Potential testing bug

[rossbar]

I noticed the following line:

deepcell-toolbox/deepcell_toolbox/deep_watershed_test.py

Line 42 in 73e8861

other = np.random.randint(0, 1, size=shape)

which stands out as a potential issue in that the upper bound for randint is exclusive, so this is equivalent (but slower than) np.zeros. I suspect the intent was a random array of 0s and 1s, which would be randint(0, 2, size=shape). In practice I don't think it matters - it looks like others is only used to test the ordering of the inputs, so the contents of the array don't really matter. In which case I think just explicitly changing to np.zeros(shape, dtype=int) makes sense - just wanted to draw attention to it in case there was indeed an intention to have a binary integer array here.