Naresh1318/Adversarial_Autoencoder

Bad looking losses

PierreBoyeau opened this issue · 3 comments

Good Morning!

I am trying to reproduce the very nice results you have in your tutorial but my training do not go as well when I execute python adversarial_autoencoder.py

Here are the losses I obtain:
image
I tried basically to decrease the learning rate without effect.
Can you help me?

Thanks in advance,

Pierre

Have you modified anything besides the learning rate?
Are you running on GPU or CPU?

Because those two things might have had an effect on why you got those losses...

Hey! Did you try looking at the generated images on Tensorboard?

Naresh, Thank you for writing your tutorial on AAEs. It's been very helpful. Unfortunately, I am running into the same problem as Pierre. See my scalar values below. Any insight you might have? Did you train for 1000 epochs or 300?

screen shot 2018-05-24 at 11 34 15 am
screen shot 2018-05-24 at 11 34 11 am
screen shot 2018-05-24 at 11 34 07 am

Thanks.
-Mimi