Fix evaluate.sh passing arguments evaluate.py does not accept - #81
Open
jonathan-politzki wants to merge 1 commit into
Open
Fix evaluate.sh passing arguments evaluate.py does not accept#81jonathan-politzki wants to merge 1 commit into
jonathan-politzki wants to merge 1 commit into
Conversation
evaluate.sh invokes evaluate.py with --temperature and --guidance_scale, but evaluate.py's main() defines neither. fire aborts with "ERROR: Could not consume arg: --temperature" before any evaluation runs, so the one-click evaluation script fails out of the box. Both flags were no-ops for the constrained beam-search path (do_sample is not enabled), so simply dropping them restores the intended behavior. Verified on a fresh checkout: with these flags removed, evaluate.py runs to completion and calc.py reports HR/NDCG. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This was referenced Aug 31, 2026
Open
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
evaluate.sh(the one-click offline evaluation script) passes two flags thatevaluate.py'smain()does not define:Because
evaluate.pyusesfire, unknown arguments are fatal. Runningbash evaluate.shon a clean checkout aborts immediately with:No evaluation runs and no metrics are produced.
Fix
Drop the two unsupported flags. Both were no-ops for this code path — generation uses constrained beam search (
do_sampleis not enabled), wheretemperaturehas no effect, andguidance_scalewas never threaded intoGenerationConfig.Verification
Hit this while running a full SFT → evaluate → calc reproduction on a single A100. With the flags removed,
evaluate.pyruns to completion andcalc.pyreports HR/NDCG as expected.This may be contributing to the reproduction difficulties reported in #35, #43 and #49, since it blocks the documented evaluation entry point.
🤖 Generated with Claude Code