Reproduce the same results

[y0un0]

Hello,
First of all, thank you for sharing your work.
I'm making this issue because i have a hard time reproducing the same results as yours. For the training i used 2200 samples from NJUD and NLPR datasets. I trained the model for 20500 iterations but it seems like the model has not fully converged because i produce this salmap

instead of this one :

What is the converged loss ?
Do you have any suggestions ? Maybe share your training set ?

[y0un0]

I solved the problem