Exercises

Derive a gradient by hand; check it numerically; diagnose an unstable run; compare embedding metrics; inspect tokenization across languages; compute attention manually; test a causal mask; implement a transformer block; compare adaptation choices; measure quantization; profile batching; run an ablation; and route a deterministic task away from the model.

Evidence and review

The final transfer exercise provides an unfamiliar architecture and incomplete documentation. Claims must trace to versioned code and artifacts. Screenshots, unexplained generated prose, and a single favorable run do not pass.

Oral defense

Explain the mechanism, identify the most dangerous assumption, interpret one failure from raw evidence, and name the condition that would reverse your decision.

Source backbone

Use Dive into Deep Learning, PyTorch documentation, primary papers, and Designing Data-Intensive Applications selectively.