CATH dataset pretraining hyper-params

[yzf-code]

Dear authors,
I have followed you description in the README to download the CATH42 datasets and processed it into protein.rball.edges.CATH42.npy, protein.knn.edges.CATH42.npy, protein.nodes.CATH42.pt.
I directly use your codes and the following commands:
python -B train.py --dataset SHS27k --split_mode random --pre_train CATH42
python -B train.py --dataset SHS27k --split_mode dfs --pre_train CATH42
python -B train.py --dataset SHS27k --split_mode bfs --pre_train CATH42

But sadly I found that my test results are much lower than those reported in your paper.

I think this may be caused by the inconsistency between the hyperparameters for pretraining with CATH42 and the hyperparameters without CATH42 which is provided in your code. Could you please provide the hyperparameters for pretraining with CATH42?
Thank you very much！