This repository contains open science data used in the paper
M. Lajkó, V. Csuvik and L. Vidacs, Towards JavaScript program repair with Generative Pre-trained Transformer (GPT-2)
submitted at the 3rd International Workshop on Automated Program Repair (APR 2022).
The repository contains 3 folders:
- source : contains the plain code, in source/README there is a descripton about how to run the code (running the code will create addition files, checkpoints etc. in source)
- APR_Data: Contains the dataset (gitter_commits), running the code described above will create a no comment version of this dataset and all experiment directories will be created in this folder (the experiment directory's name can be given to the code as a parameter: example experiment dir can be found below)
- paper_exp_dir: This is the experiment directory which is the base of the paper submitted.
In each directory we provide a README file that describes the structure of the folder in question.
- APR22-JS-GPT/paper_exp_dir/candidates.txt: Examples in Qualitative evaluation section can be found here
- APR22-JS-GPT/paper_exp_dir/Matches.txt: Base of Table 1 in Quantitative evaluation
The code was designed to work for all kind of programming languages, although we made experiments for javascript only.