about the 'col_pred' key in the folder 'data' for the bert version

[Twsschx]

I have run the code with the version of bert. But wthat does the key 'col_pred' in the dev and train data mean during the "to_batch_seq" procedure and how to generate these values? Because I couldn't find the data preprocess for this part.

[slamandar]

Hi, have you solved this problem? I met this problem as well and still do not solve it.

[AoZhang]

'col_pred' will only be used in inference mode. so you can just use IRNet released dev set instead of preprocessing by yourself.

[Twsschx]

Yes. Can you explain how to get the ‘col_pred’ result? Thanks a lot.

2020年6月11日 上午11:05，ZHANG Ao notifications@github.com 写道：

'col_pred' will only be used in inference mode. so you can just use IRNet released dev set instead of preprocessing by yourself.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/microsoft/IRNet/issues/42#issuecomment-642377555, or unsubscribe https://github.com/notifications/unsubscribe-auth/AK4XKEUMM4KZYFI4M6BKTMLRWBCXTANCNFSM4NMZKIKQ.

[marlesson]

+1

[papsipublic]

I would like to run my own question and to see the sql. Is this column, "col_pred" important for new questions? If yes, how can we generate it?

I would appreciate your answer very much.

[ygxw0909]

I also want to preprocess my own dataset, how can I get "col_pred"?