Abstract: Drug design in molecular design aims to produce molecules with desirable biological activity and physicochemical properties. With the rapid development of computer science and high performance computing, the application of deep learning technologies in the field of drug design is gaining increasing recognition. Generative deep learning models have demonstrated remarkable performance in such fields as natural language, image, music, and video, providing new ideas for molecule generation. More and more researchers have started to use deep learning technologies to complete molecule generation tasks. This article summarizes the research progress of deep learning algorithms in molecule generation, focusing on the principles, applications, molecular representation forms, and technical details of several commonly used neural network architectures for molecule generation.