Inspired by minitorch and tinygrad, smalltensor
is a toy implementation of an basic deep learning framework, with Tensor and Autograd engine.
As a learning project, its aims to be simple, readable and build from the ground up with few dependencies.
Require Python version 3.8 or above.
git clone git@github.com:hbhungg/smalltensor.git
cd smalltensor
python3 -m pip install -e .
Example usage and training can be found at \examples.
- Package as PyPi.
- Create our own backend (no longer depend on numpy)! Good exercise.
- Conv2d.
Welcome PRs and Issues.