For this code comment matching task, an ablation study is apply to a NL-PL pretrained model, CodeBERT, to analyze which layers of the model are more important for Code Comment Matching task. Afterwards, a modified CodeBERT-based network is built, which makes the model smaller ,more efficient and more effective.
This repository is tested on Python 3.7, PyTorch 1.11.0, transformers 4.20.1 and tqdm 4.64.0 with GPU Tesla V100.
First, create a virtual environment with the version of Python 3.7.
Second, you should install PyTorch 1.11.0, transformers 4.20.1 and tqdm 4.64.0 in a virtual environment.
For example,
conda install pytorch==1.11.0
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Please make sure to update tests as appropriate.
MIT