2025 USA-NA-AIO Round 1, Problem 2, Part 10

AI Olympiads 2025 USA-NA-AIO Round 1
1

Part 10 (5 points, coding task)

Rectified Linear Unit, or the “ReLU”, is one of the most common used function in deep learning. It is defined as

\text{ReLU}(x)= \max \left\{ 0,x \right\} .

In this part, you are asked to use PyTorch to build a ReLU class named My_ReLU that subclasses nn.module.

A successful class works in the following ways:

  • The initialization of an object in My_ReLU does not take any input.

  • Suppose we have a My_ReLU object called activiation0. When we call activiation0(x)with input x that is a tensor x with an arbitrary dimension and shape, we get an output y from the element-wise ReLU acvitivation on x.

2 
### WRITE YOUR SOLUTION HERE ###

class My_ReLU(nn.Module):
    def __init__(self):
        super().__init__()

    def forward(self, x):
        return torch.max(torch.zeros_like(x), x)

""" END OF THIS PART """