open-compass / open-compass/VLMEvalKit
LLAVA-one-vision MMMU 性能评测差别很大
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 4.4k
- Forks
- 768
- Avg merge
- 1d 10h
- Merged PRs (30d)
- 17
Description
这是评测的脚本：torchrun --nproc-per-node=1 run.py --data MMMU_DEV_VAL --model llava_onevision_qwen2_0.5b_ov --verbose
评测下来的指标是：
split validation dev
Overall 0.3522222222222222 0.31333333333333335
Accounting 0.43333333333333335 0.0
Agriculture 0.36666666666666664 0.2
Architecture_and_Engineering 0.23333333333333334 0.2
Art 0.3333333333333333 0.0
Art_Theory 0.43333333333333335 0.6
使用的 gpt3.5 进行评测，和llava-one-vision 论文里面给的结果0.31，有很大的出入。这是怎么回事呢。
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the reported torchrun command in run.py using MMMU_DEV_VAL and the llava_onevision_qwen2_0.5b_ov model. Compare the validation and dev metrics shown in the issue with the LLAVA-OneVision paper's reported 0.31, then identify which evaluation configuration or result differs. Done means the discrepancy is explained and the evaluation setup or expected result is clarified.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning, testing
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100