mlcommons / mlcommons/endpoints
Unify accuracy scoring path for perf+accuracy and accuracy-only runs
@Palanivelg is already working on this.
Since Jul 22, 2026.
- Dominant language
- Python
- Stars
- 21
- Forks
- 28
- Avg merge
- 3d 17h
- Merged PRs (30d)
- 13
Description
Follow-up to #346 (review thread on execute.py — accuracy scoring path).
Accuracy-only and perf+accuracy runs have parallel wiring in commands/benchmark/execute.py, which risks drift between the two modes.
Goal: drive scoring entirely off the config's declared phases so both modes share a single path (perf+accuracy should give the same accuracy behavior as accuracy-only).
The immediate single-stream duplication was consolidated in 4ed5bf8 (accuracy-only num_workers=1/max_connections=1 is baked into the config before config.yaml is persisted). This tracks the deeper unification, deferred to keep #346 scoped.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.