本文原始碼基於版本1.0，互動介面基於0.4.1

import torch

按照指定軸上的座標進行過濾

index_select()
沿著某tensor的一個軸dim篩選若干個座標

>>> x = torch.randn(3, 4)		# 目標矩陣
>>> x
tensor([[ 0.1427,  0.0231, -0.5414, -1.0009],
        [-0.4664,  0.2647, -0.1228, -1.1068],
        [-1.1734, -0.6571,  0.7230, -0.6004]])
>>> indices =
 
 torch.tensor([0, 2])	# 在軸上篩選座標
>>> torch.index_select(x, dim=0, indices)	# 指定篩選物件、軸、篩選座標
tensor([[ 0.1427,  0.0231, -0.5414, -1.0009],
        [-1.1734, -0.6571,  0.7230, -0.6004]])
>>> torch.index_select(x, dim=1, indices)
tensor([[ 0.1427, -0.5414],
        [-0.4664, -0.1228],
        [-1.1734,  0.7230
 
]])

where()
用於將兩個broadcastable的tensor組合成新的tensor，類似於c++中的三元操作符“?:”

>>> x = torch.randn(3, 2)
>>> y = torch.ones(3, 2)
>>> torch.where(x > 0, x, y)
tensor([[1.4013, 1.0000],
        [1.0000, 0.9267],
        [1.0000, 0.4302]])
>>> x
tensor([[ 1.4013, -0.9960],
        [-0.3715,
 
  0.9267],
        [-0.7163,  0.4302]])

指定條件返回01-tensor

>>> x = torch.arange(5)   
>>> x
tensor([0, 1, 2, 3, 4])
>>> torch.gt(x,1)		# 大於
tensor([0, 0, 1, 1, 1], dtype=torch.uint8)
>>> x>1					# 大於
tensor([0, 0, 1, 1, 1], dtype=torch.uint8)
>>> torch.ne(x,1)		# 不等於
tensor([1, 0, 1, 1, 1], dtype=torch.uint8)
>>> x!=1				# 不等於
tensor([1, 0, 1, 1, 1], dtype=torch.uint8)
>>> torch.lt(x,3)		# 小於
tensor([1, 1, 1, 0, 0], dtype=torch.uint8)
>>> x<3					# 小於
tensor([1, 1, 1, 0, 0], dtype=torch.uint8)
>>> torch.eq(x,3)		# 等於
tensor([0, 0, 0, 1, 0], dtype=torch.uint8)
>>> x==3				# 等於
tensor([0, 0, 0, 1, 0], dtype=torch.uint8)

返回索引

>>> x = torch.arange(5)
>>> x   # 1維
tensor([0, 1, 2, 3, 4])
>>> torch.nonzero(x)
tensor([[1],
        [2],
        [3],
        [4]])
>>> x = torch.Tensor([[0.6, 0.0, 0.0, 0.0],[0.0, 0.4, 0.0, 0.0],[0.0, 0.0, 1.2, 0.0],[0.0, 0.0, 0.0,-0.4]])
>>> x   # 2維
tensor([[ 0.6000,  0.0000,  0.0000,  0.0000],
        [ 0.0000,  0.4000,  0.0000,  0.0000],
        [ 0.0000,  0.0000,  1.2000,  0.0000],
        [ 0.0000,  0.0000,  0.0000, -0.4000]])
>>> torch.nonzero(x)
tensor([[0, 0],
        [1, 1],
        [2, 2],
        [3, 3]])

藉助nonzero()我們可以返回符合某一條件的index（https://stackoverflow.com/questions/47863001/how-pytorch-tensor-get-the-index-of-specific-value）

>>> x=torch.arange(12).view(3,4)
>>> x
tensor([[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]])
>>> (x>4).nonzero()
tensor([[1, 1],
        [1, 2],
        [1, 3],
        [2, 0],
        [2, 1],
        [2, 2],
        [2, 3]])