想要将Numpy的结果应用于权重矩阵[Python]

Question

我想对循环神经网络权重矩阵使用numpy where来更新大于阈值的值。

更新pre-> post重量还可以。

但是更新帖子->重量不起作用。

擅长numpy的人，请帮助我！

>>> import numpy as np
>>> v = np.arange(3*2*2).reshape((3,2,2))
"""
array([[[ 0,  1],
        [ 2,  3]],

       [[ 4,  5],
        [ 6,  7]],

       [[ 8,  9],
        [10, 11]]])
"""
>>> w = np.arange(3*2*2*3*2*2).reshape((3,2,2,3,2,2))
"""
array([[[[[[  0,   1],
           [  3,   3]],

          [[  6,   5],
           [  9,   7]],

          [[ 12,   9],
           [ 15,  11]]],


         [[[ 18,  13],
           [ 21,  15]],

          [[ 24,  17],
           [ 27,  19]],

          [[ 30,  21],
           [ 33,  23]]]],
~~~~~~~~~~~~~~~~~~~~~~~~~~
          [[144, 137],
           [147, 139]],

          [[150, 141],
           [153, 143]]]]]])
"""
>>> w[np.where(v>3)] += 1 # pre -> post is OK.
>>> w[:,:,:, np.where(v>3)] += 1 # post -> pre is not working!! I can't understand this result.
# incremented all elements!!