Currently, we must use scikit-learn==1.3.0 because we load model weights from a file that was generated on that version. Subsequently, we must also pin numpy<2 due to the scikit-learn pin: https://github.com/scikit-learn/scikit-learn/blob/7f9bad99d6e0a3e8ddf92a7e5561245224dab102/pyproject.toml.
We should work to make our model files less version-dependent so we can remove these dependency pins entirely. Perhaps we could switch to ONNX persistence: https://scikit-learn.org/stable/model_persistence.html#onnx-persistence
Currently, we must use
scikit-learn==1.3.0because we load model weights from a file that was generated on that version. Subsequently, we must also pinnumpy<2due to thescikit-learnpin: https://github.com/scikit-learn/scikit-learn/blob/7f9bad99d6e0a3e8ddf92a7e5561245224dab102/pyproject.toml.We should work to make our model files less version-dependent so we can remove these dependency pins entirely. Perhaps we could switch to ONNX persistence: https://scikit-learn.org/stable/model_persistence.html#onnx-persistence