We recently prepared a template to bootstrap students projects, in a DL course.
To start a new project it is enough to click on Template and everything non-project specific will be already present in the new repository.
The goal is to help in all the scaffolding needed for flexible configuration, logging, and hyperparameter sweeps while encouraging best practices.
It uses Lightning, Hydra, W&B, DVC and Streamlit:
It would be great to have feedback and opinions from you, in particular I’m interested in your working flow with streamlit in ML research