This repository contains the trained checkpoints corresponding to our paper RepoFusion: Training Code Models to Understand Your Repository. The released checkpoints are:

• RepoFusion_PPC: RepoFusion model trained with prompt proposal repo contexts. This is our best-performing model.
• RepoFusion_BM25: RepoFusion model trained with BM25 repo contexts.
• RepoFusion_RandomNN: RepoFusion model trained with RandomNN repo contexts.
• finetuned_codet5base_512: Our finetuned CodeT5-base model. This was used as initialization for our RepoFusion models.
• finetuned_codet5large_512: Our finetuned CodeT5-large model. This was used as a baseline.

For details of how these models were trained and evaluated, please check our paper RepoFusion: Training Code Models to Understand Your Repository.

Citation

@article{shrivastava2023repofusion,
  title={RepoFusion: Training Code Models to Understand Your Repository},
  author={Shrivastava, Disha and Kocetkov, Denis and de Vries, Harm and Bahdanau, Dzmitry and Scholak, Torsten},
  journal={arXiv preprint arXiv:2306.10998},
  year={2023}
}