Is the gradient penalty loss problematic when input image is large?

[shaform]

Hi,
Here the sum of squares is computed
slopes = tf.sqrt(tf.reduce_sum(tf.square(gradients), reduction_indices=[1]))

However, when the input image is extremely large, the dimension of gradients would be huge. It seems possible that the resulting slopes would be extremely large compared to 1.
Is this the case?

[kodalinaveen3]

We take gradients of D(x) in the x-space. Because we only perturb samples slightly, the size of input image shouldn't matter that much