third_codes/eland
1
0
Fork 0
mirror of https://github.com/elastic/eland.git synced 2025-07-24 00:00:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

Upgrade Sentence Transformers to v5 (#801)

Browse Source 
Sentence Transformers v5 adds support for sparse embedding models and is now necessary for importing sparse models such as https://huggingface.co/naver/splade-v3-distilbert.
This commit is contained in:
David Kyle 2025-07-23 08:07:29 +01:00 committed by GitHub
parent 117f61b010
commit bebb9d52e5
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 23 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
setup.py
View File

@ -62,10 +62,10 @@ extras = {
"requests<3",
"torch==2.5.1",
"tqdm",
"sentence-transformers>=2.1.0,<=2.7.0",
"sentence-transformers>=5.0.0,<6.0.0",
# sentencepiece is a required dependency for the slow tokenizers
# https://huggingface.co/transformers/v4.4.2/migration.html#sentencepiece-is-removed-from-the-required-dependencies
"transformers[sentencepiece]>=4.47.0",
"transformers[sentencepiece]>=4.47.0,<4.50.3",
],
}
extras["all"] = list({dep for deps in extras.values() for dep in deps})

21
tests/ml/pytorch/test_pytorch_model_upload_pytest.py
View File

@ -65,6 +65,8 @@ TEXT_EMBEDDING_MODELS = [
TEXT_SIMILARITY_MODELS = ["mixedbread-ai/mxbai-rerank-xsmall-v1"]
TEXT_EXPANSION_MODELS = ["naver/splade-v3-distilbert"]
@pytest.fixture(scope="function", autouse=True)
def setup_and_tear_down():
@ -155,3 +157,22 @@ class TestPytorchModel:
assert result.body["inference_results"][0]["predicted_value"] < 0
assert result.body["inference_results"][1]["predicted_value"] > 0
@pytest.mark.skipif(ES_VERSION < (9, 0, 0), reason="requires current major version")
@pytest.mark.parametrize("model_id", TEXT_EXPANSION_MODELS)
def test_text_expansion(self, model_id):
with tempfile.TemporaryDirectory() as tmp_dir:
ptm = download_model_and_start_deployment(
tmp_dir, False, model_id, "text_expansion"
)
result = ptm.infer(
docs=[
{
"text_field": "The Amazon rainforest covers most of the Amazon basin in South America"
},
{"text_field": "Paris is the capital of France"},
]
)
assert len(result.body["inference_results"][0]["predicted_value"]) > 0
assert len(result.body["inference_results"][1]["predicted_value"]) > 0