A tiny deep learning library from scratch in Python
What I cannot create, I do not understand.
This codebase follows the one hour coding madness of Joel Grus, building a deep learning framework in under an hour!
Also, Joel has some strong opinions on software engineering and replicable research. I wanted to follow a pro typing docstrings, use type annotations, and run mypy for type checking while building something from scratch. I'm hooked now!
See XOR or the awesome fizzbuzz example
- Tensors
- Loss Functions
- Layers
- Neural Nets
- Optimizers
- Data
- Training
- End-to-end example: XOR
- End-to-end example: FizzBuzz