Implementation:Open compass VLMEvalKit DoubaoVLWrapper

Field	Value
source	VLMEvalKit
domain	Vision, API_Integration

Overview

DoubaoVLWrapper provides a VLMEvalKit API adapter for ByteDance Doubao vision-language models via the Volcengine Ark API.

Description

DoubaoVLWrapper inherits from BaseAPI and uses the OpenAI-compatible client to communicate with the Volcengine Ark API endpoint. It supports model-specific endpoint configuration through environment variables (DOUBAO_VL_ENDPOINT_*), handles image dumping and base64 encoding, and provides custom prompt construction for MCQ datasets. Authentication requires the DOUBAO_VL_KEY environment variable.

Usage

Use this adapter when evaluating Doubao vision-language models (such as Doubao-1.5-vision-pro) through the Volcengine Ark API.

Code Reference

Source: vlmeval/api/doubao_vl_api.py, Lines: L1-210
Import: from vlmeval.api.doubao_vl_api import DoubaoVLWrapper

Signature:

class DoubaoVLWrapper(BaseAPI):
    def __init__(self, model='', retry=5, verbose=True, system_prompt=None,
                 temperature=0, timeout=60, max_tokens=4096,
                 api_base='https://ark.cn-beijing.volces.com/api/v3',
                 **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 = DoubaoVLWrapper(model='Doubao-1.5-vision-pro')
response = model.generate(message)

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment