[Python] np.where on a numpy MxN matrix but return M rows with indices where condition exists

I am trying to use np.where on a MxN numpy matrix, where I want to return the same number of M rows but the indices in each row where the element exists. Is this possible to do so? For example:

a = [[1 ,2, 2]
[2, 3, 5]]

np.where(a == 2)


I would like this to return:

[[1, 2],
[0]]

Continue reading...