Visual instruction tuning has recently shown encouraging progress with open-source large multimodal models (LMM) such as LLaVA and MiniGPT-4. However, most existing studies of open-source LMM are performed using models with 13B parameters or smaller. In this paper we present an empirical study of scaling LLaVA up to 33B and 65B/70B, and share our findings from our explorations in image resolution, data mixing and parameter-efficient training methods such as LoRA/QLoRA. These are evaluated by their impact on the multi-modal and language capabilities when completing real-world tasks in the wild. We find that scaling LMM consistently enhances model performance and improves language capabilities, and performance of LoRA/QLoRA tuning of LMM are comparable to the performance of full-model fine-tuning. Additionally, the study highlights the importance of higher image resolutions and mixing multimodal-language data to improve LMM performance, and visual instruction tuning can sometimes improve LMM's pure language capability. We hope that this study makes state-of-the-art LMM research at a larger scale more accessible, thus helping establish stronger baselines for future research. Code and checkpoints will be made public.
@article{arxiv.2309.09958,
title = {An Empirical Study of Scaling Instruct-Tuned Large Multimodal Models},
author = {Yadong Lu and Chunyuan Li and Haotian Liu and Jianwei Yang and Jianfeng Gao and Yelong Shen},
journal= {arXiv preprint arXiv:2309.09958},
year = {2023}
}
Comments
Released at LLaVA Model Zoo: https://github.com/haotian-liu/LLaVA/blob/main/docs/MODEL_ZOO.md