fs_vid2vid model inference outputs are all 0

[Linxi-ZHAO]

When I use the given test dataset and follow the instructions to run inference.py for fs_vid2vid model, the outputs are all 0. I don't know where the error might be(pre-trained model parameters? or datatype?). Please give me some suggestions.
https://github.com/NVlabs/imaginaire/blob/master/projects/fs_vid2vid/README.md

[Linxi-ZHAO]

sorry, the outputs are NaN. After self.net_G_output = net_G(data_t) in vid2vid.py line 390, I printed self.net_G_output and found some items, like 'fake_images' , are NaN. I have no idea.