Parm: Efficient Training of Large Sparsely-Activated Models with Dedicated Schedules This code repository is for 'Parm: Efficient Training of Large Sparsely-Activated Models with Dedicated Schedules,' a paper accepted at INFOCOM 2024.