univlm 1.0.2

Unified wrapper for HF, VLLM and other models

Univlm Model Framework

Description

The Univlm Model Framework provides a flexible and extensible system for loading, processing, and performing inference across various AI models. It aims to simplify interaction with multiple model types and platforms, offering a unified pipeline for Vision-Language Models (VLM). This makes it easier to integrate visual and linguistic information for a range of tasks, such as image captioning, visual question answering, and more.

Our framework supports a variety of models, including those from Hugging Face (HF), VLLM, and other models not natively supported on either platform. This flexibility allows you to easily load and use models from diverse sources, whether they are available on HF, VLLM, or custom-built models that don’t fit into standard frameworks.

Prerequisites

• We strongly recommend using Conda for a virtual environment. See the Conda Installation Guide.
• OS: Linux

Installation

A step-by-step guide on how to install the software.

1. Install using pip

pip install univlm

2. Install external files (one-time setup)

univlm-install

Quick Start

Refer to the documentation Examples.

Contributing

Contributions will be welcomed once the project is finalized.

License

This project is licensed under the Apache License, Version 2.0, January 2004. For more details, see: Apache License 2.0.

Contact

For any inquiries, reach out via email:

• Aryan Singh: [e-mail]
• Ilia Davydov: [e-mail]
• Siddhant Tyagi: [e-mail]

LinkedIn Profiles

• Project Mentor: Imbesat Rizvi
• Aryan Singh
• Ilia Davydov
• Siddhant Tyagi