Orest Kupyn

icon location Oxford icon location DPhil student at Oxford focusing on generative models for computer vision and deep learning. Creator of DeblurGAN and other open-source CV tools

Results 14 comments of Orest Kupyn

when training, an error occurred

do you use the correct package versions (specifically pytorch-lightning) from the requirements file?

There is a "RuntimeError" when I use this command to run the program, my environment is windows10 64bit system.

Hello, unfortunately I am not able to test it on Windows but it seems that you have troubles with multiprocessing, try this - https://medium.com/@mackie__m/running-a-cifar-10-image-classifier-on-windows-with-pytorch-9094e29089cd

RuntimeError: The size of tensor a (256) must match the size of tensor b (253) at non-singleton dimension 3

fixed this issue in latest commit

generate wrong blurred image

@nickolaymykhalych

When I run the model ,I meet the error：

Could you provide more details? What is the size of your images? I would try to reproduce it

Details on the pre-trained weights in the Dropbox Links

Hello, The weights on dropbox are corresponding to DeblurGAN WILD model, please send me the message if you have any problems with that, I would double check the weights.

Are the generator weights are only kept in the "test" runtime code in this repo or We have to take care of it?

You need to put the Generator weights into the appropriate place and load the model in test time. After that, the 'fake' output should be deblurred version of the image....

module 'torch._C' has no attribute '_cuda_setDevice'

Could you please specify the error and the command you use to run the code?

About the release weight

Updated in latest commit

unexpected key "model.0.weight" in state_dict

We provide the generator model, however, I have some problems with Dropbox so it was unavailable. I already updated the link to the model, please send me the message if...