3.Pytorch 中 torch.stack(）/vstack()/hstack()和torch.cat()

1.torch.stack(）

torch.stack(tensors, dim=0, *, out=None) → Tensor

作用：

Concatenates a sequence of tensors along a new dimension. All tensors need to be of the same size.

把一系列tensor沿着新的维度堆起来。注意要tensor都一样的size，并且会增加一个维度。默认，dim=0.

x = torch.arange(9).view(3,3)
print(x)
print("---")
new_x = torch.stack([x, x, x])
print(new_x.shape)
print(new_x)

================================================================
tensor([[0, 1, 2],
        [3, 4, 5],
        [6, 7, 8]])
---
torch.Size([3, 3, 3])
tensor([[[0, 1, 2],
         [3, 4, 5],
         [6, 7, 8]],

        [[0, 1, 2],
         [3, 4, 5],
         [6, 7, 8]],

        [[0, 1, 2],
         [3, 4, 5],
         [6, 7, 8]]])

如果指定维数的话，

• dim=0时， 这个维度是3， 将其看作3行，那么特征数是2。将2个特征列依次叠加。
• dim=1时， 这个维度是2， 将其看作2列，那么特征数是3。将3个特征行依次叠加。

a = torch.arange(0, 6).view((3, 2))
b = torch.arange(6, 12).view((3, 2))
print('a:', a)
print('b:', b)
ab0= torch.stack((a, b), dim=0)
ab1 = torch.stack((a, b), dim=1)
print(ab0, '\n', ab1)

+++++++++++++++++++++++++++++++++++++++++++
a: tensor([[0, 1],
        [2, 3],
        [4, 5]])
b: tensor([[ 6,  7],
        [ 8,  9],
        [10, 11]])
tensor([[[ 0,  1],
         [ 2,  3],
         [ 4,  5]],

        [[ 6,  7],
         [ 8,  9],
         [10, 11]]]) 
 tensor([[[ 0,  1],
         [ 6,  7]],

        [[ 2,  3],
         [ 8,  9]],

        [[ 4,  5],
         [10, 11]]])

2. torch.vstack()和 torch.hstack()

torch.vstack(tensors, *, out=None) → Tensor 和hstack()

作用：这两个方法在1.8.0之后才支持，没有就用torch.cat()

在竖直、水平方向上堆tensor

ab_vstack_0 = torch.vstack((a, b))
ab_vstack_1 = torch.vstack((a, b))
print('ab_vstack_0 :', ab_vstack_0 )
print('ab_vstack_1 :', ab_vstack_1 )
print(torch.__version__)

++++++++++++++++++++++++++++++++++++++++++++
ab_vstack_0 : tensor([[ 0,  1],
        [ 2,  3],
        [ 4,  5],
        [ 6,  7],
        [ 8,  9],
        [10, 11]])
ab_vstack_1 : tensor([[ 0,  1],
        [ 2,  3],
        [ 4,  5],
        [ 6,  7],
        [ 8,  9],
        [10, 11]])
1.8.0+cu101

3. torch.cat()

torch.cat(tensors, dim=0, *, out=None) → Tensor与 torch.stack()区别是：不增加维度

作用：

Concatenates the given sequence of seq tensors in the given dimension. All tensors must either have the same shape (except in the concatenating dimension) or be empty.

按照指定维度连接tensor，所有tensor必须有同样的shape， 除了指定合并的维度或者是空tensor。

ab_cat_0 = torch.cat((a, b), dim=0)
ab_cat_1 = torch.cat((a, b), dim=1)
print('ab_cat_0 :', ab_cat_0 )
print('ab_cat_1 :', ab_cat_1 )

++++++++++++++++++++++++++++++++++++
ab_cat_0 : tensor([[ 0,  1],
        [ 2,  3],
        [ 4,  5],
        [ 6,  7],
        [ 8,  9],
        [10, 11]])
ab_cat_1 : tensor([[ 0,  1,  6,  7],
        [ 2,  3,  8,  9],
        [ 4,  5, 10, 11]])