How to include your trained model (forward tacotron) in your project (code), like sova-tts do?

Question

Python

I trained the model in the ForwardTacotron repository. I tried to save via torch. But I can’t figure out how to embed it in my project so that it gives me a chalk-spectrogram and connects griffin-lim to get audio?
I got confused about the organization of the sova-tts project, but I would like to do the same thing, only without the flag. Just like a module that gives audio.
PS Trained the model in the colab.
I'm trying to write the module in pycharm on windows 10. It doesn't write cuda, and the hardware is old (but I'm not going to teach), while sova-tts works.
How do I include my model and get speech synthesis from it in my project without cuda?