MoE-LLaVA: Mixture of Experts for Large Vision-Language Models
This addresses efficiency issues for researchers and practitioners in multi-modal AI by reducing training and inference costs while maintaining performance, though it is incremental as it builds on existing MoE and LVLM methods.
The paper tackles the high computational costs of scaling large vision-language models by proposing MoE-LLaVA, a sparse model that activates only top-k experts during deployment, achieving performance comparable to LLaVA-1.5-7B with 3B parameters and surpassing LLaVA-1.5-13B in object hallucination benchmarks.
Recent advances demonstrate that scaling Large Vision-Language Models (LVLMs) effectively improves downstream task performances. However, existing scaling methods enable all model parameters to be active for each token in the calculation, which brings massive training and inferring costs. In this work, we propose a simple yet effective training strategy MoE-Tuning for LVLMs. This strategy innovatively addresses the common issue of performance degradation in multi-modal sparsity learning, consequently constructing a sparse model with an outrageous number of parameters but a constant computational cost. Furthermore, we present the MoE-LLaVA, a MoE-based sparse LVLM architecture, which uniquely activates only the top-k experts through routers during deployment, keeping the remaining experts inactive. Extensive experiments show the significant performance of MoE-LLaVA in a variety of visual understanding and object hallucination benchmarks. Remarkably, with only approximately 3B sparsely activated parameters, MoE-LLaVA demonstrates performance comparable to the LLaVA-1.5-7B on various visual understanding datasets and even surpasses the LLaVA-1.5-13B in object hallucination benchmark. Through MoE-LLaVA, we aim to establish a baseline for sparse LVLMs and provide valuable insights for future research in developing more efficient and effective multi-modal learning systems. Code is released at https://github.com/PKU-YuanGroup/MoE-LLaVA.