Skip to content

fix: normalize custom token embeddings along the embedding axis #793

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

Closed
ktz03 wants to merge 1 commit into from
Closed

fix: normalize custom token embeddings along the embedding axis #793

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
2 changes: 1 addition & 1 deletion fastembed/text/custom_text_embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,7 +119,7 @@ def _pool(
)

def _normalize(self, embeddings: NumpyArray) -> NumpyArray:
return normalize(embeddings) if self._normalization else embeddings
return normalize(embeddings, dim=-1) if self._normalization else embeddings

@classmethod
def add_model(
Expand Down
36 changes: 36 additions & 0 deletions tests/test_custom_models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -284,6 +284,42 @@ def test_custom_text_model_lookup_is_case_insensitive():
assert model.model._normalization is True


@pytest.mark.parametrize("dtype", [np.float16, np.float32, np.float64])
@pytest.mark.parametrize("token_output", [False, True])
@pytest.mark.parametrize("normalization", [False, True])
def test_custom_disabled_pooling_normalizes_each_vector(
dtype, token_output: bool, normalization: bool
) -> None:
"""Unpooled token outputs normalize along the embedding axis, like sentence outputs."""
model_name = "local/disabled-pooling"
TextEmbedding.add_custom_model(
model_name,
pooling=PoolingType.DISABLED,
normalization=normalization,
sources=ModelSource(hf="local/unused"),
dim=2,
)
model = CustomTextEmbedding(model_name, lazy_load=True, specific_model_path="./")
vectors = np.array([[[3, 4], [-5, 12], [0, 0]], [[3, 4], [3, 4], [0, 0]]], dtype=dtype)
expected = np.array(
[[[0.6, 0.8], [-5 / 13, 12 / 13], [0, 0]], [[0.6, 0.8], [0.6, 0.8], [0, 0]]],
dtype=dtype,
)
if not token_output:
vectors, expected = vectors[:, 0], expected[:, 0]
if not normalization:
expected = vectors.copy()
original = vectors.copy()

output = np.stack(
list(model._post_process_onnx_output(OnnxOutputContext(model_output=vectors)))
)

np.testing.assert_allclose(output, expected, rtol=1e-3 if dtype == np.float16 else 1e-6)
np.testing.assert_array_equal(vectors, original)
assert output.dtype == dtype


def test_do_not_add_existing_model():
existing_base_model = "sentence-transformers/all-MiniLM-L6-v2"
custom_model_name = "intfloat/multilingual-e5-small"
Expand Down