Hi, I noticed that the repo only provides models trained on the CleanSplit dataset; would it be possible to also share checkpoints trained on the original PDBbind training set, as well as the Pafnucy checkpoint, since I’d like to run a benchmark comparison?