Yaqing Wang
Authored Publications
Sort By
Preview abstract
Specialized Large multi-modal models (LMMs) have exhibited remarkable performance across numerous tasks, however, generalist LMMs suffer from performance degradation when training with a large collection of tasks. Recent research suggests Mixture of Experts (MoE) Models help instruction tuning, however, for LMMs of parameter size around O(50-100B), the prohibitive cost of replicating and storing the expert models severely limits the number of experts we can use.
We propose Omni-SMoLA that softly mixes many multimodal low rank experts to large models without introducing significant new parameter count compared to conventional MoE models. The core idea is that the large model provides a foundational backbone and different lightweight experts learn specialized knowledge residually. Extensive experiments demonstrate that the SMoLA approach helps improve the generalist performance across a broad range of visual question answering and captioning tasks, achieving a new state-of-the-art generalist performance that matches or outperforms single specialized LMM baselines.
View details
You Need Multiple Exiting: Dynamic Early Exiting for Accelerating Unified Vision Language Model
Shengkun Tang
Zhenglun Kong
Tianchi Zhang
Yao Li
Caiwen Ding
Yanzhi Wang
Dongkuan Xu
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023) (to appear)
Preview abstract
Large-scale Transformer models bring significant improvements for various downstream vision language tasks with a unified architecture. The performance improvements come with increasing model size, resulting in slow inference speed and increased cost for severing. While some certain predictions benefit from the full complexity of the large-scale model, not all of inputs need the same amount of computation to conduct, potentially leading to computation resource waste. To handle this challenge, early exiting is proposed to adaptively allocate computational power in term of input complexity to improve inference efficiency. The existing early exiting strategies usually adopt output confidence based on intermediate layers as a proxy of input complexity to incur the decision of skipping following layers. However, such strategies cannot apply to encoder in the widely-used unified architecture with both encoder and decoder due to difficulty of output confidence estimation in the encoder. It is suboptimal in term of saving computation power to ignore the early exiting in encoder component. To handle this challenge, we propose a novel early exiting strategy for unified visual language models, which allows dynamically skip the layers in encoder and decoder simultaneously in term of input layer-wise similarities with multiple times of early exiting, namely MuE. By decomposing the image and text modalities in the encoder, MuE is flexible and can skip different layers in term of modalities, advancing the inference efficiency while minimizing performance drop. Experiments on the SNLI-VE and MS COCO datasets show that the proposed approach MuE can reduce expected inference time by up to 50% and 40% while maintaining 99% and 96% performance respectively.
View details
