KADEL: Knowledge-Aware Denoising Learning for Commit Message Generation

This is the replication package for "KADEL: Knowledge-Aware Denoising Learning for Commit Message Generation" accepted to ACM Transactions on Software Engineering and Methodology (TOSEM).

Citation for this work can be found here.

Environment

Anaconda

conda create -n KADEL python==3.8.5 -y
conda activate KADEL
conda install pytorch==1.13 torchvision torchaudio cudatoolkit=10.0 -c pytorch -c nvidia -y
pip install tensorboard==2.4.1 tree-sitter==0.2.2 gsutil tqdm wandb notebook notebook gsutil scikit-learn protobuf protobuf==3.20.*

Data

Step 1: Download MCMD

# Downlaod MCMD filtered_data
cd $MCMD_PATH
mkdir filtered_data
wget https://zenodo.org/record/5025758/files/filtered_data.tar.gz
tar -zxvf filtered_data.tar.gz -C filtered_data/

Step 2: Preprocessing (e.g.: Get the indexes of commit which match the template (<type>(<scope>):<subject>))

# Time Cost: about 5 min per programming language.
LANGUAGE=javascript
python preprocess_MCMD.py --MCMD_data_folder_path $MCMD_PATH//filtered_data/$LANGUAGE/sort_random_train80_valid10_test10

Evaluation

Details can be seen at evaluation_results.ipynb

Thanks for the evaluation scripts from CommitMsgEmpirical!

Experiment

Our Model(full version)

cd sh/
python run_KADEL.py

Ablation Study

w/o knowledge version

cd sh/
python run_KADEL_without_knowledge.py

w/o denoising version

cd sh/
python run_KADEL_without_denoising.py

Thanks for the code from CodeT5!

Citation

If you use this code, please consider citing us:)

@article{KADEL_CMG_24,
  author    = {Wei Tao and
               Yucheng Zhou and
               Yanlin Wang and
               Hongyu Zhang and
               Haofen Wang and
               Wenqiang Zhang},
  title     = {KADEL: Knowledge-Aware Denoising Learning for Commit Message Generation},
  journal   = {{ACM} Trans. Softw. Eng. Methodol.},
  year      = {2024},
  month     = {jan},
  publisher = {Association for Computing Machinery},
  issn      = {1049-331X},
  url       = {https://doi.org/10.1145/3643675},
  doi       = {10.1145/3643675}
}