Question about the pretrain model

[satyremusique]

Thank you for this great work. Just a simple question, is the pretrain model that you share was train with exactly the code in the github. And is is the same(same hyperparameter) than the article.

I am able to use that pretrain model to do inference, but I am not able to use it to fine-tune with different data. It's seem that the miss match of tensor in the discriminator(when given the output of the generator). But if I retrain from scratch and save result, I seem to be able to fine tune it after that.