KDE/C++ Code Summarization using LLM

[tm243 at yahoo.de]

Hi KDE developers,
I created a small dataset of code-description pairs of KDE/C++ code to see how a language model (LLM) can be used to describe KDE code. I fine-tuned a T5 model with this data, and the results are promising. The model is open-source and can run on a laptop.
However, the model's accuracy would significantly improve with a larger dataset. Therefore, I would like to invite anyone who is interested to add to the dataset. You can find the model and a link to the dataset here: https://github.com/tm243/CodeT5-KDE
Please note that this project is still in its early stages and the results may not be accurate in all cases.
Best,Thomas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-devel/attachments/20230508/6ee30d20/attachment.htm>