Summary
scikit-learn example scripts exercise realistic end-to-end workflows (fit, predict, plot) that unit tests don't cover. Running them under cuml.accel catches regressions that affect user experience.
Motivation
A 100% pass rate on scikit-learn examples is an achievable and meaningful target — more so than the full test suite, where many failures are internal test-infrastructure mismatches rather than real user-facing issues. Example scripts mirror what users actually do, so failures here are failures users would hit. Passing all of them is a strong signal that cuml.accel has reached maturity for real-world use cases, and something we can document to set expectations.
Summary
scikit-learn example scripts exercise realistic end-to-end workflows (fit, predict, plot) that unit tests don't cover. Running them under
cuml.accelcatches regressions that affect user experience.Motivation
A 100% pass rate on scikit-learn examples is an achievable and meaningful target — more so than the full test suite, where many failures are internal test-infrastructure mismatches rather than real user-facing issues. Example scripts mirror what users actually do, so failures here are failures users would hit. Passing all of them is a strong signal that cuml.accel has reached maturity for real-world use cases, and something we can document to set expectations.