Dear Sir/Madam,
Thank you for your fantastic paper GearNet.
I wish to finetune Gearnet (distance prediction) for some other downstream tasks.
Now I have all atom positions and residue type and atom type. I used the same operation as fold3d to get protein. However, I do not know how can I get input (Tensor) – input node representations. I could not understand how those APIs passed this input into GearNet.
Yours Sincerely,
Jiale Zhao