Hi! Great work and congratulations!
I noticed that your team also won the first price in Track 2(Illumination settings estimation), but I failed to find your code on Github. Could you tell me how I could find your code?

I was in the training phase when I finished the first epoch, and the second epoch started to run 5 iterations. The code stagnated. May I ask which part of the problem occurred? I tried to reduce the number of workers also did not work.

thanks for the source, and I`m new in deep learning
when training the model, I got an error that
"Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same"
which means that the model was not in gpu
but when I add "model.to('cuda')" the transfer it to my gpu, I stack with another error says
"Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!"
how can I solve it?
I run the code in centos7 with nvidia 2080ti

Hello, Thank you for your contribution!
When I testing the trained model, I found that the result of LPIPS is lower than your paper. The results just like

PSNR:17.586011871965436
SSIM:0.6088370930977324
LPIPS:0.3919581416580412

What could be causing it?

Hello！ First of all, thank you for your open source code, your project is great. But when I was reproducing your code, I could only set the batchsize to 2. I noticed that you stated in the paper that your mini-batch size is 6, use two 2080ti, I also use the same device to run but the batchsize can only be set to 2 (this is the default in the code), so are there any other settings that need to be changed?

Thank you for your nice contribution! BTW, could you please share the code to generate wide-range images used in stage 1?

python test.py

Thanks for your open source.
In train.py, I saw that the code commented out the part to update the discriminator weights.
I am a beginner, after I run train.py, do I need to comment out the updated part of the generator and let the discriminator train again, or is the code automatically trained alternately?