-
Notifications
You must be signed in to change notification settings - Fork 116
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Adds Global MLMU #426
Adds Global MLMU #426
Conversation
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Hi! It could also make sense to support "Unannotated" (which is |
I don't think that's particulary intersting for anyone. I don't see a reason in suporting unannotated |
Comparative analysis, to compare un-annotated to CS or CA and therefore get a better idea of which of the subsets it's closer to |
It's a very minor change to add no? |
|
Just one nit in the doc, if it works for you merge it and then you can squash and merge |
Co-authored-by: Clémentine Fourrier <22726840+clefourrier@users.noreply.github.com>
Supports 3 version (ca (cult. agnostic), cs (cult sensitive), all (everything)):
Example how to run:
lighteval vllm pretrained="Qwen/Qwen2.5-7B,dtype=bfloat16,max_model_length=4096,pairwise_tokenization=True,tensor_parallel_size=1,gpu_memory_utilisation=0.6" --custom-tasks lighteval.tasks.multilingual.tasks --max-samples 100 --output-dir ./results "lighteval|global_mmlu_all_ces_cf|3|1"
Results for Qwen2.5-7B:
Examples of prompts (ces):
cc @clefourrier, as you had opinion on subset