LogSoftmax# class torch.nn.LogSoftmax(dim=None)[source]# Applies the log(Softmax(x))\log(\text{Softmax}(x))log(Softmax(x)) function to an n-dimensional input Tensor. The LogSoftmax formulation can be simplified as: LogSoftmax(xi)=log(exp(xi)∑jexp(xj))\text{LogSoftmax}(x_{i}) = \log\left(\frac{\exp(x_i) }{ \sum_j \exp(x_j)} \right) LogSoftmax(xi)=log(∑jexp(xj)exp(xi)) Shape: Input: (∗)(*)(∗) where * means, any number of additional dimensions Output: (∗)(*)(∗), same shape as the input Parameters dim (int) – A dimension along which LogSoftmax will be computed. Returns a Tensor of the same dimension and shape as the input with values in the range [-inf, 0) Return type None Examples: >>> m = nn.LogSoftmax(dim=1) >>> input = torch.randn(2, 3) >>> output = m(input) extra_repr()[source]# Return the extra representation of the module. Return type str forward(input)[source]# Runs the forward pass. Return type Tensor
