Hi, thank you for sharing this great project!
I have a question regarding the atom-level model weights provided in antigenTCR/weights/atom_level_parameters.pt. Based on the training logic in run_antigenTCR_atom.py, it seems that each peptide is used as a validation set in a leave-one-out fashion, and a separate checkpoint is generated for each.
Could you clarify which specific peptide was used as the validation set to produce the shared atom_level_parameters.pt file? Or is this file a general pre-trained model not tied to any single validation peptide?
Thanks in advance for your help!
Hi, thank you for sharing this great project!
I have a question regarding the atom-level model weights provided in antigenTCR/weights/atom_level_parameters.pt. Based on the training logic in run_antigenTCR_atom.py, it seems that each peptide is used as a validation set in a leave-one-out fashion, and a separate checkpoint is generated for each.
Could you clarify which specific peptide was used as the validation set to produce the shared atom_level_parameters.pt file? Or is this file a general pre-trained model not tied to any single validation peptide?
Thanks in advance for your help!