Implementation:Open compass VLMEvalKit TaiyiWrapper

Field	Value
source	VLMEvalKit
domain	Vision, API_Integration

Overview

TaiyiWrapper provides a VLMEvalKit API adapter for Megvii's Taiyi vision-language model.

Description

TaiyiWrapper inherits from BaseAPI and communicates with the Taiyi chat completions endpoint at taiyi.megvii.com. It supports custom prompt construction for Y/N, MCQ, and VQA dataset types, handles image-first reordering for single-image inputs, and encodes images to base64 for API transmission. Authentication uses the TAIYI_API_KEY environment variable.

Usage

Use this adapter when evaluating Megvii Taiyi vision-language models through the Taiyi API.

Code Reference

Source: vlmeval/api/taiyi.py, Lines: L1-185
Import: from vlmeval.api.taiyi import TaiyiWrapper

Signature:

class TaiyiWrapper(BaseAPI):
    def __init__(self, model='taiyi', retry=5, key=None, verbose=False,
                 system_prompt=None, temperature=0, timeout=60,
                 url='https://taiyi.megvii.com/v1/chat/completions',
                 max_tokens=1024, **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 = TaiyiWrapper(model='taiyi')
response = model.generate(message)

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment