leaky_relu_op

→TensorImplPtr

int leaky_relu_op(const int & a, double slope)

Leaky ReLU $y = x$ if $x > 0$ else $\text{slope} \cdot x$ .

See Also

LeakyReluBackward : Autograd node implementing the gradient rule.

aTensorImplPtr

Input tensor.

slopedouble

Negative-side slope; 0.01 is the standard default.

TensorImplPtr

Output tensor with the same shape as a.

Loading…