edge model does not converge

[Turlan]

Hi, thanks for releasing codes. I am using this code to train my own data from scratch . After I changed the sigma of canny calculator to 1, the feature matching loss keeps increasing at first 10 epochs. I am wondering if it's a normal phenomenon ?

[neouyghur]

@Turlan my curve is similar to you. I think it is because discriminator is learning in the beginning.

[12ycli]

How can I draw the curve like you ?

[12ycli]

How can I draw the curve like you ?

[liuqi-liuqi]

Hi, thanks for releasing codes. I am using this code to train my own data from scratch . After I changed the sigma of canny calculator to 1, the feature matching loss keeps increasing at first 10 epochs. I am wondering if it's a normal phenomenon ?

[liuqi-liuqi]

How can I draw the curve like you ?

[12ycli]

I offer my solution here. I just use pandas to read the log file and then draw them. However, you better change the EVAL_INTERVAL and change eval() function to log the validation dataset, which can avoid overfitting.

[liuqi-liuqi]

I offer my solution here. I just use pandas to read the log file and then draw them. However, you better change the EVAL_INTERVAL and change eval() function to log the validation dataset, which can avoid overfitting.

thanks!

[DwanZhang-AI]

@Turlan my curve is similar to you. I think it is because discriminator is learning in the beginning.

I meet the same problem, will you froze the discriminator for the first several iters?