The task of KG-to-text generation is to produce sentences that accurately describe the information contained in a given knowledge graph. Pre-trained language models have become the mainstream method for handling the KG-to-text task because of their powerful capabilities in text understanding and generation. Nevertheless, when transforming the graphical information of the linearized knowledge graph into textual information via PLMs, issues such as the loss of triple structure information and the inadequacy of syntactic information will surface. Hence, we propose the Dual Attention KG-to-text Generation with Dual-view Graph Attention model. To avoid the additional representation alignment issue caused by the inconsistency between text embeddings and graph node embeddings, we put forward a graph-text dual attention mechanism, sharing text representations and graph representations, and subsequently employ the dual-view graph attention module to fuse graph structure information. Additionally, we introduce an enhanced pointer network to enhance the generated text’s accuracy and fluency. A significant amount of experiments conducted on the benchmark dataset show that our approach can significantly improve the performance of KG-to-text generation.