Enhancing intelligent IoT services development by integrated multi-token code completion

Yu Xia, Tian Liang, Weihuan Min, Li Kuang, Honghao Gao

Abstract: Highlights•Code completion techniques in IoT service layer accelerate development and enhance intelligent sensing.•Graph Neural Networks improve smaller language models for multi-token code completion.•The S2RCC framework predicts code skeleton and refines it for accurate multi-token code completion.•Incorporating a modified Transformer, GGNN, and a heterogeneous code graph, demonstrates superior performance in multi-token code completion.