contextual_loss_pytorch icon indicating copy to clipboard operation
contextual_loss_pytorch copied to clipboard

Published 20 hours ago • verified publisher icon S-aiueo32

Fixed some bugs about parameters and updated comments

Open Kitsunetic opened this issue 4 years ago • 0 comments

I found some bugs and fixed it. Please check my pull request.

  1. Parameter loss_type is being ignored in both ContextualLoss and ContextualBilateralLoss.
  2. Parameter band_width in ContextualBilateralLoss is treated as weight_sp.
  3. Parameter loss_type is better to compared in lower case.
  4. Default value of parameter band_width in ContextualBilateralLoss is 0.5 but F.contextual_bilateral_loss have default value of 1.0.

Thank you.

Kitsunetic avatar Mar 22 '20 01:03 Kitsunetic