This repository contains notebooks that replicate the usage of adapters for the task of issue report classification.
The data used is available at: NLBSE'24 competition.
The Python packages required to run the code are listed in requirements.txt.
The requirements can easily installed via pip install -r requirements.txt.
For saving the model on Huggingface and for experiment tracking on wandb, you need to have your API keys.
The code is simplified and available in several Python Jupyter Notebooks, compatible to run on Google Colab.
| Notebook | Description |
|---|---|
| AdaptIRC_unsplit.ipynb | This notebook trains adapter per repo and does not split the training data into training and validation. |
| AdaptIRC_FT.ipynb | This notebook utilises full fine-tuning if the RoBERTa model. |
| AdaptIRC_SA.ipynb | This notebook trains a single adapter for all repositories. |
| eIRC_Demo.ipynb | This notebook shows a demo of the eIRC web application. |
The work during the NLBSE'24 competition is available on another branch at: AdaptIRC
The adapters created ara available on the following link: Adapters
The JSON format results are avaialble on the results directory at Results
All the results and metrics ara available at: IRC wandb
The eIRC web application demonstration is available on HuggingFace on the following link: eIRC
This repository used codes of different libraries and other notebooks:
The work is based on the NLBSE'24 competition.
@inproceedings{kallis2024nlbse,
title={The NLBSE'24 Tool Competition},
author={Kallis, Rafael and Colavito, Giuseppe and Al-Kaswan, Ali and Pascarella, Luca and Chaparro, Oscar and Rani, Pooja},
booktitle={Proceedings of the Third ACM/IEEE International Workshop on NL-based Software Engineering},
pages={33--40},
year={2024}
}It is also an extention of the AdaptIRC model.
@inproceedings{ebrahim2024few,
title={Few-Shot Issue Report Classification with Adapters},
author={Ebrahim, Fahad and Joy, Mike},
booktitle={Proceedings of the Third ACM/IEEE International Workshop on NL-based Software Engineering},
pages={41--44},
year={2024}
}The Journal will be shared later.