Real to Complex FFT

[tknopp]

I wonder why the FourierTransform operator uses a regular fft instead of an rfft in:
https://github.com/SciML/NeuralOperators.jl/blob/main/src/Transform/fourier_transform.jl#L10
Or am I missing something there?

If I read the code further, it seems that the negative frequencies are never recovered but keep zero after padding. This, however, would mean that the ifft will not generate a real signal and even the real part does not correspond to what one would expect when doing rfft and irfft.