Issues with preprocessing the raw spider data for IRNet

[saxenakanishk]

Hello all,

I am having some difficulties preprocessing the raw spider data using the preprocessing scripts provided with irnet. After executing the run_me.sh file for preprocessing, I am getting an error message ("column * table error" ) for 72 queries but it also outputs 'Finished 7000 datas and failed 0 datas' message which is confusing me.

When I am looking at the preprocessed train.json (output) file and comparing it with the preprocessed train.json file provided by the authors, I could see a lot of data (values/ keys) missing like 'truth_label', 'stanford_dependencies', 'true_label' etc. for each database and hence there is also a difference in their file size. When I am training with my preprocessed file, I am getting 0.0 Accuracy which indicates there are some issues with the data files.

Any help or pointers would be appreciated.

Thanks, Kanishk

[liguozhanglearner]

I'm facing same problem, but having no idea.

[yucheng-zeng]

I'm facing same problem, anyone could help?