The models does not convergence

[wanmeiziyoulan]

Hi, when I train the model with using SHA, SHB or my own datasets, the loss will never show a downward trend and et_cnt also keep empty. I I keep the default parameter settings.
I would appreciate your reply! Thank you!

[Song-gusu]

yeah. when i trained the model on SHA with eopch = 450 , i got the same problem with you. After that i runned nowtest.py and the final mae is very high，it is quite different from the experimental results in this paper