-
Hi there, I follow this repo (https://github.com/luyug/COIL) to train a UniCOIL model with my own dataset (in Vietnamese), now I want to create a search API base on this UniCOIL model, follow this code (https://github.com/castorini/pyserini/blob/master/pyserini/search/lucene/__main__.py) and this script (https://github.com/castorini/pyserini/blob/master/docs/experiments-unicoil.md) I have to use LuceneImpactSearcher instead LuceneSearcher, but LuceneImpactSearcher need query_encoder argument which use to encode query for searching, but I use another encoder to encode query first then do searcher.search(text) with text was encoded by the encoder, is there any way to ignore query_encoder argument when using LuceneImpactSearcher |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
Yes, see the |
Beta Was this translation helpful? Give feedback.
Yes, see the
--encoded-queries
options here as an example: https://github.com/castorini/pyserini/blob/master/docs/experiments-tct_colbert-v2.md