What are the data splits and hyperparameters to get the reported results?

[LITONG99]

Dear authors, thank you for your inspiring works. I am especially interested in the fixed- and dynamic-context prompt LM tuning settings.

According to the paper, 15% of the original training set is reserved for validation. Could you please publish the new data splits? Also, what are the hyperparameters for every setting? These would be very helpful to align with the experiment results.

Thank you!

PS: The code cannot directly run, but I fixed it after checking with the PET repo.