LGESQL validation with Cspider datasets, upgrade Pytorch Version to 1.13.Cspider need a pre-trained LM to encode both Chinese and English, so we used multilingual-BERT instead.

Dev dataset exact match/checker/beam acc is 0.5551/0.5609/0.6470

This is the project containing source code for the paper LGESQL: Line Graph Enhanced Text-to-SQL Model with Mixed Local and Non-Local Relations in ACL 2021 main conference. If you find it useful, please cite our work.

    @inproceedings{cao-etal-2021-lgesql,
            title = "{LGESQL}: Line Graph Enhanced Text-to-{SQL} Model with Mixed Local and Non-Local Relations",
            author = "Cao, Ruisheng  and
            Chen, Lu  and
            Chen, Zhi  and
            Zhao, Yanbin  and
            Zhu, Su  and
            Yu, Kai",
            booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
            month = aug,
            year = "2021",
            address = "Online",
            publisher = "Association for Computational Linguistics",
            url = "https://aclanthology.org/2021.acl-long.198",
            doi = "10.18653/v1/2021.acl-long.198",
            pages = "2541--2555",
    }

The following commands are provided in setup.sh.

1. Firstly, create conda environment text2sql:

• In our experiments, we use torch==1.13.0 and dgl==1.0.0 with CUDA version 11.6

• We use one GeForce RTX 2080 Ti for GLOVE and base-series pre-trained language model~(PLM) experiments, one Tesla V100-PCIE-32GB for large-series PLM experiments

  conda create -n text2sql python=3.8
  source activate text2sql
  pip install torch==1.13.0 torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
  pip install -r requirements.txt
  

2. Next, download dependencies:

    python -c "import stanza; stanza.download('en'); stanza.download('zh', processors='tokenize,pos')"
    python -c "from embeddings import GloveEmbedding; emb = GloveEmbedding('common_crawl_48', d_emb=300)"
    python -c "import nltk; nltk.download('stopwords')"
   

3. Download pre-trained language models from Hugging Face Model Hub, such as bert-large-whole-word-masking and electra-large-discriminator, into the pretrained_models directory. The vocab file for glove.42B.300d is also pulled: (please ensure that Git LFS is installed)

    mkdir -p pretrained_models && cd pretrained_models
    git lfs install
    git lfs clone https://huggingface.co/bert-base-multilingual-uncased
   

1. Download, unzip and rename the [Cspider.zip] into the directory data.

2. Preprocess the train and dev dataset, including input normalization, schema linking, graph construction and output actions generation. (Our preprocessed dataset can be downloaded here)

    ./run/run_preprocessing.sh
   

Training LGESQL models with GLOVE, BERT and ELECTRA respectively:

• msde: mixed static and dynamic embeddings

• mmc: multi-head multi-view concatenation

  ./run/run_lgesql_plm.sh [mmc|msde] bert-base-multilingual-uncased
  

1. Create the directory saved_models, save the trained model and its configuration (at least containing model.bin and params.json) into a new directory under saved_models, e.g. saved_models/electra-msde-75.1/.

2. For evaluation, see run/run_evaluation.sh and run/run_submission.sh (eval from scratch) for reference.

We would like to thank Tao Yu, Yusen Zhang and Bo Pang for running evaluations on our submitted models. We are also grateful to the flexible semantic parser TranX that inspires our works.