University of Twente Student Theses
Fine-tuning transformer models for commit message generation and autocompletion
Miksik, M. (2023) Fine-tuning transformer models for commit message generation and autocompletion.
|
PDF
1MB |
| Abstract: | Commit messages provide insight into the developer’s intentions and motivations — a fundamental source of information in the exceptionally collaborative discipline of software development. To assist the process of commit message writing, we fine-tune two transformer models for commit message generation and integrate them into popular code editors. We 1) collect and publish a dataset of commit message and patch pairs for 6 different programming languages, 2) fine-tune two generative language models for commit message autocompletion and generation tasks, and 3) provide integration for these models with popular code editors (IntelliJ, VSCode). Lastly, we show that on the test dataset, our fine-tuned models perform 2x and 10x times better than the base models for the completion and generation tasks, respectively. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Business & IT BSc (56066) |
| Link to this item: | https://purl.utwente.nl/essays/94380 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page