分类目录：《深入浅出Pytorch函数》总目录
相关文章：
· 机器学习中的数学——激活函数：Softmax函数
· 深入浅出Pytorch函数——torch.softmax/torch.nn.functional.softmax
· 深入浅出Pytorch函数——torch.nn.Softmax

---

将Softmax函数应用于$n$维输入张量，重新缩放它们，使得$n$维输出张量的元素位于$[0,1]$的范围内，且总和为1。当输入张量是稀疏张量时，未指定的值被视为-inf。

语法

torch.nn.Softmax(dim=None)

参数

• dim：[int] Softmax函数将沿着dim轴计算，即沿dim的每个切片的和为1

返回值

与输入张量具有相同尺寸和形状的张量，且其元素值在$[0，1]$范围内。

实例

>>> m = torch.nn.Softmax(dim=1)
>>> input = torch.randn(2, 3)
>>> output = m(input)
tensor([[0.4773, 0.0833, 0.4395],[0.0281, 0.6010, 0.3709]])

函数实现

class Softmax(Module):r"""Applies the Softmax function to an n-dimensional input Tensorrescaling them so that the elements of the n-dimensional output Tensorlie in the range [0,1] and sum to 1.Softmax is defined as:.. math::\text{Softmax}(x_{i}) = \frac{\exp(x_i)}{\sum_j \exp(x_j)}When the input Tensor is a sparse tensor then the unspecifiedvalues are treated as ``-inf``.Shape:- Input: :math:`(*)` where `*` means, any number of additionaldimensions- Output: :math:`(*)`, same shape as the inputReturns:a Tensor of the same dimension and shape as the input withvalues in the range [0, 1]Args:dim (int): A dimension along which Softmax will be computed (so every slicealong dim will sum to 1)... note::This module doesn't work directly with NLLLoss,which expects the Log to be computed between the Softmax and itself.Use `LogSoftmax` instead (it's faster and has better numerical properties).Examples::>>> m = nn.Softmax(dim=1)>>> input = torch.randn(2, 3)>>> output = m(input)"""__constants__ = ['dim']dim: Optional[int]def __init__(self, dim: Optional[int] = None) -> None:super().__init__()self.dim = dimdef __setstate__(self, state):super().__setstate__(state)if not hasattr(self, 'dim'):self.dim = Nonedef forward(self, input: Tensor) -> Tensor:return F.softmax(input, self.dim, _stacklevel=5)def extra_repr(self) -> str:return 'dim={dim}'.format(dim=self.dim)