Applies a softmax followed by a logarithm.
nnf_log_softmax(input, dim = NULL, dtype = NULL)
input | (Tensor) input |
---|---|
dim | (int) A dimension along which log_softmax will be computed. |
dtype | ( |
While mathematically equivalent to log(softmax(x)), doing these two operations separately is slower, and numerically unstable. This function uses an alternative formulation to compute the output and gradient correctly.