Description
When --resume is passed and a partial results.json exists in the output directory, skip already-completed scenarios and continue from where the run stopped.
Part of epic #2827. See spec: .local/specs/zeph-bench/spec.md FR-008.
Scope
- On startup with
--resume: read results.json if it exists, extract completed scenario IDs
- Skip scenarios whose IDs appear in the existing results
- Append new results to existing list; update aggregate on completion
- If
results.json does not exist, behave as a fresh run
- Warn if the existing results are from a different dataset or model
Acceptance Criteria
Description
When
--resumeis passed and a partialresults.jsonexists in the output directory, skip already-completed scenarios and continue from where the run stopped.Part of epic #2827. See spec:
.local/specs/zeph-bench/spec.mdFR-008.Scope
--resume: readresults.jsonif it exists, extract completed scenario IDsresults.jsondoes not exist, behave as a fresh runAcceptance Criteria
--resumeon a partial results file skips completed scenariosresults.jsoncontains all scenarios (original + resumed)--resumewith no existing file behaves as fresh run