This project template implements a simple MNIST classification code with a fully connected network. Uses:
- Lightning
- Hydra
- Have uv installed in the system.
- Clone this repository.
- Run
uv init
, thenuv sync
.
- For training:
python -m scripts.train
. - For testing loss on test set:
python -m scripts.test
- For placing the trained weights in a more accessible folder weights/ run
python -m scripts.pull_last_weights
, use the--del
flag to delete the checkpoint history to reduce clutter and save space. - notebooks/run_network.ipynb is a notebook for visualizing network results.