Repository navigation
fix: bump lm_eval to 0.4.11 for transformers 4.56+ compatibility - #225
Merged
mwiebe merged 1 commit intoMay 22, 2026
Merged
Conversation
github-actions
Bot
added
the
waiting-on-maintainers
folouiseAWS
approved these changes
May 22, 2026
mwiebe
approved these changes
May 22, 2026
Signed-off-by: Anika Suman <anika-suman-amazon@users.noreply.github.com>
mwiebe
force-pushed
the
fix/lm-eval-version-bump
branch
from
May 22, 2026 23:24
299932a to
330a952
Compare
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What was the problem/requirement? (What/Why)
The conda-forge
vllm=0.19.1package was rebuilt with a dependency ontransformers>=4.56.0. In transformers 4.56+,AutoModelForVision2Seqwas removed/renamed.lm_eval=0.4.9still references this attribute at import time, causing an immediate crash before any evaluation runs.What was the solution? (How)
Bumped
lm_evalfrom0.4.9to0.4.11in theCondaPackagesparameter default intemplate.yaml. Version 0.4.11 is compatible withtransformers>=4.56.0.What is the impact of this change?
Without this fix, the job bundle fails during benchmark evaluation due to the import error. With this fix, the bundle works as intended again.
How was this change tested?
Ran the job bundle -- all 3 default models (Qwen/Qwen2.5-0.5B, Qwen/Qwen2.5-1.5B, EleutherAI/pythia-1.4b) evaluated successfully across all 4 default benchmarks (hellaswag, arc_easy, arc_challenge, winogrande). The Aggregate step produced the expected leaderboard.csv and leaderboard.md with results matching the README examples.
Was this change documented?
No documentation changes needed, the README does not reference the specific lm_eval version.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.