spotlight

Why we included this project

Spotlight is a PyTorch library that tackles recommendation from two directions in one API: classic factorization models for rating prediction and ranking, plus sequential models for next-item prediction. The sequential side covers pooling, LSTM, and causal convolution, and everything shares the same data structures and evaluation helpers, so you can switch between approaches without rebuilding your pipeline. It also bundles MovieLens loaders and synthetic sequence generators, which makes it easy to test an idea quickly. The modular design is simple to extend if you want to experiment with loss functions or representations. One caveat: the project hasn't been updated since late 2022, so it's best treated as a prototyping base rather than something to run in production.