MoTE: Mixture of Ternary Experts for Memory-efficient Large Multimodal Models

MoERe-assign community