Validation loop in PyTorch - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of a simple PyTorch validation loop

What will be the printed output after running this PyTorch validation loop snippet?

PyTorch

import torch
from torch import nn

model = nn.Linear(2, 1)
model.eval()

inputs = torch.tensor([[1.0, 2.0], [3.0, 4.0]])
targets = torch.tensor([[1.0], [2.0]])

criterion = nn.MSELoss()

with torch.no_grad():
    outputs = model(inputs)
    loss = criterion(outputs, targets)
    print(f"Loss: {loss.item():.4f}")

A Loss: 2.0000

B Loss: 1.0000

C Loss: 0.0000

D Loss: 0.5000

💡 Hint

Remember that the model is untrained and initialized randomly, so the loss won't be zero.

The model is a linear layer with random weights. The mean squared error loss between the model's output and targets will be around 0.5 for this input.

❓ Model Choice

intermediate

Choosing the correct model mode during validation

Which PyTorch model mode should be set during the validation loop to ensure correct behavior of layers like dropout and batch normalization?

A model.train()

B model.eval()

C model.freeze()

D model.validate()

💡 Hint

Think about which mode disables dropout and uses running statistics for batch norm.

During validation, model.eval() sets the model to evaluation mode, disabling dropout and using running statistics for batch normalization, which is the correct behavior.

❓ Metrics

advanced

Calculating average validation accuracy

Given a validation loop that accumulates correct predictions and total samples, which formula correctly computes the average validation accuracy?

A accuracy = total_correct / total_samples

B accuracy = total_samples / total_correct

C accuracy = total_correct * total_samples

D accuracy = total_correct - total_samples

💡 Hint

Accuracy is the fraction of correct predictions out of all predictions.

Accuracy is calculated as the number of correct predictions divided by the total number of samples evaluated.

🔧 Debug

advanced

Identifying the error in this validation loop snippet

What error will this PyTorch validation loop code raise?

PyTorch

model.eval()
correct = 0
total = 0

for inputs, labels in val_loader:
    outputs = model(inputs)
    predicted = torch.argmax(outputs, dim=1)
    correct += (predicted == labels).sum().item()
    total += labels.size(0)

accuracy = correct / total
print(f"Validation accuracy: {accuracy:.2f}")

A RuntimeError due to missing torch.no_grad() context

B ValueError because torch.argmax needs a dimension argument

C TypeError because predicted and labels have incompatible shapes

D No error, code runs correctly

💡 Hint

Check the shapes of predicted and labels tensors before comparison.

torch.argmax without dimension returns a single value, but labels is a batch tensor. Comparing them causes a shape mismatch and TypeError.

🧠 Conceptual

expert

Purpose of torch.no_grad() in validation loops

Why is it important to use torch.no_grad() during the validation loop in PyTorch?

A It disables gradient calculation to reduce memory usage and speed up computations

B It enables gradient calculation for validation metrics

C It resets model weights to initial values before validation

D It automatically switches the model to evaluation mode

💡 Hint

Think about what happens when gradients are not needed.

torch.no_grad() disables gradient tracking, which reduces memory use and speeds up validation since no backward pass is needed.