二、Softmax函数

Softmax函数

代码：

import torch
input = torch.randn(2,2,3,3) # 生成随机张量
print("input",input)
print("input.size",input.size())
output = torch.nn.Softmax(dim=1)(input) # 沿着维度1进行Softmax计算
print("output", output)
print("output.size", output.size())

结果：

input tensor([[[[-1.0284,  0.3340,  0.8394],
          [ 2.0868, -0.4440,  1.1711],
          [-0.4021, -1.5332,  1.3740]],

         [[-1.7498, -2.3110, -0.9069],
          [ 1.3210, -2.1890,  0.8349],
          [-0.0612,  0.3885, -1.3221]]],


        [[[ 0.9202,  0.4145,  0.5510],
          [-1.2614,  1.1752, -0.7427],
          [ 0.1773, -0.1683, -0.4466]],

         [[ 1.7270,  0.7876, -0.0378],
          [ 1.1749,  0.1121, -0.0742],
          [-1.4793, -0.5104,  1.9719]]]])
input.size torch.Size([2, 2, 3, 3])
output tensor([[[[0.6729, 0.9337, 0.8515],
          [0.6826, 0.8513, 0.5833],
          [0.4156, 0.1277, 0.9368]],

         [[0.3271, 0.0663, 0.1485],
          [0.3174, 0.1487, 0.4167],
          [0.5844, 0.8723, 0.0632]]],


        [[[0.3086, 0.4078, 0.6431],
          [0.0804, 0.7433, 0.3388],
          [0.8398, 0.5847, 0.0818]],

         [[0.6914, 0.5922, 0.3569],
          [0.9196, 0.2567, 0.6612],
          [0.1602, 0.4153, 0.9182]]]])
output.size torch.Size([2, 2, 3, 3]) # 维度1之和为1