Skip to main content

CI compatibility

This matrix shows pipeline CI compatibility on demo datasets for model, dataset, and task combinations collected by MEDS-DEV's full-integration suite. It does not measure predictive performance or establish clinical validity. See the benchmark for predictive-performance results such as AUROC.

Statuses describe GitHub Actions job conclusions, not individual pytest case outcomes. A passed job can contain skipped pytest cases. The pytest step conclusion is shown separately: for example, tests can succeed while a later artifact upload causes the job to fail. Follow a status link to inspect the job logs; the run also provides the CI matrix report and available JUnit artifacts.

Loading CI compatibility report…

Reading the matrix​

  • Passed: the job completed successfully; this does not imply every pytest case ran and passed.
  • Failed: the job failed or timed out, potentially outside the pytest step.
  • Skipped: GitHub Actions skipped the job, distinct from individual skipped pytest cases.
  • Cancelled: the job was cancelled.
  • Not run: the combination was expected from pytest collection, but no matching job was found.
  • Other: the job has another or unavailable conclusion; inspect its logs for details.
  • — (not included): the combination was not in the collected matrix. It may be unsupported or untested; it is not a failure and is excluded from the counts.

Use the dataset, task, and model selectors to narrow the matrix. Counts reflect only collected combinations matching all selected filters, rather than the full cross-product of rows and columns.

Snapshot availability​

The report is the latest published full-integration snapshot from MEDS-DEV's main branch, not a live view of every pull request or the latest commit. Use its generated timestamp, commit, and run link to assess freshness. Pull request and dev runs do not update this public snapshot.

Publication is introduced by MEDS-DEV PR #345. The feed will become available after that change reaches main and a qualifying full run publishes it. If the report is absent, invalid, or unavailable, use MEDS-DEV GitHub Actions to inspect runs directly.