Why are two apparently equal distributions not equal?

Consider the following program

import tensorflow_probability as tfp

d = tfp.distributions.Normal(loc=[0], scale=[1])
d2 = tfp.distributions.Normal(loc=[0, 1], scale=[[0, 4],
                                                 [1, 3]])

assert d2[1, 0].loc == d.loc and d2[1, 0].scale == d.scale
print("d2[1, 0] == d?", d2[1, 0] == d)

which outputs

d2[1, 0] == d? False

Why this behavior? I haven't looked at the implementation, but I suppose you haven't overridden __eq__ and similar methods of the Normal class. If that's the case, you may want to do that, or is there any particular reason why you haven't done it?

Start by running the Python example with tensorflow_probability.distributions.Normal and inspect how Normal instances compare. Determine whether distributions with equivalent loc and scale should compare equal, and define the expected equality behavior before making a change; the issue is done when the behavior is documented and covered by an appropriate test.