Repository navigation
fix: reject resumes across incompatible dataset and model identities - #5
Open
rudycelekli wants to merge 1 commit into
Open
rudycelekli wants to merge 1 commit into
rudycelekli wants to merge 1 commit into
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Resuming a label currently inherits matching qids even after the dataset content or driver model changes. The runner then stamps the current config onto inherited rows, concealing mixed measurements. Records already carry
gt_shaandmodel, but resume never checks them.Require retained rows to match the current dataset hash and model. Missing legacy identity fails with an actionable fresh-run hint; existing records remain unchanged. Dead rows being retried still work normally. Propagate the reserved configuration-error status through orchestration and the public CLI so rejected saved cells cannot subsequently be rescored or presented under the new invocation.
Validation: 9 new regressions fail against upstream production code; all 144 tests pass after the fix on Python 3.11. Tests cover unknown/mismatched identity, compatible resume, dead retries,
RESUME_KEEP_DEAD, original file preservation, production identity wiring, suite rejection, and public CLI propagation. No provider calls.git diff --checkpasses.Compatibility: legacy retained rows without dataset/model stamps need a fresh run. This validates dataset/model identity; other harness and remote-server changes remain outside this fix.