Using Condenser models with pyserini #1351
-
Hi, I'm trying to use the pyserini framework with the pre-trained condenser models available here https://github.com/luyug/Condenser . I can encode the corpus yet I get errors when I am trying to create a FaissSearcher because the model cannot be found and downloaded from HF. This happens in spite of successfully running the encoding. Encoding runs ok. It downloads the model, and encodes the output in spite of warning for not having all the layers initialized. python -m pyserini.encode \
input --corpus data/raw-collection/ \
--fields title text \
--delimiter "\n" \
output --embeddings embeddings/condenser/ \
--to-faiss \
encoder --encoder Luyu/condenser \
--fields title text \
--batch 32 \
--max-length 512 \
--device cpu When creating the Searcher the 'Luyu/condenser ' cannot be found in spite of being used during the encoding phase.from pyserini.search import FaissSearcher
searcher = FaissSearcher(
'index/condenser',
'Luyu/condenser'
)
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
Hi @vicpara , thanks for reporting! the issue should be fixed now with #1355. |
Beta Was this translation helpful? Give feedback.
Hi @vicpara , thanks for reporting! the issue should be fixed now with #1355.