Open Issues Need Help
View All on GitHubAI Summary: This issue proposes improving the fairness model card generation to prevent crashes when using intersectional analysis with too few subgroups. The current implementation requires at least one subgroup, but this threshold could be made more intelligent, for example, by requiring at least two groups for comparison or basing the minimum on a percentage of attribute intersections. The issue is marked as 'good first issue' but also 'wontfix', suggesting it's a good learning opportunity but not intended for immediate merging.
Contains modules with MAMMOths research results, and a lightweight demonstrator for running those
AI Summary: Debug an ONNX model loading error within the MAI-BIAS toolkit. The error occurs after data encoding, suggesting a problem with the pre-processing or the ONNX model itself. Debugging will likely involve inspecting the data pipeline, the ONNX model structure, and potentially the MAI-BIAS toolkit's ONNX loading mechanisms.
Contains modules with MAMMOths research results, and a lightweight demonstrator for running those