Adrian Seyboldt

I hope I don't seem like a broken record when saying this: But implementing the backward mode gradient operations through eg the schur decomposition isn't ideal. Let's say we want...

My summary from that chat (slightly edited, I hope I didn't break anything @brandonwillard ): We know that if we split the solve op and then chain the backward ops...

Is it possible that codecov doesn't pick up changes when I force-push? Somehow I still see the old version of the patch in the coverage report...

Comparison of the constant index case with the non-constant index case: ```python %env NUMBA_BOUNDSCHECK=0 import aesara import aesara.tensor as at import numpy as np n, k = 100_000, 100 idxs_vals...

This also raises the question, how we want to deal with out of bounds access by default. I'm not really comfortable with a default implementation that doesn't check bounds for...

I turned part of it into a rewrite, that makes it a bit cleaner. Apart from that I'm not really sure what extra code you are referring to.

numbas default for boundschecks is False, so unless we change that, this means that `at.as_tensor_variable(np.zeros(2))[3]` has undefined behavior. I think this is *terrible* API. We don't have this problem in...

I'm actually a bit shocked you would accept something in aesara where we access invalid memory for wrong user input by default. I am not going to remove boundchecks from...

I kind of hope we are just talking past each other here, so I'll just summarize a bit, and hopefully that helps: `AdvancedIncSubtensor1` is using *user defined* indices, so I...

I give up. Sorry @ricardoV94...