pytorch --- torch.cat()

最新推荐文章于 2025-02-14 01:24:14 发布

原创最新推荐文章于 2025-02-14 01:24:14 发布 · 2.5k 阅读

6 ·

CC 4.0 BY-SA版权

文章标签：

#pytorch

pytorch 专栏收录该内容

30 篇文章

订阅专栏

本文详细介绍了PyTorch中两种重要的张量操作：torch.cat和torch.stack。通过具体的代码示例，展示了如何使用这两种函数来连接张量，以及它们在维度处理上的不同之处。torch.cat用于沿指定维度拼接张量，而torch.stack则在原维度基础上新增一维进行堆叠。

部署运行你感兴趣的模型镜像

torch.cat((tensor1,tensor2), dim)

将两个tensor连接起来，具体如何连接见下面例子

x = torch.rand((2,2,3))
y = torch.rand((2,2,3))
print("x:",x)
print("y:",y)
print("dim=0:", torch.cat((x,y),dim=0))
print("dim=1:", torch.cat((x,y), dim=1))
print("dim=2:", torch.cat((x, y), dim=2))

输出：

x: tensor([[[0.2571, 0.9011, 0.7935],
         [0.9308, 0.3267, 0.3290]],

        [[0.6155, 0.4739, 0.7251],
         [0.8025, 0.0424, 0.8101]]])

y: tensor([[[0.8813, 0.1149, 0.7757],
         [0.4733, 0.9003, 0.3300]],

        [[0.2597, 0.5810, 0.2507],
         [0.1220, 0.2260, 0.5620]]])

dim=0: tensor([[[0.2571, 0.9011, 0.7935],
         [0.9308, 0.3267, 0.3290]],

        [[0.6155, 0.4739, 0.7251],
         [0.8025, 0.0424, 0.8101]],

        [[0.8813, 0.1149, 0.7757],
         [0.4733, 0.9003, 0.3300]],

        [[0.2597, 0.5810, 0.2507],
         [0.1220, 0.2260, 0.5620]]])

dim=1: tensor([[[0.2571, 0.9011, 0.7935],
         [0.9308, 0.3267, 0.3290],
         [0.8813, 0.1149, 0.7757],
         [0.4733, 0.9003, 0.3300]],

        [[0.6155, 0.4739, 0.7251],
         [0.8025, 0.0424, 0.8101],
         [0.2597, 0.5810, 0.2507],
         [0.1220, 0.2260, 0.5620]]])
         
dim=2: tensor([[[0.2571, 0.9011, 0.7935, 0.8813, 0.1149, 0.7757],
        		[0.9308, 0.3267, 0.3290, 0.4733, 0.9003, 0.3300]],

        [[0.6155, 0.4739, 0.7251, 0.2597, 0.5810, 0.2507],
         [0.8025, 0.0424, 0.8101, 0.1220, 0.2260, 0.5620]]])
[Finished in 2.1s]

torch.stack((tensor1, tensor2), dim)

x = torch.rand((2,2,3))
y = torch.rand((2,2,3))
print("x:",x)
print("y:",y)
print("dim=0:", torch.stack((x,y),dim=0))
print("dim=1:", torch.stack((x,y), dim=1))
print("dim=2:", torch.stack((x, y), dim=2))
print("dim=3", torch.stack((x, y), dim=3))

输出：

x: tensor([[[0.5099, 0.3434, 0.3731],
         [0.8523, 0.4672, 0.4163]],

        [[0.3364, 0.4910, 0.2302],
         [0.7896, 0.8119, 0.3978]]])

y: tensor([[[0.3843, 0.7627, 0.9757],
         [0.0065, 0.5462, 0.2765]],

        [[0.1890, 0.1698, 0.4486],
         [0.3459, 0.5552, 0.1908]]])

dim=0: tensor([[[[0.5099, 0.3434, 0.3731],
          [0.8523, 0.4672, 0.4163]],

         [[0.3364, 0.4910, 0.2302],
          [0.7896, 0.8119, 0.3978]]],


        [[[0.3843, 0.7627, 0.9757],
          [0.0065, 0.5462, 0.2765]],

         [[0.1890, 0.1698, 0.4486],
          [0.3459, 0.5552, 0.1908]]]])

dim=1: tensor([[[[0.5099, 0.3434, 0.3731],
          [0.8523, 0.4672, 0.4163]],

         [[0.3843, 0.7627, 0.9757],
          [0.0065, 0.5462, 0.2765]]],


        [[[0.3364, 0.4910, 0.2302],
          [0.7896, 0.8119, 0.3978]],

         [[0.1890, 0.1698, 0.4486],
          [0.3459, 0.5552, 0.1908]]]])
          
dim=2: tensor([[[[0.5099, 0.3434, 0.3731],
          [0.3843, 0.7627, 0.9757]],

         [[0.8523, 0.4672, 0.4163],
          [0.0065, 0.5462, 0.2765]]],


        [[[0.3364, 0.4910, 0.2302],
          [0.1890, 0.1698, 0.4486]],

         [[0.7896, 0.8119, 0.3978],
          [0.3459, 0.5552, 0.1908]]]])

dim=3 tensor([[[[0.5099, 0.3843],
          [0.3434, 0.7627],
          [0.3731, 0.9757]],

         [[0.8523, 0.0065],
          [0.4672, 0.5462],
          [0.4163, 0.2765]]],


        [[[0.3364, 0.1890],
          [0.4910, 0.1698],
          [0.2302, 0.4486]],

         [[0.7896, 0.3459],
          [0.8119, 0.5552],
          [0.3978, 0.1908]]]])
[Finished in 2.2s]