[WIP][Benchmark] Add MEGA-Bench core and core_single_image support #724
Conversation
This file contains 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
TianhaoLiang2000
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Hi, VLMEvalKit team,
This PR incorporates our recent work, MEGA-Bench, a multimodal evaluation suite with over 500 real-world tasks and 45 metrics.
The evaluation process involves two steps: 1) run VLMEvalKit to produce the response/submission file; 2) run our evaluator with 45 diverse metrics to get the scores.
Example usage:
python3 run.py
--data MEGABench_core_single_image_16frame
--model Qwen2-VL-7B-Instruct
--verbose
--work-dir ~/LMUData
This PR implements the MEGA-Bench core subset and the core_single_image subset, and the core_single_image subset has been successfully tested using the Qwen2VL-7B model, confirming its functionality and compatibility. The open-ended subset will be committed soon in the same PR.