This is the source code for the Shallow Dives course on building neural networks from scratch in Rust, inspired by Karpathy's Zero To Hero course, and enabled by Candle, a minimalist ML framework for Rust developed by HuggingFace.
This course starts all the way at the basics of and progresses through developing and training Rust versions of micrograd, makemore, and nanoGPT as described in original Python repo here: https://github.com/karpathy/nn-zero-to-hero
Why learn Rusty ML? Full machine learning frameworks like PyTorch are very large, whereas Candle allows deployment of lightweight binaries and lets you remove Python from production workloads for drastically improved performance.
You can follow along with this course by progressing through the interactive notebooks in the /notebooks directory -
either within the Shallow Dives course interface or locally by cloning this repository and following
the instructions in notebooks/GettingStarted.md.
Karpathy's Zero to Hero Course
User Guide for the Candle Framework
Original Source Code for Micrograd
Original Source Code for Makemore