open-compass / open-compass/VLMEvalKit

[Help Wanted] Supporting the `chat_inner` API for existing VLMs.

Open
#323 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Feature Request
Dominant language
Python
Stars
4.4k
Forks
768
Avg merge
1d 10h
Merged PRs (30d)
17

Description

Since we have now supported the multi-turn benchmark MMDU, we would like to implement the chat_inner function for existing VLMs in VLMEvalKit add support for multi-turn chatting.
Currently, we have already supported the method for: GPT series, Claude series, QwenVL APIs, qwen_chat, MiniCPM-v2.5, Idefics2_8b, llava-v1.5, deepseek-vl series. We need help from the community to support it for more models.

If you would like to help, you can refer to the development doc: https://github.com/open-compass/VLMEvalKit/blob/main/docs/en/Development.md.

Implementing three chat_inner api can be viewed as a major contribution. However, only implement this function if you are sure the VLM has the capability to handle the multi-turn input.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with docs/en/Development.md and inspect the existing chat_inner implementations for the model families listed in the issue. Select a VLM that can handle multi-turn input, implement its support, and validate the result against the MMDU multi-turn benchmark.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.