Neural network task processing method and device, electronic equipment and storage medium

• Main Authors: MA SONGCHEN, SHI LUPING, PEI JING, QU HUANYU, ZHAO RONG, ZHANG WEIHAO
• Format: Patent
• Language: Chinese; English
• Published: 23.09.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention relates to a neural network task processing method and device, electronic equipment and a storage medium, which are used for grouping a plurality of calculation cores included in the electronic equipment under the condition of receiving a plurality of neural network tasks to obtain calculation core groups for executing each neural network task, and determining an operation cycle corresponding to each neural network task. And executing the neural network task according to the corresponding operation cycle through each calculation core group. Wherein the calculation core groups carry out at least one time of data exchange in one operation cycle, and at least two calculation cores of which the data exchange moments are kept synchronous exist in each calculation core group. Different tasks are asynchronously processed through different calculation core groups, an independent running environment is developed for each task, and delay and unnecessary waiting caused by task switching are reduced. Meanwh