Skip to content

fix: accept tuple batches in rerank_pairs #710

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xblwh wants to merge 1 commit into
base: main
Choose a base branch
from
+34 −3
Open

fix: accept tuple batches in rerank_pairs #710

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 8 additions & 3 deletions fastembed/rerank/cross_encoder/onnx_text_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,11 +103,16 @@ def _rerank_pairs(
) -> Iterable[float]:
is_small = False

if isinstance(pairs, tuple):
pairs = [pairs]
if (
isinstance(pairs, tuple)
and len(pairs) == 2
and isinstance(pairs[0], str)
and isinstance(pairs[1], str)
):
pairs = [(pairs[0], pairs[1])]
is_small = True

if isinstance(pairs, list):
if isinstance(pairs, (list, tuple)):
if len(pairs) < batch_size:
is_small = True

Expand Down
26 changes: 26 additions & 0 deletions tests/test_text_cross_encoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,6 +93,32 @@ def test_batch_rerank(model_cache, model_name: str) -> None:
), f"Model: {model_name}, Scores: {scores}, Expected: {canonical_scores}"


@pytest.mark.parametrize("parallel", [None, 2])
@pytest.mark.parametrize("pair_count", [0, 1, 2, 3])
def test_rerank_tuple_of_pairs(model_cache, pair_count: int, parallel: int | None) -> None:
model_name = "Xenova/ms-marco-MiniLM-L-6-v2"
query = "What is the capital of France?"
documents = ["Paris is the capital of France.", "Berlin is the capital of Germany."] * 2
pairs = tuple((query, document) for document in documents[:pair_count])

with model_cache(model_name) as model:
scores = np.array(list(model.rerank_pairs(pairs, batch_size=2, parallel=parallel)))

expected = np.tile(CANONICAL_SCORE_VALUES[model_name], 2)[:pair_count]
np.testing.assert_allclose(scores, expected, atol=1e-3)


@pytest.mark.parametrize("parallel", [None, 2])
def test_rerank_single_pair(model_cache, parallel: int | None) -> None:
model_name = "Xenova/ms-marco-MiniLM-L-6-v2"
pair = ("What is the capital of France?", "Paris is the capital of France.")

with model_cache(model_name) as model:
scores = np.array(list(model.rerank_pairs(pair, batch_size=1, parallel=parallel)))

np.testing.assert_allclose(scores, CANONICAL_SCORE_VALUES[model_name][:1], atol=1e-3)


@pytest.mark.parametrize("model_name", ["Xenova/ms-marco-MiniLM-L-6-v2"])
def test_lazy_load(model_name: str) -> None:
is_ci = os.getenv("CI")
Expand Down