独热编码
one_hot(tensor, num_classes=-1) -> LongTensor

tensor:要编码的张量

num_calsses:类别数，就是下面例子中的列数

import torch
from torch.nn import functional as F

1、无num_classes 

列数与最大值有关:0 -> max,比如下面例子中最大值为8，所以0 -> 8,共9列（相当于labels共9类）
行数与数的个数有关，该例子共8个数，所以8行。
所以shape=(8,9)

x = torch.tensor([1, 1, 1, 3, 3, 4, 8, 5])

y1 = F.one_hot(x) # 只有一个参数张量x
print(f'x = ',x)  
print(f'x_shape = ',x.shape) 
print(f'y1 = ',y1)  
print(f'y1_shape = ',y1.shape)

结果 

x = tensor([1, 1, 1, 3, 3, 4, 8, 5])
x_shape = torch.Size([8])
y = tensor([[0, 1, 0, 0, 0, 0, 0, 0, 0],
        [0, 1, 0, 0, 0, 0, 0, 0, 0],
        [0, 1, 0, 0, 0, 0, 0, 0, 0],
        [0, 0, 0, 1, 0, 0, 0, 0, 0],
        [0, 0, 0, 1, 0, 0, 0, 0, 0],
        [0, 0, 0, 0, 1, 0, 0, 0, 0],
        [0, 0, 0, 0, 0, 0, 0, 0, 1],
        [0, 0, 0, 0, 0, 1, 0, 0, 0]])
y_shape = torch.Size([8, 9])

 

2、有num_classes

num_classes参数可以自行设置列数

y2 = F.one_hot(x, num_classes = 10)  # 这里num_classes设置为10，其中 10 > max{x}
print(f'x = ',x)  # 输出 x
print(f'x_shape = ',x.shape)  # 查看 x 的形状
print(f'y2 = ',y2)  # 输出 y
print(f'y2_shape = ',y2.shape)

结果

x =  tensor([1, 1, 1, 3, 3, 4, 8, 5])
x_shape =  torch.Size([8])
y2 =  tensor([[0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
        [0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
        [0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
        [0, 0, 0, 1, 0, 0, 0, 0, 0, 0],
        [0, 0, 0, 1, 0, 0, 0, 0, 0, 0],
        [0, 0, 0, 0, 1, 0, 0, 0, 0, 0],
        [0, 0, 0, 0, 0, 0, 0, 0, 1, 0],
        [0, 0, 0, 0, 0, 1, 0, 0, 0, 0]])
y2_shape =  torch.Size([8, 10])