different lpips distance value obtained on rerun with same set of test images

[PoulamiSM]

Thank you for the code. I am training conditional GAN. Whenever I run the code using same set of images, on each run the distance values between a given pair of images changes. The change is not too significant (order of 10 e-03) , but I expected the distance to be constant between a given set of real and synthesised image?