[BUG] Numpy Version inconsistency in 03-Real-Time Fraud Detection sample notebook #734
Description
Community Note
- Please use a 👍 reaction to provide a +1/vote. This helps the community and maintainers prioritize this request.
- If you are interested in working on this issue or have submitted a pull request, please leave a comment.
Describe the bug
Sample Notebook Neptune/03-Neptune-ML/03-Sample-Applications/03-Real-Time Fraud Detection Using Inductive Inference fails at !pip install -q scikit-learn umap-learn matplotlib with the error
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
graph-notebook 4.6.2 requires numpy<1.24.0, but you have numpy 2.2.4 which is incompatible.
sagemaker 2.243.0 requires numpy<2.0,>=1.9.0, but you have numpy 2.2.4 which is incompatible.
To Reproduce
Run pip install -q scikit-learn umap-learn matplotlib after installing graph-notebook
Root Cause Analysis
The issue is caused by a recent update to the numba package, which is a dependency of umap-learn. In their latest release, numba updated their minimum supported numpy version to 1.24.0.
This creates a conflict because:
- graph-notebook v4.6.2 requires
numpy <1.24.0 - sagemaker v2.243.0 requires
numpy >=1.9.0,<2.0
What's happening?
Initially, numpy v1.23.5 is installed, which meets these requirements.
However, when running the command pip install -q scikit-learn umap-learn matplotlib, numpy gets automatically upgraded to v2.2.4 due to the new numba requirements.
Expected behavior
The required packages get installed without any errors and the sample notebook runs successfully.