The tests fail because of this (significant) diff in the data ...

@dstansby I claim a fix here 🚀

The

OverflowError: Python integer -1 out of bounds for uint8

was fixed by -1 → -1.0 🙃

I've somehow managed to push to your fork, so I can approve this, but I'll wait foryou to click merge.
(Also, we need to be sure to squash merge because there are images changed and changed back in the history here.)

pre-commit.ci autofix

pre-commit.ci run

I think the mypy error is because pre-commit is running a version of numpy (I can't work out which...) that doesn't have ashape argument tonp.reshape(), which indeed seems only to have been added in 2.1.0:numpy/numpy#26292. I think we want to support older numpy versions than that, so I think the fix is to revert to using thenewshape argument, and ignoring the deprecation warning (for now, until we depend on numpy>=2.1)

so I think the fix is to revert to using thenewshape argument, and ignoring the deprecation warning (for now, until we depend on numpy>=2.1)

Thanks for the detective work! I was just playing 🔨