Skip to content
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

Add MMMU scenario and support multimodal multiple choice adaptation #2259

Merged
merged 25 commits into from
Jan 24, 2024
Merged

Conversation

teetone
Copy link
Member

@teetone teetone commented Jan 18, 2024

Resolves #2114, #2068

@teetone teetone added the VHELM Holistic Evaluation of Vision-Language Models (VLM) label Jan 18, 2024
options: List[str] = row["options"]
answer: str = row["answer"]

# Create the question. Questions can have text and images interleaved
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh, I am pretty sure in the PR for Llava and OpenFlamingo we assumed the image was at the top

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think we can evaluate llava and openflamingo with MMMU. There can be multiple images (up to 7) in the question and even in the answer choices.

src/helm/benchmark/vlm_run_specs.py Show resolved Hide resolved
src/helm/benchmark/vlm_run_specs.py Show resolved Hide resolved
@teetone
Copy link
Member Author

teetone commented Jan 20, 2024

@JosselinSomervilleRoberts I answered your concerns. Let me know if something is unclear!

Copy link
Contributor

@JosselinSomervilleRoberts JosselinSomervilleRoberts left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the changes, just make sure to make the change when loading the huggingface dataset, otherwise looks good!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
VHELM Holistic Evaluation of Vision-Language Models (VLM)
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Add MMMU for VLM evaluation
2 participants