Structs

  • | Softsign takes one input data tensor | $X$ and produces one output data $Y,$ | where the softsign function, $y = \frac{x}{1+ | |x|}$, is applied to $X$ elementwise. | | This operation can be done in an in-place | fashion too, by providing the same input | and output blobs. | | Github Links: | | - https://github.com/pytorch/pytorch/blob/master/caffe2/operators/softsign_op.cc |
  • | Calculates the softsign gradient (sgn(x)/(1+|x|)^2) | of the given input tensor element-wise. |