SupplementaryMaterialsforM3ViT: Mixture-of-ExpertsVision TransformerforEfficientMulti-taskLearning withModel-AcceleratorCo-design