Dropout

Train/eval mode, inverted scaling

Easy Fundamentals

Problem Description

Implement Dropout regularization from scratch.

Signature

class MyDropout(nn.Module):
    def __init__(self, p: float = 0.5): ...
    def forward(self, x: Tensor) -> Tensor: ...

Rules

• During training: zero each element with probability p, scale remaining by 1/(1-p)

• During eval: return input unchanged (identity)

• Do NOT use nn.Dropout or F.dropout

Template

Implement the function below. Use only basic PyTorch operations.

# ✏️ YOUR IMPLEMENTATION HERE

class MyDropout(nn.Module):
    def __init__(self, p=0.5):
        super().__init__()
        pass

    def forward(self, x):
        pass

Test Your Implementation

Use this code to debug before submitting.

# 🧪 Debug
d = MyDropout(p=0.5)
d.train()
x = torch.ones(10)
print('Train:', d(x))
d.eval()
print('Eval: ', d(x))

Reference Solution

Try solving it yourself first! Click below to reveal the solution.

# ✅ SOLUTION

class MyDropout(nn.Module):
    def __init__(self, p=0.5):
        super().__init__()
        self.p = p

    def forward(self, x):
        if not self.training or self.p == 0:
            return x
        mask = (torch.rand_like(x) > self.p).float()
        return x * mask / (1 - self.p)

Tips

Run Locally

For interactive practice with auto-grading, run TorchCode locally:
pip install torch-judge then use check("dropout")

Key Concepts

Train/eval mode, inverted scaling

Cross-Entropy Loss Embedding

Dropout

Description Template Test Solution Tips