Training problem about simplefold

[Paulie-ai]

Hi, I am trying to reproduce SimpleFold. I noticed that the residues generated using my weights have severe steric clashes (atoms are clumped together). My training MSE loss is around 0.3 before fine-tuning(lddt_loss: 0.23 and mse loss 0.026 ), after fine-tuning the lddt_loss was increased to ~0.5-0.7. Any advice on how to fix this ?

[yuyangw]

Hi @Paulie-ai, thanks for your interest in our work! Since a higher LDDT loss weight is applied in finetuning phase, it is normal to see an increased loss value. The training MSE loss and LDDT loss values look reasonable to me. What is the inference configuration used in your case?

[Paulie-ai]

Thanks for your reply. I rewrote the simplefold training and inference scripts, and the configuration filde is same as the original. The raw weight work fine when tested with my inference code. Therefore, could the issue be related to the batch size/the dataset size/epochs used for training?

[yuyangw]

Yes, it could be related to batch size / epochs. How are these hyper-parameters set in your case?