Implementation:Open compass VLMEvalKit ARM thinker

Field	Value
source	VLMEvalKit
domain	Vision, API_Integration

Overview

ARM_thinker provides a VLMEvalKit API adapter for the ARM-Thinker agent-based vision-language model.

Description

ARM_thinker inherits from LMDeployWrapper (which itself inherits from BaseAPI) and extends it with ARM-Thinker agent capabilities from the InternLM ARM-Thinker repository. It requires an external agent_repo_root path pointing to a cloned ARM-Thinker installation and integrates the VerlAgent for agentic reasoning during evaluation.

Usage

Use this adapter when evaluating ARM-Thinker agent models that combine LMDeploy-served vision-language models with agentic reasoning tools.

Code Reference

Source: vlmeval/api/arm_thinker.py, Lines: L1-274
Import: from vlmeval.api.arm_thinker import ARM_thinker

Signature:

class ARM_thinker(LMDeployWrapper):
    def __init__(self, **kwargs): ...
    def generate(self, message, dataset=None): ...

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 = ARM_thinker(model='internvl2', api_base='http://localhost:23333/v1', agent_repo_root='/path/to/ARM-Thinker')
response = model.generate(message)

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment