1
1
Fork
You've already forked micrograd
0
A tiny scalar-valued autograd engine and a neural net library on top of it
  • Common Lisp 67.6%
  • HTML 32.4%
2026年02月10日 17:35:30 +01:00
hyperdoc Adapt to new Wikipedia naming scheme 2026年02月10日 17:35:30 +01:00
AUTHORS Add AUTHORS file to satisfy HAL requirements 2025年12月16日 09:47:09 +01:00
codemeta.json Added codemeta.json 2025年10月31日 18:02:27 +01:00
explore.lisp Adapt to renames in HyperDoc 2025年02月27日 16:38:56 +01:00
LICENSE Updated skeleton, almost complete engine code 2025年07月01日 15:56:59 +02:00
micrograd.asd Adapt to changes in HyperDoc 2025年12月11日 20:31:40 +01:00
README.md Updated README 2025年10月31日 18:02:21 +01:00

Micrograd

Micrograd is a pedagogical implementation of neural networks using a reverse-mode automatic differentiation engine. This is the algorithm used in Deep Learning methods. Micrograd is limited to scalar values, and optimized for clarity rather than for performance.

Micrograd started out as a port of Andrej Karpathy's micrograd package from Python to Common Lisp. The overall architecture and functionality are still the same, but the code has evolved substantially in the process of writing the accompanying explanations and exploration tools. Lecture 1 from Andrej Karpathy's neural networks course remains a good complement to the explanations in this HyperDoc. Be aware though that many names of classes and functions have changed compared to his Python code.

License

GPL 3 or higher

Copyright (c) 2025 Konrad Hinsen