TinyLLaVA: A Framework of Small-scale Large Multimodal Models
This work addresses the problem of resource efficiency in multimodal AI for researchers and practitioners, offering a baseline for future studies, though it is incremental as it builds on existing LMM concepts.
The paper tackles the challenge of designing small-scale Large Multimodal Models (LMMs) by introducing the TinyLLaVA framework, which shows that with better data and training recipes, smaller models like TinyLLaVA-3.1B can achieve performance on par with larger 7B models such as LLaVA-1.5 and Qwen-VL.
We present the TinyLLaVA framework that provides a unified perspective in designing and analyzing the small-scale Large Multimodal Models (LMMs). We empirically study the effects of different vision encoders, connection modules, language models, training data and training recipes. Our extensive experiments showed that better quality of data combined with better training recipes, smaller LMMs can consistently achieve on-par performances compared to bigger LMMs. Under our framework, we train a family of small-scale LMMs. Our best model, TinyLLaVA-3.1B, achieves better overall performance against existing 7B models such as LLaVA-1.5 and Qwen-VL. We hope our findings can serve as baselines for future research in terms of data scaling, training setups and model selections. Our model weights and codes will be made public.