workaround in my fork, but it feels extremely hacky

I have quite a bit of code that I would like to differentiate through, but it seems fairly cumbersome to avoid mutation. As a workaround I tried this quick'n dirty...

Alternatively, is there a more canonical workaround? It seems more people require something like this ( https://gist.github.com/sdewaele/9a9c705eb4e8da9d798056d18d04c383 , https://discourse.julialang.org/t/zygote-and-setindex/44554 ). ------ I tried to fix it like : ``` function...

I have pushed a potential fix, but I'm not yet completely happy (I want entropy(finitemps) to just work, but currently it complains about the non-existence of similar(CRView)). I'm against using...