Numpy has changed its string representations. With numpy 1.26.4, I get:

>>> import numpy as np
>>> np.float64(3)
3.0

With numpy 2.0, though:

>>> import numpy as np
>>> np.float64(3)
np.float64(3.0)

This isn’t mentioned in the migration guide. It breaks tons of my doctests. I could update the doctests, but they would become harder to read. Is there a way to revert to the old way?