TensorNetwork complex support on pytorch backend

I am a physicist and I have recently started working with TensorNetwork for my simulations. With the NumPy backend, everything is great. However, the pytorch_backend.py seems to only support real tensors: the backend implementation of ‘conj’ is just returning the tensor, assuming that PyTorch does not support complex numbers.
However, it seems like PyTorch has recently started supporting complex numbers - https://github.com/pytorch/pytorch/issues/33152
Is it possible to change the support in the backend? As far as I can tell, only the conj and sign function need to be adjusted, and it should be quite easy.

Thank you,

1 Like