pytorch计算一个矩阵每一行分别与另一个矩阵的差值

import torch

x = torch.randint(10, 20, (3,4))  # [3,1,4]
y = torch.randint(0, 9, (2,4))   # [2,1,4]

print(x)
print(y)

z = x.unsqueeze(1) - y.unsqueeze(0) # p[3, 2, 4]
'''
tensor([[10, 19, 14, 14],
        [17, 15, 16, 12],
        [15, 16, 12, 17]])
tensor([[7, 3, 3, 8],
        [7, 8, 5, 3]])
tensor([[[ 3, 16, 11,  6],
         [ 3, 11,  9, 11]],

        [[10, 12, 13,  4],
         [10,  7, 11,  9]],

        [[ 8, 13,  9,  9],
         [ 8,  8,  7, 14]]])
'''

上面的代码就是计算x矩阵的每一行分别与y矩阵的元素的差值。