Abstract: The deployment of Vision Transformers (ViTs) on resource-constrained devices faces significant challenges due to the computational overhead of floating-point nonlinear operations (e.g., ...