LESSON 2.9 · Assessment · 180 min
Mastery Gate 2: build minBPE + makemore
Mastery Gate 2: build minBPE + makemore. Learn the complete tokenizer and language-model loop and complete: Train on unseen data, evaluate, and explain failure cases. Part of the “
Learning objectives
- Explain what problem “Mastery Gate 2: build minBPE + makemore” solves without hiding behind terminology.
- Trace the variables and causal links across the complete tokenizer and language-model loop.
- Complete “Train on unseen data, evaluate, and explain failure cases” and judge the result with evidence rather than intuition.
Core concepts
the complete tokenizer and language-model loop
the complete tokenizer and language-model loop is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment before moving to an optimized implementation.
Build and verify
Train on unseen data, evaluate, and explain failure cases
- Predict: write the expected output, trend, or failure before running code.
- Build: implement only the minimum components needed to answer the question.
- Verify: compare with a baseline or trusted implementation; save seeds, parameters, and raw outputs.
- Transfer: change one shape, dataset, scale, or workload condition and explain whether the conclusion still holds.