The GitHub Actions job "Benchmarks" on texera.git/feat/standalone-ml-m3 has failed. Run started by GitHub user kz930 (triggered by kz930).
Head commit for run: abf2c352e0f1b5bdda6b5df543b37e16b366d12c / kary zheng <[email protected]> fix(sklearn): score the rows the executor scores The executor drops every row holding a missing value before it scores, because the model arrives already fitted and the operator cannot ask which estimator it holds. The exported script scored all of them, so it answered with a number the run never reported, or failed inside scikit-learn on a value the estimator refuses. The verification runner cannot catch this one: a fixture written from the JVM cannot carry a fitted model on an input port, so the two paths are never run side by side. The test instead pins the property that follows from the drop, which is testable on its own: a row the executor would drop must not move the score. Its label is one the model never saw, so scoring that row is wrong whatever the estimator answers; a plausible label would have agreed by luck, since a decision tree predicts through a missing feature rather than refusing it. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]> Report URL: https://github.com/apache/texera/actions/runs/33917737629 With regards, GitHub Actions via GitBox
