Transformers based Python Code Generation from Natural Language

Code generation is the process of creating source code by utilizing user inputs in either natural language or code form to produce code snippets. It facilitates expedited completion of programming jobs for developers by lowering code development time and minimizing maintenance expenses. This allows them to concentrate on the strategic aspects of their business and innovative problem-solving.Transformers is the cutting-edge model for natural language challenges. This work utilizes the capabilities of transformers to mitigate long-term dependencies and exploit attention mechanisms to enhance programme structural context. Our technique produces code by forecasting the structure of the abstract syntax tree of the desired programme. We employ a BERT encoder and a grammar-based decoder to forecast the abstract syntax tree. This paper conducts a comprehensive assessment of the model’s performance by utilizing the CoNaLa and Django datasets.