Python Tensor计算求和时，怎样保持轴数不变

目录

1.求和 sum()

2.求和的时候，保持轴数不变，需要用到keepdim

1.求和 sum()

import torch
x = torch.arange(12).reshape(3,4)
x,x.sum(axis=0),x.sum(axis=1) # axis=0,表示竖着计算和，axis=1表示横着计算和

#运行结果
x:  (tensor([[ 0,  1,  2,  3],
         [ 4,  5,  6,  7],
         [ 8,  9, 10, 11]]), 

x.sum(axis=0):  tensor([12, 15, 18, 21]), 
x.sum(axis=1):  tensor([ 6, 22, 38]))

2.求和的时候，保持轴数不变，需要用到keepdim

#计算总和或均值时保持轴数不变

import torch
x = torch.arange(12).reshape(3,4)
Sum_x = x.sum(axis=0,keepdim=True)
Sum_x
Sum_x2 = x.sum(axis=1,keepdim=True)
Sum_x,Sum_x2


#运行结果
(tensor([[12, 15, 18, 21]]), 

tensor([[ 6],
         [22],
         [38]]))

只有保持轴数不变，才能计算x /Sum_x, 或x/Sum_x2