Prerequisites
Python programming. Calculus (derivatives, chain rule, gradients). Linear algebra (eigenvalues, matrix decomposition). Probability and statistics (Bayes' theorem, distributions). Recommended: one introductory ML course.
Focus: Architecture internals, training methods, interpretability. Students leave with a deep understanding of how LLMs work and how to study them. This pathway trades breadth for depth: it covers the same foundations as Course A but then dives into pre-training, scaling laws, PEFT, and alignment. The reasoning is that future researchers need to understand the training pipeline end to end, since that is where novel contributions happen.
14-Week Syllabus
| Week | Topics | Lab / Assignment |
|---|---|---|
| 1 | ML and PyTorch Foundations | Build and train an image classifier in PyTorch |
| 2 | NLP, Text Representation, Tokenization (Ch 01 through 02) | Compare tokenizer vocabulary coverage across languages |
| 3 | Sequence Models and Attention | Implement attention from scratch, visualize attention weights |
| 4 | The Transformer Architecture | Build a minimal transformer (encoder + decoder) |
| 5 | Decoding Strategies | Implement nucleus sampling, measure diversity vs. quality |
| 6 | Pre-training and Scaling Laws | Reproduce a scaling law curve on a small model |
| 7 | Modern LLM Landscape and Reasoning Models (Ch 07 through 08) | Compare model architectures (paper reading assignment) |
| 8 | Inference Optimization | Benchmark KV-cache and quantization effects |
| 9 | Synthetic Data Generation | Generate and validate a synthetic training dataset |
| 10 | Fine-Tuning and PEFT (Ch 14 through 15) | Compare full fine-tuning vs. LoRA on the same task |
| 11 | Alignment (RLHF, DPO) | Implement DPO training on a preference dataset |
| 12 | Interpretability | Probe internal representations with logit lens |
| 13 | Emerging Architectures and AI and Society (Ch 34 through 35) | Write a research proposal on an open problem |
| 14 | Final project presentations | Research paper replication or extension (individual project) |
- Appendix D: Environment Setup – set up your development environment before Week 1
- Appendix K: HuggingFace: Transformers, Datasets, and Hub – access pretrained models and datasets for labs
- Appendix A: Math Foundations – review the linear algebra and probability behind attention
- Appendix R: Experiment Tracking – log experiments systematically for your research project
What Comes Next
Return to the Course Syllabi overview to explore other courses and reading tracks, or proceed to FM.4: How to Use This Book for a quick orientation on conventions and callout types.