search

dsgibbons / shap

A game theoretic approach to explain the output of any machine learning model.
https://shap-community.readthedocs.io/en/latest/
MIT License
25 stars 5 forks source link

Random floating-point errors in GitHub Actions #98

Closed dsgibbons closed 1 year ago

dsgibbons commented 1 year ago

We have seen GitHub Actions randomly fail occasionally due to floating-point precision issues. See, for example:

In both cases, the tests passed after being re-run with no source code changes. We should consider either increasing some of the floating-point tolerances, or investigate why we are seeing such large random deviations in test_deep.py.

dsgibbons commented 1 year ago

Closing in favour of https://github.com/slundberg/shap/issues/2960