Arghyadeep Giri

I think if we can produce c++ equivalent code for inference and convert the model using torch.jit.trace it could be exported to both android and ios.

Try lowering learning rate.

https://montreal-forced-aligner.readthedocs.io/en/latest/

You are facing this issue since there are two states saved in this checkpoint. The model and the optimizer. To fine tune you need to load only the model component...