Following the softmax exercise shows me the following problem.

Could you help me with an explanation for my mistake?

thanks.

Hi, the x_sum sum each row of x_exp, so x_sum = np.sum (x_exp, axis = 1, keepdims = True)