INNER CODE UNIT · Python
queries_per_passage
UKPLab/gpl · gpl/train.py:99
if queries_per_passage == -1:
assert "corpus.jsonl" in os.listdir(
path_to_generated_data
), "At least corpus should exist!"
corpus = GenericDataLoader(path_to_generated_data).load_corpus()
if len(corpus) * 3 < 250e3:
queries_per_passage = math.ceil(
250e3 / len(corpus)
) # Here use ceil to guarantee the QPP will not be too small
else:
queries_per_passage = 3
logger.info(f"Automatically set `queries_per_passage` to {queries_per_passage}")
#### Synthetic query generation ####
#### This will be skipped if there is an existing `gen-queries.jsonl`file under `path_to_generated_data` ####
if use_train_qrels == True:
if qgen_prefix is not None:
logger.warning(