Hi there,
I wonder is there any other difference between GOOD-HIV and mobg-molhiv except for the split setting? Because I found the dataset size of GOOD-HIV and ogbg-molhiv are equal (41127) and both of them are adapted from MoleculeNet so I assume the full set of them are the same at first.
But if I pre-train a model on the ogbg-molhiv dataset and then finetune it on the GOOD-HIV dataset, an error will be reported. Is this because the node features of GOOD-HIV and ogbg-molhiv are different?
Hi!
The full set of GOOD-HIV is the same as the original one. Our node features follows PyG while OGB defines their node features OGB. I think it is the dilemma you may have to tackle.
Hi there, I wonder is there any other difference between GOOD-HIV and mobg-molhiv except for the split setting? Because I found the dataset size of GOOD-HIV and ogbg-molhiv are equal (41127) and both of them are adapted from MoleculeNet so I assume the full set of them are the same at first. But if I pre-train a model on the ogbg-molhiv dataset and then finetune it on the GOOD-HIV dataset, an error will be reported. Is this because the node features of GOOD-HIV and ogbg-molhiv are different?