Question

我有一个方矩阵，喜欢将其分解为一些较小的矩阵。例如，假设我们有一个[4,4]形状的矩阵，并想将其转换为大小为[2,2]的4个较小的矩阵。

输入：

[9, 9, 9, 9,
 8, 8, 8, 8,
 7, 7, 7, 7,
 6, 6, 6, 6]

输出：

[[9, 9  | [9, 9,
 8, 8]  | 8, 8],
 ---------------
 [7, 7  | [7, 7,
 6, 6]  | 6, 6]]

Answer 1

为此，您可以重复调用torch.split。

>>> x
tensor([[ 1,  2,  3,  4],
        [ 5,  6,  7,  8],
        [ 9, 10, 11, 12],
        [13, 14, 15, 16]])
>>> [z for y in x.split(2) for z in y.split(2, dim=1)]
[tensor([[1, 2],
        [5, 6]]), tensor([[3, 4],
        [7, 8]]), tensor([[ 9, 10],
        [13, 14]]), tensor([[11, 12],
        [15, 16]])]

Answer 2

给出具有4*4或1*16形状的张量，最简单的方法是通过视图函数或整形：

a = torch.tensor([9, 9, 9, 9, 8, 8, 8, 8, 7, 7, 7, 7, 6, 6, 6, 6])
# a = a.view(4,4)
a = a.view(2, 2, 2, 2)

# output:
tensor([[[[9, 9],
          [9, 9]],

         [[8, 8],
          [8, 8]]],


        [[[7, 7],
          [7, 7]],

         [[6, 6],
          [6, 6]]]])

Mat2cell Matlab在张量流或pytorch中等效

2 个答案: