This appendix is a compact quick-reference companion to Chapter 0: ML and PyTorch Foundations. It distills the key tables, formulas, and cheat sheets from that chapter into a lookup format: learning paradigms, loss functions, optimization algorithms, regularization techniques, and evaluation metrics. For full explanations, worked examples, and code walkthroughs, Chapter 0 remains the authoritative source.
ML essentials matter for LLM work because the full model lifecycle, from pretraining to fine-tuning to evaluation, rests on these fundamentals. Knowing when cross-entropy loss is appropriate, how gradient clipping prevents instability, or how to read a confusion matrix correctly saves debugging time when results do not behave as expected.
This reference is most valuable for practitioners who understand the material but benefit from a single page of formulas and tables rather than prose. Graduate students entering LLM work from adjacent fields (vision, speech, classical NLP) will find it a useful orientation to the conventions used throughout the book.
This appendix maps directly onto Chapter 0 (ML and PyTorch Foundations). The evaluation metrics in Section B.4 also connect to the model assessment framework in Chapter 29 (Evaluation). Loss functions and optimization methods reappear throughout fine-tuning in Chapter 14.
This appendix assumes you have read or are reading Chapter 0 (ML and PyTorch Foundations). It is not a self-contained introduction; the entries here are summaries rather than explanations. For mathematical grounding behind the formulas, see Appendix A (Mathematical Foundations).
Reach for this appendix when you need a quick reminder during active coding or reading: "What is the formula for F1 score?" or "Which optimizer should I use for fine-tuning?" rather than re-reading Chapter 0. Each section links back to the relevant Chapter 0 section for deeper coverage. If you are new to ML entirely, read Chapter 0 first. For hardware and cost considerations related to training, see Appendix G (Hardware and Compute).