Hi Mattias,

Thanks for your feedback, there is known issues with this implementation, I’m currently improving them, and train them with GPU (Tensorflow and PyTorch).

Have a nice day,