Randomized leaky ReLU.
Usage
nnf_rrelu(input, lower = 1/8, upper = 1/3, training = FALSE, inplace = FALSE)
nnf_rrelu_(input, lower = 1/8, upper = 1/3, training = FALSE)
Arguments
- input
(N,*) tensor, where * means, any number of additional dimensions
- lower
lower bound of the uniform distribution. Default: 1/8
- upper
upper bound of the uniform distribution. Default: 1/3
- training
bool wether it's a training pass. DEfault: FALSE
- inplace
can optionally do the operation in-place. Default: FALSE