-
Notifications
You must be signed in to change notification settings - Fork 7.5k
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
Fix PixelShuffle_icnr weight #3322
Conversation
…ight_norm and icnr_init modifies derived weight instead of weight_g and weight_v
Work in Progress. Checking out nbdev and following the development process used for fastai. |
…t test to test for weight_norm
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
Updated the notebook with the fix and also the unit test. The unit test passes with weight_norm enabled (earlier weight_norm was explicitly disabled in the unit test, even though the description said that weight_norm is what worked best with super_resolution) which is the expected behavior. |
… and disabled (both cases)
Updated PixelShuffle_icnr unit test to check for both cases, weight_norm enabled and disabled. Both the unit tests pass. |
Would you mind adding the spectral norm init also?
…On Thu, Apr 15, 2021, 8:50 AM pratX ***@***.***> wrote:
Updated PixelShuffle_icnr unit test to check for both cases, weight_norm
enabled and disabled.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#3322 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEMWOAK4ANPJKXU26YDGKX3TI2EDDANCNFSM42542OBA>
.
|
Added unit test for PixelShuffle_icnr with sepctral_norm. |
graet! |
Many thanks! |
When using weight_norm (the current default), PixelShuffle_icnr needs to icnr_init weight_g and weight_v.
Issue #3315