A strange production of a tensor and a matrix-CodePudding

When I have a tensor and a matrix below:

A = np.array([[['a', 'b'], ['c', 'd']], [['e', 'f'], ['g','h']]], dtype=object)
x = np.array([[1, 2], [3, 4]])

How to generate a tensor shown below?

array([[['abb' 'eff'],
        ['cdd' 'ghh']],

       [['aaabbbb' 'eeeffff'],
        ['cccdddd' 'ggghhhh']]]

I examined np.tensordot(A, x, 1) and A@x with .transpose(,,), but they looked other forms.

CodePudding user response：

You just need to keep transposing!

>>> (A @ x.T).T
array([[['abb', 'eff'],
        ['cdd', 'ghh']],

       [['aaabbbb', 'eeeffff'],
        ['cccdddd', 'ggghhhh']]], dtype=object)