README.md

NumNet+

This is the official code repository for NumNet+(https://leaderboard.allenai.org/drop/submission/bm60vq8f7g2p7t2ld0j0). NumNet(https://github.com/ranqiu92/NumNet) was used as a basis for our work.

If you use the code, please cite the following paper:

@inproceedings{ran2019numnet,
  title={{NumNet}: Machine Reading Comprehension with Numerical Reasoning},
  author={Ran, Qiu and Lin, Yankai and Li, Peng and Zhou, Jie and Liu, Zhiyuan},
  booktitle={Proceedings of EMNLP},
  year={2019}
}

Requirements

pip install -r requirements.txt

Usage

Data and pretrained roberta-large preparation.

• Download drop data.

  wget -O drop_dataset.zip https://s3-us-west-2.amazonaws.com/allennlp/datasets/drop/drop_dataset.zip

  unzip drop_dataset.zip

• Download roberta model.

  cd drop_dataset && mkdir roberta.large && cd roberta.large

  wget -O pytorch_model.bin https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-pytorch_model.bin

• Download roberta config file.

  wget -O config.json https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-config.json

• Download roberta vocab files.

  wget -O vocab.json https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-vocab.json

  wget -O merges.txt https://s3.amazonaws.com/models.huggingface.co/bert/roberta-large-merges.txt

Train

sh train.sh 345 5e-4 1e-4 5e-4 0.01

Eval

• Save your model as model.pt.

  sh eval.sh drop_dataset/drop_dataset_dev.json prediction.json

  python drop_eval.py --gold_path drop_dataset/drop_dataset_dev.json --prediction_path prediction.json