Face attribute editing is a popular direction in face generation. Its purpose is to modify the face attributes in a face image while keeping other attributes unchanged. However, editing multiple facial attributes often results in unintentional alterations to other unedited attributes. This primarily occurs because current generative models do not consider the hierarchical relationships among facial attributes, treating each attribute independently. Consequently, even models that achieve semantic disentanglement can still experience editing distortions. To address this issue, we propose a method that automatically discovers hierarchical attribute relationships in this paper. Specifically, We select 13 significantly varying attribute labels on the CelebA dataset and designed a hierarchical detection module that leverages prior knowledge and trained high-precision classifiers to demonstrate features affected by the interrelationships of face attributes. Easily learn the hierarchical relationship between attributes. In addition, we comprehensively verified the learned hierarchical relationship structure through quantitative experiments and qualitative experiments and achieved better division results than the baseline model.