Project installation fails: tensorflow conflicting dependencies

[djbb7]

Describe the bug

The installation of the project dependencies fails, when following the steps outlined in the readme. The error message indicates that it was not possible to find a version of tensorflow that satisfies requirements.

To Reproduce

1. git clone [email protected]:openai/evals.git
2. cd evals
3. git lfs fetch --all
4. git lfs pull
5. pip install -e
6. See error message:

ERROR: Cannot install evals because these package versions have conflicting dependencies.

The conflict is caused by:
    spacy-universal-sentence-encoder 0.4.6 depends on tensorflow<3.0.0 and >=2.4.0
    spacy-universal-sentence-encoder 0.4.5 depends on tensorflow<3.0.0 and >=2.4.0
    spacy-universal-sentence-encoder 0.4.4 depends on tensorflow<3.0.0 and >=2.4.0
    spacy-universal-sentence-encoder 0.4.3 depends on tensorflow<3.0.0 and >=2.4.0
    spacy-universal-sentence-encoder 0.4.1 depends on tensorflow<3.0.0 and >=2.4.0
    spacy-universal-sentence-encoder 0.4.0 depends on tensorflow<3.0.0 and >=2.4.0
    spacy-universal-sentence-encoder 0.3.4 depends on tensorflow==2.3.0
    spacy-universal-sentence-encoder 0.3.3 depends on tensorflow==2.3.0
    spacy-universal-sentence-encoder 0.3.2 depends on tensorflow==2.3.0
    spacy-universal-sentence-encoder 0.3.1 depends on tensorflow==2.2.0
    spacy-universal-sentence-encoder 0.3.0 depends on tensorflow==2.2.0
    spacy-universal-sentence-encoder 0.2.1 depends on tensorflow==2.1.0

To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip to attempt to solve the dependency conflict

Code snippets

No response

OS

macOS Sonoma 14.5

Python version

3.13.0

Library version

3.0.1.post1