lukasheinrich
commented
5 years ago I did at some point and didn't see much improvement. I think opt.einsum is useful if you do einsums with more than two tensors which I think we don't. (but maybe we can rewrite things in this way?) definitely worth looking at again
Description
See opt-einsum here: https://optimized-einsum.readthedocs.io/en/latest/ . A large majority of the mathematical code we write, especially for tensorization, relies on being able to use einsum effectively. It seems possible to get speed-ups, even there.