factorisation subpaterns, float compatible, inplace and memory optimisation....

[robinechuca]

Hi, I've also coded a function equivalent to this one on my side.
Thank you very much for publishing your project as open source.
One of the problems with open-source is its the diversification.
Our 2 projects seem very complementary, as they both have different strengths. Yours is more torch-friendly with elegant integration, mine is more optimized on the points described in the titles. I'd be happy to merge my project with yours. That way, I can delete mine. It's better to have one single library that does everything well than 2 that do everything halfway!

Here's the source code for my project, which is perhaps a little more "sympy" than "torch" oriented:
https://framagit.org/robinechuca/cutcutcodec/-/blob/main/cutcutcodec/core/compilation/sympy_to_torch.py

Are you interested to implement some of this optimizations in your project?