The default run results differed from the results reported in the paper #3
Description
Hello everyone, I encountered a problem while running your code. I ran the 'train.py' with default settings, but the results in 'wandb' differed from the results reported in the paper. Specifically, the result I obtained was 0.665(val_auc), while the result reported in the paper was 0.772. Therefore, I'm curious whether the hyperparameter settings in 'train.py' are the same as those in the paper, or if there are other reasons that prevent me from reproducing the results in the paper.