This is the implementation of EMNLP 2020 paper IGSQL: Database Schema Interaction Graph Based Neural Model for Context-Dependent Text-to-SQL Generation

folder data/ data/database model/bert/data/annotated_wikisql_and_PyTorch_bert_param/pytorch_model_uncased_L-12_H-768_A-12.bin can be found here

use ./run_sparc.sh or ./run_cosql.sh to train and evaluate on the best model.

use ./test_sparc.sh or ./test_cosql.sh (change the save_file path) to evaluate the specific model.

We adapt the code from editsql. Thanks to the open-source project.