Deep learning task scheduling method based on reinforcement learning

The invention relates to a deep learning task scheduling method based on reinforcement learning. The method aims to face a deep learning multi-task scheduling scene, self-adaptively learn and adjust ascheduling strategy based on task online performance feedback, and improve the task completion efficiency and the utilization efficiency of cluster resources as much as possible. According to the method, adaptive learning is carried out based on performance online feedback of the deep learning task under different scheduling strategies, and the scheduling decision is updated adaptively, so that the task execution efficiency and the cluster resource utilization rate are maximized. The design and implementation of the method belong to the lightweight class, a programming mode and a task submission mode of a user do not need to be modified, and meanwhile, the method is friendly to operation and maintenance personnel and convenient and simple to deploy.