Home

Awesome

Plug-and-Play Adaptation for Continuously-updated QA

This includes an implementation of "Plug-and-Play Adaptation for Continuously-updated QA"

Executing program

To train the model with source Knowledge, please use the following command

python pretrain.py --train_path $train_path --dev_path $dev_path

Here are some arguments(included but not all) which might be useful.

To update the model with target Knowledge, please use the following command

python update.py --checkpoint $model_checkpoint --train_path $train_path --dev_path $dev_path --adapter $adapter --freeze_orig_param $params

Here are some arguments(included but not all) which might be useful.

To evaluate the model, please use the following command

python eval.py --checkpoint $model_checkpoint --dev_path $dev_path --adapter $adapter

Here are some arguments(included but not all) which might be useful.

Version History

Citation

If you find this repo useful, please cite our preprint:

@article{lee2022plug,
  title={Plug-and-Play Adaptation for Continuously-updated QA},
  author={Lee, Kyungjae and Han, Wookje and Hwang, Seung-won and Lee, Hwaran and Park, Joonsuk and Lee, Sang-Woo},
  journal={arXiv preprint arXiv:2204.12785},
  year={2022}
}

License

Copyright
Copyright 2022-present SNU-NAVER Hyperscale AI Center

Acknowledgement
This research was supported by SNU-NAVER Hyperscale AI Center, and IITP grants funded by the Korea government (MSIT) [2021-0-02068 SNU AIHub, IITP-2022-2020-0-01789].