Powerful unsupervised domain adaptation method for dense retrieval. Requires only unlabeled corpus and yields massive improvement: "GPL: Generative Pseudo Labeling for Unsupervised Domain Adaptation of Dense Retrieval" https://arxiv.org/abs/2112.07577
The previous code does not give enough hint about missing evaluation data
gpl/toolkit/evaluation.py
: Added checking for missing evaluation datatests/unit/conftest.py
: Separated sbert and sbert_path fixturestests/unit/test_eval.py
: Added test