reproduction of the result on the paper and purporse of 'mean'

[myungjoon]

Thank you for your amazing work.

I clone the repository and run the code. The test error is about 0.208, which is much higher value than the value on the paper. How can I get the MAE value on the paper (0.029)?

Plus, what is the purpose of means in waveynet_trainer.py? It has value of [1e-3, 1e-3] and used when calculate Maxwell loss.

Thanks!