pytorch中沿着维度操作的理解torch.any 和 torch.sum 为列子

 

沿着某个维度操作，即压缩指定的维度，而其它维度保持不变

 

In [3]:

import torch
from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = "all"

a = torch.randint(0,10,[4,3]) < 5
a
torch.any(a,-1,True)
torch.any(a,0,True)

 

Out[3]:

tensor([[False,  True,  True],
        [ True, False,  True],
        [False, False,  True],
        [ True,  True,  True]])

Out[3]:

tensor([[True],
        [True],
        [True],
        [True]])

Out[3]:

tensor([[True, True, True]])

 

In [4]:

a = torch.randint(0,2,[4,3,3])
a
torch.any(a,1,True)   #shape (4,1,3)

 

Out[4]:

tensor([[[1, 1, 1],
         [0, 1, 0],
         [1, 0, 1]],

        [[1, 1, 0],
         [1, 0, 1],
         [1, 0, 0]],

        [[1, 1, 1],
         [1, 0, 0],
         [0, 0, 1]],

        [[1, 0, 1],
         [1, 1, 1],
         [1, 1, 0]]])

Out[4]:

tensor([[[True, True, True]],

        [[True, True, True]],

        [[True, True, True]],

        [[True, True, True]]])

 

In [6]:

a = torch.randint(0,5,[4,3,3])
a
torch.sum(a,1,True)

 

Out[6]:

tensor([[[3, 4, 0],
         [4, 4, 0],
         [2, 2, 2]],

        [[0, 0, 3],
         [1, 3, 4],
         [4, 3, 1]],

        [[0, 2, 2],
         [1, 3, 4],
         [0, 3, 3]],

        [[1, 2, 2],
         [0, 0, 4],
         [1, 1, 3]]])

Out[6]:

tensor([[[ 9, 10,  2]],

        [[ 5,  6,  8]],

        [[ 1,  8,  9]],

        [[ 2,  3,  9]]])