<p>Traditional vehicular edge computing (VEC) research does not fully consider the high-speed mobility and dynamics of the vehicular edge environment. At the same time, when using Deep Reinforcement Learning (DRL) to solve the vehicle task offloading problem, it was not fully considered that DRL requires a large amount of training data, and the computing resources of the vehicle and edge servers were not fully utilized, nor was the problem of coordinating cooperation between edge servers. To this end, this paper proposes a hybrid framework for vehicular edge computing offloading that combines informer-based prediction and deep reinforcement learning. To solve the problem of DRL requiring a large amount of training data, this paper designs a vehicular edge system based on digital twin assistance, using Digital Twin (DT) technology to obtain a large amount of real-time information. Considering the high-speed mobility and dynamics of the vehicular edge environment, this paper proposes a task prediction model based on Informer. Informer is used to predict the vehicular edge environment and assist the system in making accurate offloading decisions. To fully use the computing resources of the vehicle and each edge server, this paper adopts a partial offloading strategy and proposes a task offloading decision based on TD3. Based on the current environment of the vehicular edge system and the prediction results of Informer, the TD3 offloading model is used to determine the task offloading strategy. Experimental results show that the computing offloading scheme proposed in this paper performs better than others and can effectively reduce task processing delays and energy consumption. The key code for this article is available at <a href="https://github.com/UAMN-project/models">https://github.com/UAMN-project/models</a></p>