Repository navigation
Conversation
…view - bench_typesafe: falsy-zero on confidence=0.0 no longer crashes - probe: fix recency tie-break; stop silently dropping classify() errors - gate: clear error instead of KeyError when thresholds/summary keys diverge - model-gate.yml: cache key also covers the vocab file's LFS oid - distill_report: --version now reaches the heldout-set key - loaders: shared confusion_metrics(), used by run_eval and finetune_distill instead of two separately maintained metric implementations - label_audit: shared ask_typesafe_batch(), used by distill_labels instead of a second retry/concurrency/resume harness
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes from the code review on this branch's diff:
bench_typesafe.py: confidence of exactly 0.0 no longer crashes the run (falsy-zero bug in attribute extraction)probe.py: training-needs list now actually sorts most-recent-first within a severity tier; a classify() error no longer disappears silently instead of showing up as a visible error rowgate.py: a threshold key with no matching summary key now fails with a clear message instead of a rawKeyErrormodel-gate.yml: the LFS cache key now also covers the vocab file's oid, so a vocab-only change can't reuse a stale cached vocabdistill_report.py:--versionnow actually reaches the heldout-set lookup key instead of it being hardcoded to 2.9loaders.py/run_eval.py/finetune_distill.py: extracted the sharedconfusion_metrics()helper so accuracy/precision/recall/f1/block-accuracy are defined once, not twice; verified numerically equivalent to both prior implementationslabel_audit.py/distill_labels.py: extracted the sharedask_typesafe_batch()helper so the retry/concurrency/incremental-write TypeSafe query loop isn't maintained in two placesNo behavior change intended outside of the bug fixes themselves.