class

GS::Autograd::ReluBackward

Inherits GS::Autograd::GradFn < Reference < Object

Gradient for ReLU: out = max(0, x) d(relu)/dx = 1 if x > 0 else 0

Constructors

new(input_data : Tensor)
Source

Instance methods

backward(grad_output : Tensor) : Array(Tensor | Nil)

Compute gradients w.r.t. inputs given gradient of output Returns array of gradients, one per input (or nil if input doesn't require grad)

Source