We have the following main points of running experiments in the matchmaker library:

• train.py Main training runner for both re-ranking & retrieval training of all models supported by matchmaker.

• eval.py Allows to evaluate a re-ranking model.

• dense-retrieval.py Allows to encode, index, and evaluate a trained dense retrieval model

• pre-train.py Currently a bit abandoned, but allows to pre-train model weights on an mlm or other self-supervised pre-training task