relu_op

→TensorImplPtr

int relu_op(const int & a)

Element-wise rectified linear unit $y = \max(0, x)$ .

See Also

ReluBackward : Autograd node implementing the gradient rule.

aTensorImplPtr

Input tensor of any real dtype.

TensorImplPtr

Output tensor with the same shape and dtype as a; autograd-aware when a.requires_grad is true.

Loading…