In industrial-grade applications, the efficiency of algorithms and models takes precedence, ensuring a certain level of performance while aligning with the specific requirements of the application and the capabilities of the underlying equipment. In recent years, the Vision Transformer has been introduced as a powerful approach to significantly improve recognition accuracy in various tasks. However, it faces challenges concerning portability, as well as high computational and input requirements. To tackle these issues, a dynamic adaptive transformer (DAT) has been proposed. This innovative method involves dynamic parameter pruning, enabling the trained Vision Transformer to adapt effectively to different tasks. Experimental results demonstrate that the dynamic adaptive transformer (DAT) is capable of reducing the model’s parameters and Gmac with minimal accuracy loss.