Cao Ruisheng
Cao Ruisheng
It is sql acc (exact set match accuracy without values, but we use a small widely-used trick which returns the first prediction in the beam without execution error). Not Beam...
The official evaluation script (exact set match without values) does not compare ON conditions, thus for simplified outputs and better performance, we do not add this clause in the current...
Thanks a lot for pointing out this bug. We also find this problem when loading from checkpoints. Honestly, we never used this interface for training from checkpoints in our experiments...
I suppose these two statements are the same. If you mean why we do not directly use the CrossEntropyLoss API from pytorch for the main text-to-SQL loss, type constaints are...
Already on the schedule, wil let you know once prepared.
Working towards that. Hopefully, before the end of August this year. Will let you know.
Working towards that. Codes will be released if everything is ok.
This is little weird. My collaborator can re-train and reproduce the experiment~(even a little higher), and the open-sourced Electra-large model **should** achieve 75% EM acc on dev set. 1. Please...
Sorry for the late reply. I just checked my experiment logs. It takes roughly 1200 seconds for training with `large`-series PLM per epoch, with running script in `run/run_train_lgesql_plm.sh`. Sadly, it...
One GeForce RTX 2080 Ti for GLOVE and base PLM, one Tesla V100-PCIE-32GB for large PLM. 50g is enough for CPU memory.