Skip to content

Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering

Notifications You must be signed in to change notification settings

bibisbar/LLaVA-Steering

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

LLaVA-Steering

Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering

News

  • Dec. 16rd, 2024: We release paper for LLaVA-Steering.
  • Code will be released asap after internal review.

Abstract

Multimodal Large Language Models (MLLMs) have significantly advanced visual tasks by integrating visual representations into large language models (LLMs). The textual modality, inherited from LLMs, equips MLLMs with abilities like instruction following and in-context learning. In contrast, the visual modality enhances performance in downstream tasks by leveraging rich semantic content, spatial information, and grounding capabilities. These intrinsic modalities work synergistically across various visual tasks. Our research initially reveals a persistent imbalance between these modalities, with text often dominating output generation during visual instruction tuning. This imbalance occurs when using both full fine-tuning and parameter-efficient fine-tuning (PEFT) methods. We then found that re-balancing these modalities can significantly reduce the number of trainable parameters required, inspiring a direction for further optimizing visual instruction tuning. Hence, in this paper, we introduce Modality Linear Representation-Steering (MoReS) to achieve the goal. MoReS effectively re-balances the intrinsic modalities throughout the model, where the key idea is to steer visual representations through linear transformations in the visual subspace across each model layer. To validate our solution, we composed LLaVA Steering, a suite of models integrated with the proposed MoReS method. Evaluation results show that the composed LLaVA Steering models require, on average, 500 times fewer trainable parameters than LoRA needs while still achieving comparable performance across three visual benchmarks and eight visual question-answering tasks. Last, we present the LLaVA Steering Factory, an in-house developed platform that enables researchers to quickly customize various MLLMs with component-based architecture for seamlessly integrating state-of-the-art models, and evaluate their intrinsic modality imbalance. This open-source project enriches the research community to gain a deeper understanding of MLLMs.

Installation

Step 1: Clone LLaVA-Steering repository:

git clone https://github.com/bibisbar/LLaVA-Steering.git
cd LLaVA-Steering

Step 2: Environment Setup:

Create and activate a new conda environment

conda create -n LLaVASteering
conda activate LLaVASteering

Install Dependencies

pip install -r requirements.txt

Quick Start

1. Datasets Preparation

Please refer to the Data Preparation section in LLaVA Documenation.

2. Train

Pretraining

Visual Instruction Tuning

The scripts will be released shortly after completing the code cleaning process.

3. Evaluation

The scripts will be released shortly after completing the code cleaning process.

Custom Your MLLM!

In the LLaVA Steering Factory, we establish standardized training and evaluation pipelines, along with flexible data preprocessing and model configurations. Our framework allows researchers to easily customize their models with various training strategies without the need for additional coding. We implement all mainstream LLMs and vision encoders, including multiple PEFT methods and our proposed MoReS technique. Furthermore, we support a wide range of benchmarks and integrate our intrinsic modality imbalance evaluation. The goal of the LLaVA Steering Factory is to facilitate research in MLLMs, particularly in addressing intrinsic modality imbalance to optimize visual instruction tuning.

Model Zoo

which are trained using LLaVA Steering Factory.

Train Recipe

The scripts will be released shortly after completing the code cleaning process.

Contact

If you have any questions, feel free to either initiate an Issue or contact us by email ([email protected]).

❤️ Community efforts

About

Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published