KeyNotes is a bilingual learning guide to machine learning and the infrastructure behind it. It connects first principles, worked examples, diagrams, and engineering trade-offs so newcomers can build intuition and experienced readers can trace how systems work.
- ML: foundations, language and multimodal models, training, inference, and data tools.
- Infra: accelerators, memory, communication, data movement, and performance.
Read the English documentation or switch to 简体中文 with the language icon on any page. The two editions share the same topic structure.
Corrections, clearer examples, diagrams, and new topics are welcome. Open an issue to discuss a larger addition, or submit a pull request with a focused change.
- Keep the English and Chinese versions of a topic at matching paths under
docs/en/anddocs/zh/. - Explain the motivation, mechanism, assumptions, and a concrete example where useful. Cite primary sources for implementation-specific claims.
- Update
mkdocs.ymlwhen adding a topic. Keep shared images indocs/assets/and preserve the original imported figures. - In the pull request, describe what changed and how you checked the affected pages. Read the Docs provides a preview for pull requests.
The original Notion material and image checksums are recorded in migration/.
Original KeyNotes documentation text and diagrams are shared under CC BY 4.0 where the contributors hold the necessary rights. You may share and adapt them, including commercially, with attribution, a license link, and an indication of changes. Imported images and third-party material are excluded unless their own notices say otherwise; see LICENSE.md for scope.
Suggested credit: nbbb24, “KeyNotes,” project website, CC BY 4.0, plus the page URL and any changes you made. GitHub's “Cite this repository” menu also reads CITATION.cff.