Can't reproduce the same psnr result with paper
fefang opened this issue · comments
I try to train the model using the training data in this git and use the default train config, but when testing the testing data, the reproduced model has lower psnr result(36.3), compared with the pretrained model(37.3).
Do I need to adjust the training config, and how to adjust it, to get the same model with pretrained model?