Conflicting packages optimum-haystack and instructor-embedders-haystack

[ArzelaAscoIi]

Describe the bug

• for optimum-haystack the we request sentence-transformers>=2.3
• for nstructor-embedders-haystack we exclude 2.3 or greater

These packages are incompatible

To Reproduce

• just install both packages

Describe your environment (please complete the following information):

• OS: [e.g. iOS]
• Haystack version:
• Integration version:

[julian-risch]

@ArzelaAscoIi After a first look at the dependencies my guess is that we can remove the <2.3.0 pin from the instructor-embedder integration because there is also no upper limit for the version set here https://github.com/xlang-ai/instructor-embedding/blob/main/requirements.txt#L9

[anakin87]

I'm not sure... I pinned the dependency because Instructor Embeddings were not working with sentence_transformers>=2.3.0. #289

Then they updated the main branch (xlang-ai/instructor-embedding#112), but a new package was not released. https://pypi.org/project/InstructorEmbedding/#history; xlang-ai/instructor-embedding#112 (comment).

So, we should try...

[anakin87]

I tried to unpin the dependency and using sentence-transformers==3.0.1 (latest),
I get these errors when running the tests:

FAILED tests/test_instructor_document_embedder.py::TestInstructorDocumentEmbedder::test_run - TypeError: INSTRUCTOR._load_sbert_model() got an unexpected keyword argument 'token'
FAILED tests/test_instructor_text_embedder.py::TestInstructorTextEmbedder::test_run - TypeError: INSTRUCTOR._load_sbert_model() got an unexpected keyword argument 'token'

[julian-risch]

Thanks @anakin87 ! That means at the moment the two packages are not compatible and there is nothing we can do about it, which is why I suggest we close this issue.