The repository holds the artifacts resulting from the research process for the dissertation paper entitled: CODE GENERATION FOR MOBILE LANGUAGES USING LARGE LANGUAGE MODELS, authored by Mircea Vasiliniuc and Prof. Adrian Groza (UTCN), published in 2023.
You will find here:
- the steps of training a GPT-2 model with public datasets for Swift and Kotlin languages in an attempt to provide code generation specific to mobile development languages.
- the complete dissertation and associated papers
- an extensive case study involving 16 participants from a software development department designed to understand the impact of using LLM trained for code generation in specific phases of the team, more specifically, technical onboarding and technical stack switch.
Other links:
- https://huggingface.co/mvasiliniuc (Huggingface datasets and models associated with the papers)
- https://arxiv.org/abs/2308.04736 (Chapter 5 extracted into an Arxiv Paper/Article)