maximlf

Hi, Thanks for the implementations, I just discovered this repo., very interesting. I tried running the train.py with default settings, The results shown is bit far from what claimed on...