PdD is an artifact for the paper "Formalizing Robustness Against Character-Level Perturbations for Neural Network Language Models" (ICFEM'23), which provides a tool for adversarial training by generating character-level perturbation