castorini / pyserini

Pyserini is a Python toolkit for reproducible information retrieval research with sparse and dense representations.
http://pyserini.io/
Apache License 2.0
1.63k stars 356 forks source link

On-the-fly vs. pre-encoded queries for all dense retrieval models #445

Closed lintool closed 3 years ago

lintool commented 3 years ago

For some dense retrieval models, we use on-the-fly, for others, we use pre-encoded... and we have some models where we have both.

We should go back and make consistent - i.e., we should have all three options available - and tested.

MXueguang commented 3 years ago

this issue will be solved together with https://github.com/castorini/pyserini/issues/459