Add stargan model (#119)
* Initial commit of pytorch_stargan
* added benchmark for stargan
* Reduced training time for stargan 1/10x
* Training iternum=1
* remove tf from deps
* fix parser args
* gpu training for stargan on PR"
Co-authored-by: Ed Tyantov <tyantov@devvm2029.lla0.facebook.com>
