Skip to content

TRANSLATIONS.md

Latest commit

 

History

History
56 lines (40 loc) · 2.64 KB

TRANSLATIONS.md

File metadata and controls

56 lines (40 loc) · 2.64 KB

Translations

no-magic scripts teach algorithms through code comments. Translating those comments into other languages makes this resource accessible to engineers worldwide who think more fluently in their native language.

This initiative translates comments, docstrings, section headers, and print statements in all 41 scripts. The code itself stays in English — variable names, function names, and logic are universal.

Status

Language Locale Status Maintainer Progress
Spanish es 🔴 Not Started TBD 0/41 scripts
Brazilian Portuguese pt-BR 🔴 Not Started TBD 0/41 scripts
Simplified Chinese zh-CN 🔴 Not Started TBD 0/41 scripts
Japanese ja 🔴 Not Started TBD 0/41 scripts
Korean ko 🔴 Not Started TBD 0/41 scripts
Hindi hi 🔴 Not Started TBD 0/41 scripts

Guidelines for Translators

What to Translate

  • File thesis docstrings
  • Section headers (# === SECTION NAME ===)
  • All inline comments (why comments, math-to-code mappings, intuition comments, signpost comments)
  • Print statements (training progress, inference output)
  • Docstrings on functions

What NOT to Translate

  • Variable names, function names, class names
  • Code logic, operators, syntax
  • Mathematical notation in comments (equations are universal)
  • English technical terms that are standard in the target language's ML community (e.g., "softmax", "embedding", "attention")

File Naming and Location

Translated scripts live in translations/<locale>/. The filename stays the same as the original:

translations/es/microgpt.py      # Spanish translation of 01-foundations/microgpt.py
translations/ja/microlora.py     # Japanese translation of 02-alignment/microlora.py
translations/zh-CN/microbeam.py  # Chinese translation of 03-systems/microbeam.py

Quality Bar

  1. Technical accuracy over literary polish. An incorrect translation is worse than an awkward one.
  2. Preserve all 7 comment types from the commenting standard (thesis, section headers, why comments, math-to-code mappings, intuition comments, signpost comments, no obvious comments).
  3. Preserve math notation as-is. Equations are universal.
  4. Use domain-standard terminology for the target language's ML community.
  5. When in doubt, keep the English term with a parenthetical translation: # softmax (normalización exponencial).
  6. The translated script must still run. python translations/es/microgpt.py must produce the same training and inference results as the original.