MOSAIC: Efficient Collaborative Inference for Multi-Modal Large Language Models with Distributed Edge Computing
MOSAIC is presented, which adopts a component-level decomposition and distributed deployment paradigm for efficient MLLM inference at the edge, realizing three core ideas: parallel encoding, token pruning, and device-server collaboration.
Kongyange Zhao, Zhi-Xuan Liao, Sheng-Yuan Ye et al.
· 0 citations