Week 8: PyTorch

Objective

Rebuild earlier ideas with a real deep learning framework while still keeping the model small and understandable.

Required Videos

• PyTorch 60-minute blitz: tensors, autograd, and training loop sections
• Optional: Karpathy or another framework demo after the code makes sense

Tasks

1. Run python week8_pytorch.py.
2. Inspect the printed training progress and final accuracy.
3. Change one hyperparameter such as learning rate, hidden size, or epochs.
4. Compare the result to the original run.

Deliverables

• One successful PyTorch training run
• A note describing what PyTorch automated for you
• One hyperparameter comparison

Checkpoint Questions

• What does autograd save you from doing manually?
• What are tensors doing differently from plain Python lists?
• Why do we still need a loss function and optimizer?
• What changed when you tuned a hyperparameter?