Pix2pix inference results in the same outputs

I am using Pix2pix model to generate new images based on my data (image). I am using pytorch_pix2pix. I trained the model using my own data and get pretty good results. However, I have a problem of having the same outputs the same input images even though they are trained using images with various shapes.

I think it is an issue to have the same seed for the inference but I could not find any seed variable to change in the test.py. If there is any tip to get various results please let me know. Thank you in advance.

Is there any variable (for example, seed) to get various images during inference?