【Torch API】pytorch 中index_copy_函数详解

Tensor.index_copy_按照index索引，将Tensor的元素复制到张量中

 

    def index_copy_(self, dim, index, tensor): # real signature unknown; restored from __doc__
        """
        index_copy_(dim, index, tensor) -> Tensor
        
        Copies the elements of :attr:`tensor` into the :attr:`self` tensor by selecting
        the indices in the order given in :attr:`index`. For example, if ``dim == 0``
        and ``index[i] == j``, then the ``i``\ th row of :attr:`tensor` is copied to the
        ``j``\ th row of :attr:`self`.
        
        The :attr:`dim`\ th dimension of :attr:`tensor` must have the same size as the
        length of :attr:`index` (which must be a vector), and all other dimensions must
        match :attr:`self`, or an error will be raised.
        
        .. note::
            If :attr:`index` contains duplicate entries, multiple elements from
            :attr:`tensor` will be copied to the same index of :attr:`self`. The result
            is nondeterministic since it depends on which copy occurs last.
        
        Args:
            dim (int): dimension along which to index
            index (LongTensor): indices of :attr:`tensor` to select from
            tensor (Tensor): the tensor containing values to copy
        
        Example::
        
            >>> x = torch.zeros(5, 3)
            >>> t = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=torch.float)
            >>> index = torch.tensor([0, 4, 2])
            >>> x.index_copy_(0, index, t)
            tensor([[ 1.,  2.,  3.],
                    [ 0.,  0.,  0.],
                    [ 7.,  8.,  9.],
                    [ 0.,  0.,  0.],
                    [ 4.,  5.,  6.]])
        """
        return _te.Tensor(*(), **{})

• dim(int) 需要插入的索引维度
• index(张量) 被插入的索引
• tensor 包含要复制的值的张量

官方注释例子讲解

这里x是指被复制的张量 ，如例子中是一个 5×3 的矩阵
t 里面包含了需要被插入的值， 如

t = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]]

我们希望把t[0]、t[1]、t[2] 分别复制到 x的 第0、第4、第2 维度
将index设置为 index = torch.tensor([0, 4, 2]) 即可

官方例子如下：

x = torch.zeros(5, 3)
t = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]], dtype=torch.float)
index = torch.tensor([0, 4, 2])
x.index_copy_(0, index, t)

输出

tensor([[ 1.,  2.,  3.],
                [ 0.,  0.,  0.],
                [ 7.,  8.,  9.],
                [ 0.,  0.,  0.],
                [ 4.,  5.,  6.]])