Applies a softmax followed by a logarithm.

## Arguments

- input
(Tensor) input

- dim
(int) A dimension along which log_softmax will be computed.

- dtype
(

`torch.dtype`

, optional) the desired data type of returned tensor. If specified, the input tensor is casted to`dtype`

before the operation is performed. This is useful for preventing data type overflows. Default:`NULL`

.