Repository navigation
Conversation
…ields The branch carried an older copy of task-groups.yaml, which reverted OpenEuroLLM#114, OpenEuroLLM#122, OpenEuroLLM#123, OpenEuroLLM#124 and the FLORES/Global-MMLU/INCLUDE changes on main and failed 8 tests. Keep main's file and add only the four benczechmark-* groups and the oellm-benczechmark super group. Every task now declares its HF dataset (and the umimeto subject subset), so `oellm-eval schedule` pre-downloads them for air-gapped compute nodes and tests/test_datasets.py covers them.
transformers 5 removed transformers.data.metrics, so importing utils.py failed and every BenCzechMark task failed to load with the [eval] extra or the UFAL requirements (transformers 5.5.4). Vendor the SQuAD exact-match/F1 helpers verbatim instead, so cs_sqad32 scores are unchanged. Also fix the ruff errors that failed the CI lint job and a SyntaxWarning from an un-escaped backslash in a docstring.
The ported tasks reference only _n_shot.yaml, qa/_qa_template.yaml and metrics/_qa_metrics.yaml. The other templates (multichoice/*, the classification metrics, summarization, CNEC NER) came along from HPLT-e for tasks that were not ported; _cnec_ner_template.yaml even points at a utils.process_docs_cs_nec that does not exist. Dropping them also removes the MC-AUROC/ROUGE helpers and with them the runtime dependency on evaluate.load(), which needs network access.
Collaborator
|
Thanks for porting this. I tried running it and hit a few things that block it on current main, so I pushed three commits on top:
With these, all 80 tasks load with Qwen3-0.6B-Base. @varisd the groups are 0-shot while the ported |
Collaborator
|
here is a results with Qwen3-0.6B-Base, there might be a bug with
|
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
2 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.
This PR moves the BenCzechMark task configs from the HPLT-e project.
The directory structure and filenames has been adjusted to reflect the naming in
custom_lm_eval_tasks/.Furthermore, it adds the
benczechmark-*task groups andoellm-benczechmarksuper_group.