PyLoomer - The Automated Weaving of Python Text

The introduction of attention mechanisms has evolved NLP methodologies, giving rise to Large Language Models (LLMs) that excel in various sequence-to-sequence tasks like translation, question-answering, summarization, intent, and entity classification. This research focuses on text to code translation, converting human language into executable code. Code generating bots enhances the quality and efficiency of developers and facilitates to tackle the repetitive or simple tasks. A novel and clever fusion of Python Code and the creative ability of Large Language Models to generate text is proposed in this paper named PyLoomer. PyLoomer is an innovative text-to-python code chatbot that bridges the gap between narrative expression and technical execution which uses attention mechanism. PyLoomer seeks to improve the performance of existing models for python code generation by using in-context learning and instruction fine-tuning. Rigorous evaluation metrices, including Rougel, Rouge2, RougeL, RougeLsum and model loss, demonstrated significant improvements in results. Results indicate that CodGen, among the three models, achieves the highest Rouge scores, demonstrating superior performance, the Rouge score improvements for Codegen, StarCoder and CodeParrot are 0.0224, 0.0097 and 0.0223 respectively.