LMFusion: Adapting Pretrained Language Models for Multimodal Generation

TL;DR

LMFusion is a novel framework that enhances pretrained text-only language models with multimodal capabilities, enabling efficient understanding and generation of both text and images without retraining the entire model.

Contribution

It introduces a method to adapt existing LLMs for multimodal tasks by adding parallel transformer modules for images, training only the new modules while freezing the original language model.

Findings

Improves image understanding by 20%

Enhances image generation by 3.6%

Uses only 50% of the FLOPs compared to training from scratch

Abstract

We present LMFusion, a framework for empowering pretrained text-only large language models (LLMs) with multimodal generative capabilities, enabling them to understand and generate both text and images in arbitrary sequences. LMFusion leverages existing Llama-3's weights for processing texts autoregressively while introducing additional and parallel transformer modules for processing images with diffusion. During training, the data from each modality is routed to its dedicated modules: modality-specific feedforward layers, query-key-value projections, and normalization layers process each modality independently, while the shared self-attention layers allow interactions across text and image features. By freezing the text-specific modules and only training the image-specific modules, LMFusion preserves the language capabilities of text-only LLMs while developing strong visual understanding and generation abilities. Compared to methods that pretrain multimodal generative models from scratch, our experiments demonstrate that, LMFusion improves image understanding by 20% and image generation by 3.6% using only 50% of the FLOPs while maintaining Llama-3's language capabilities. We also demonstrate that this framework can adapt existing vision-language models with multimodal generation ability. Overall, this framework not only leverages existing computational investments in text-only LLMs but also enables the parallel development of language and vision capabilities, presenting a promising direction for efficient multimodal model development.