Jun-Yan Zhu

We normalized the input from [0, 1] to [-1, 1]. To undo it, you can call tensor2im. https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/util/util.py#L9

Not sure what happended. Could you post the command line that you are using?

Which training command line did you use? Did you train a model on the default datasets?

There could be due to several reasons. 1. You may use different preprocessing codes during training and test time. You want to make sure that you use the same. One...

Could you share with us the training script and debugging information (loss plots)?

I am not quite familiar with the function `get_graph_node_names`. The model is not a network module. Maybe you could try `model.netG` or `model.netD`.

model.set_input(data) does not do too much work except moving data to gpu devices. Most of the preprocessing code is implemented in the data loader. You can also just modify the...

Would it be possible to provide more information? It's hard to identify the reasons without knowing your task (input and output), your data (dataset size), and your training and test...

#1445 looks like a good fix. Let us know if it works for @Sroebel. We are still evaluating whether we want to merge the PR as it may require many...

it might be a data loading issue. You may want to use SSD or other fast file systems.