Comment on page

nn.relu

   fn relu(tensor: @Tensor<T>) -> Tensor<T>;
Applies the rectified linear unit function element-wise
ReLU(x)=(x)+=max(0,x)ReLU(x)=(x)^+=max(0,x)ReLU(x)=(x)+=max(0,x)
Args
tensor(@Tensor<T>) - The input tensor.
Returns
A Tensor<T> with the same shape as the input tensor.
Examples
use array::{ArrayTrait, SpanTrait};
​
use orion::operators::tensor::{TensorTrait, Tensor, I32Tensor};
use orion::operators::nn::{NNTrait, I32NN};
use orion::numbers::{i32, IntegerTrait};
​
fn relu_example() -> Tensor<i32> {
    let tensor = TensorTrait::<i32>::new(
        shape: array![2, 2].span(),
        data: array![
            IntegerTrait::new(1, false),
            IntegerTrait::new(2, false),
            IntegerTrait::new(1, true),
            IntegerTrait::new(2, true),
        ]
            .span(),
    );
​
    return NNTrait::relu(@tensor);
}
>>> [[1,2],[0,0]]

Neural Network

nn.leaky_relu

Last modified 2mo ago