Daniel Smith comments

Results 258 comments of


                                            Daniel Smith

Optimal path less efficient than greedy path

Thank you for the report, this is quite odd behavior and defiantly a bug. Even setting `memory_limit=-1` does not find the correct path. For now I would recommend the `dp`...

Fixes `operands could not be broadcast` error when the subscripts contain ellipsis and the operands are shapes.

This is a subtle bug related to a blocking issue in #204, where `possibly_convert_to_numpy`'s behavior is incorrect and we're not quite normalizing on internal representations the way we should. I'll...

Fixes `operands could not be broadcast` error when the subscripts contain ellipsis and the operands are shapes.

Yes, I can confirm #204 fixes it: https://github.com/dgasmith/opt_einsum/pull/204/commits/602e61ba38f43d4ab06501156ad6abd7ce002382

Fixes `operands could not be broadcast` error when the subscripts contain ellipsis and the operands are shapes.

Closed by #204.

opt_einsum.contract is slower than raw numpy.einsum

Closing as inactive.

Fix: _einsum backend handling when optimize=False

Not sure why CI isn't triggering- can you push again as well as provide a test?

upgrade optimal and greedy algorithms with versions from cotengra

I wonder if we can get these to be an exact drop in replacement. A few thoughts: - Can we add memory_limit support? - For 1) how do we view...

Add `--converge maxiter` option to conda release of geometric

@leeping Happy to help guide this through the releases as I believe I'm still a conga-forge maintainer. PS: 👋 it's good to see you again!