RepBERT

[wangjiajia5889758]

Is the generating of passage_embeddings of this program the same as the RepBERT program?:

python precompute.py --load_model_path ./data/ckpt-350000 --task doc
python precompute.py --load_model_path ./data/ckpt-350000 --task query_dev.small
python precompute.py --load_model_path ./data/ckpt-350000 --task query_eval.small

[jingtaozhan]

This repo uses the output embedding of [cls] as the query/passage embedding.
RepBERT uses the average token embedding as the query/passage embedding.
The encoder implementation is different.