Why do I open basic_cnn_model.py under the SAE folder and make is_training = True,loss=0,but can attack successfuly?

[Kitzzaaa]

[Kitzzaaa]

And in the picture,why x_input = x_input *2.0 -1.0?