Type annotations for Approximations

[kamathhrishi]

Where?

Currently, the approximations module of SyMPC does not have type annotations. The following are the files with missing type annotations:

• [ ] exponential.py
• [ ] log.py
• [ ] reciprocal.py
• [ ] sigmoid.py
• [ ] softmax.py
• [ ] tanh.py
• [ ] utils.py

[rhit-namburs]

Hi I would like to contribute to this. Note that this is my first time contributing to a open source project.

[rasswanth-s]

Sure @rhit-namburs , go for it :+1:

[rasswanth-s]

@rhit-namburs are you still working on it?

[manel1874]

Hi! I am not sure if @rhit-namburs is still working on this issue so I decided to take a look.

I noticed that softmax.py and tanh.py are already annotated, so I decided to follow their approach. However, this leads to a circular import issue:

1. sympc.tensor.mpc_tensor.py imports APPROXIMATIONS from sympc.approximations, which includes func = (sigmoid.py, log.py, exponential.py, reciprocal.py, utils.py).
2. softmax.py and tanh.py import MPCTensor from sympc.tensor to be used for the type annotations. If we do the same for func we go into a loop of imports.

Without restructuring both sympc.approximations folder and MPCTensor class, I think the easiest way to go around this issue is to use TypeVar from typing to define an MPCTensor annotation. I tested and this solution passes the pre-commit tests.

Another possible solution could be to include the approximation functions defined in sympc.approximations folder inside MPCTensor class.