Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Generating black images and model not getting trained #4

Open
musafirsafwan opened this issue Jan 15, 2017 · 9 comments
Open

Generating black images and model not getting trained #4

musafirsafwan opened this issue Jan 15, 2017 · 9 comments

Comments

@musafirsafwan
Copy link

Same issue as #1

Downloaded the code and ran it without any editing.
The saved images after each epoch all are completely black. The numbers reported after each epoch are almost same. Issue is not solved even after updating to the latest versions of tensorflow and keras.
Versions:
tensorflow-gpu=0.12.1
Keras=1.2.0

@tpbarron
Copy link

I have tried to run the code and am experiencing the same problem. After 10, 20, and 30 batches the images look like this:

plot_epoch_000_generated

plot_epoch_001_generated

plot_epoch_002_generated

@maayanfa
Copy link

Also for me with latest versions for tensorflow-gpu and keras

@newikov
Copy link

newikov commented Feb 7, 2017

Also for me :((( help guys!

@dremovd
Copy link

dremovd commented Feb 10, 2017

+1

@dremovd
Copy link

dremovd commented Feb 11, 2017

@newikov @maayanfa @tpbarron @musafirsafwan
Managed to reproduce original results by decreasing learning rates, see #5

@sautin1
Copy link

sautin1 commented Mar 14, 2017

Problem is still actual for me (even after decreasing the lr).
Win10, Python 3.5.2 | Anaconda 4.2.0 (64-bit)
tensorflow-gpu==0.12.1,
Keras==1.2.2

@xhchrn
Copy link

xhchrn commented Mar 30, 2017

dremoved's fix is right. I decrease learning rate from 0.0002 to 0.0001, and the model works.

@xuewenyuan
Copy link

xuewenyuan commented Apr 9, 2017

I have the same problem, which are solved when I decrease learning rate to 0.00002.

@showskyzt
Copy link

Thanks to #4, I changed the lr from 0.0002 to 0.00005, and the model works!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

9 participants