Lukas Devos

> > 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...

Can I maybe add that this still exists, and also is not consistent? It works for `-()`, but not for `conj`, `real`, `imag`. ```julia-repl julia> A = Fill(1im) 0-dimensional Fill{Complex{Int64}},...

Sadly it's quite a bit of work since cuTENSOR has changed their interface. It's definitely somewhere on my to do list, but for now I think cuTENSOR v1 works just...

I started some work on moving to the new interface. I think it should be working for plain `CuArray`s, but I am still deciding on how to implement views/stridedviews, so...

I also just noticed that cuTENSOR 2 requires julia 1.8, which I am not too happy about. I think this means we either need to keep two different versions of...

Awaiting the result of https://github.com/JuliaGPU/CUDA.jl/pull/2356 to simplify the implementation further.

My main argument would be that `@tensor` does in fact automatically insert `tensorscalar`, thus this would improve consistency in that regard ```julia-repl julia> using TensorOperations julia> A = rand(2,2); julia>...

I think I agree with setting the default values as stable as possible, and afterwards trying to be able to incorporate the option of having the faster version. The time...

I think in any case we should look for a solution to determine how the `rrule` is computed, even just changing the tolerances should be accessible. Also, I think @pbrehmer...

> @Gertian Time evolving a finite window in a bath of the groundstate is already possible in MPSKit. This PR aims to extend the time evolution so that the infinite...