I repeated the training step and followed the scripts provided, but I found that the training loss rises suddenly. What could be the issue? Also, I noticed that the train_batch_size is set to 1 in your training process, and EGNN uses 12 layers. Could this easily lead to overfitting?
I repeated the training step and followed the scripts provided, but I found that the training loss rises suddenly. What could be the issue? Also, I noticed that the train_batch_size is set to 1 in your training process, and EGNN uses 12 layers. Could this easily lead to overfitting?
