Dlats stochastic clipping range

[oneiroid]

Hey! I've checked distribution of dlatent's components (W space) - they seem to be not in [-2,2] range, and not normal at all contrary to what's said in the readme... It's more like gamma distribution in range [-0.25, 1.5]. I hardcoded the latter in you repo - quality seems to improve.

[pbaylies]

Nice! Patches welcome, of course. Clipping can take a number already, it's set to 2 (i.e., [-2,2] because that is far enough out of the usual range, but it wouldn't be clipping values as quickly as you would be... Also, idk if the distribution is always like that for other StyleGAN models.