SummaryIt is expected that the demand for quality of service (QoS) and quality of experience (QoE) in future 6G scenarios will continue to increase, and edge computing (EC) will continue to receive widespread attention. However, the highly dynamic and connectivity complexity of edge computing networks (ECNs) pose severe challenges to its resource allocation issues. In addition, network virtualization (NV) is widely applied to increase the flexibility of network architectures. Based on the above inspirations, we adopt virtual network embedding (VNE) to make decisions on the resource allocation of ECN. Specifically, we deploy a deep reinforcement learning (DRL)‐based multi‐layer policy network, which is applied to extract environmental information, calculate the available resources of edge nodes, and screen candidate edge nodes that satisfy resource allocation conditions. Second, the resources of edge links are allocated according to the shortest path algorithm. Finally, we build a simulation environment to demonstrate the advantages of the proposed policy network through rich experiments.