FAQ and Troubleshooting
Why do I get missing artifacts instead of execution?
Stages check expected files and may skip or report planned mode. Confirm dataset layout and artifact names in io.dataset.
How do I skip already completed stages?
When expected artifacts exist, pipeline reports skipped_existing. Use a fresh run copy that still needs the required outputs when you want to execute those stages again.
Where are verification failures documented?
Review Verification for failure causes, tolerance tuning guidance, and the correct verify command flow.
GPU backend requested but unavailable
Install CuPy for your CUDA target or set backend to non-GPU mode; kernel wrapper raises BackendUnavailableError.
Known limitations
- Parity claims require fresh verification or audit evidence; interrupted or stale run copies are intentionally rejected by the validation workflow.
- MAT v7.3 reading requires fallback libraries depending on file content.