Implementation:Open compass VLMEvalKit TaichuVLWrapper

Field	Value
source	VLMEvalKit
domain	Vision, API_Integration

Overview

TaichuVLWrapper provides a VLMEvalKit API adapter for the Taichu vision-language model series.

Description

TaichuVLWrapper inherits from BaseAPI and communicates with the Taichu VL API endpoint. It supports custom prompt construction for MCQ, VQA, and Y/N dataset types, with special handling for system prompts, hint prompts, and multi-choice prompts. The class optionally integrates OpenAI-based evaluation when an OpenAI key is available.

Usage

Use this adapter when evaluating Taichu VL models (such as Taichu-VL-2B) through the Taichu API service.

Code Reference

Source: vlmeval/api/taichu.py, Lines: L1-355
Import: from vlmeval.api.taichu import TaichuVLWrapper

Signature:

class TaichuVLWrapper(BaseAPI):
    def __init__(self, model='Taichu-VL-2B', retry=5, verbose=True,
                 temperature=0.0, system_prompt=None, max_tokens=4096,
                 key=None, url=None, **kwargs): ...
    def generate_inner(self, inputs, **kwargs): ...

I/O Contract

Direction	Description
Inputs	message — text/image/video content list; model-specific params via kwargs
Outputs	generate() returns str prediction; generate_inner() returns (int, str, str) tuple

Usage Examples

# Example instantiation
model = TaichuVLWrapper(model='Taichu-VL-2B')
response = model.generate(message)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment