search

c-yn / SFNet

[ICLR 2023] Selective Frequency Network for Image Restoration
MIT License
97 stars 9 forks source link

bug #2

Closed peylnog closed 1 year ago

peylnog commented 1 year ago

I tried to run python3 main.py in motion deblur, but I had the following problem at this

ValueError: Expected more than 1 value per channel when training, got input size torch.Size([1, 72, 1, 1])

This error seems to be because, in the training phase, batchnorm receives input tensor ( bxcx1x1 ).

I would appreciate it if you could tell me how to debug or update the repo

stacyyy214 commented 2 months ago

I tried to run python3 main.py in motion deblur, but I had the following problem at this

ValueError: Expected more than 1 value per channel when training, got input size torch.Size([1, 72, 1, 1])

This error seems to be because, in the training phase, batchnorm receives input tensor ( bxcx1x1 ).

I would appreciate it if you could tell me how to debug or update the repo

I've encountered this issue too. How did you solve it?