sgl-project/sglang

[Bug] Mllama ambiguous multi-image behaviour

Open

#8,174 opened on Jul 19, 2025

View on GitHub
 (3 comments) (0 reactions) (0 assignees)Python (6,216 forks)auto 404
Multi-modalgood first issuehelp wanted

Repository metrics

Stars
 (28,442 stars)
PR merge metrics
 (Avg merge 2d 1h) (1,000 merged PRs in 30d)

Description

Checklist

  • 1. I have searched related issues but cannot get the expected help.
  • 2. The bug has not been fixed in the latest version.
  • 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.
  • 4. If the issue you raised is not a bug but a question, please raise a discussion at https://github.com/sgl-project/sglang/discussions/new/choose Otherwise, it will be closed.
  • 5. Please use English, otherwise it will be closed.

Describe the bug

Run:

cd /test/srt
python -m unittest test_vision_openai_server_a.TestMllamaServer.test_multi_images_chat_completion

Get: Multi-image response: The first image shows a man in a yellow shirt and shorts, standing on a beach with a surfboard under his arm, looking out at the ocean. The second image shows a yellow taxi cab with the letters "SGI" on the side, and a person in a yellow shirt and shorts standing next to it.

Which is not correct, given these two pics: pic1

pic2

Reproduction

.

Environment

.

Contributor guide