Use native PyTorch complex number support

PyLops / pylops-gpu

PyLops-gpu – Linear Operators on GPUs

https://pylops-gpu.readthedocs.io/

GNU Lesser General Public License v3.0

12 stars 6 forks source link

Use native PyTorch complex number support #17

Open anjali411 opened 2 years ago

anjali411 commented 2 years ago

I noticed that you are using ComplexTensor and ComplexScalar classes to store complex numbers https://github.com/PyLops/pylops-gpu/blob/eed70f1973f01f31ef905fe09bb88c8783fd9681/pylops_gpu/utils/complex.py However, that's not necessary since latest releases of PyTorch (1.7 and after) support native complex support with specialized kernels for complex numbers for most PyTorch ops!

mrava87 commented 2 years ago

Thanks. If you are interested to contribute changes we will be happy to switch to native pytorch complex number.

However, note that we are not developing this library further because we have added full GPU support in pylops via a cupy backend. I am going to update the README file to make people aware of this :)