fix: preserve worker failures and enforce scientific acceptance - #17
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A worker can commit valid numerical results and then time out or fail during teardown. The supervisors can currently return that run as accepted, and artifact refresh can replace the committed scientific state with an orchestration failure.
Record post-commit timeout, cancellation, teardown failure, and unexpected exit codes consistently in Python/CLI and MCP execution. Preserve the numerical outcome in
result.json; inspection reports the interruption throughrun.json. Refresh artifact hashes without collapsing those two states, and preserve completed items when a batch worker fails.Also validate requested properties when importing completed evaluation results. Relaxation acceptance requires convergence plus its required passed check; reject contradictory convergence claims and forces exceeding the requested threshold. Existing valid unconverged results remain readable and unaccepted.
Validation: focused API, worker, CLI, and MCP regression tests; source lint and formatting. Tests launch real worker processes backed by synthetic calculations, including failure after persistence. The combined three-fix wheel suite passed 393 tests with 1 skipped outside the checkout. No scientific engine or GPU validation is claimed.