TTS Model not providing clear speech

I have trained my Text-to-Speech model on tacotron few months back for speech synthesis. At that time it was generating a clear audio with proper text recognition but now after sometime I again tested the results for the same trained model. However, this time it is not providing clear output. All it is inferencing is noise only. Please help me figure out the issue.
The model file is loading correctly, so there could be no chance of it getting corrupt and if that’s so the would could be the reason as my model is saved in s3 bucket.

@ptrblck please acknowledge

I don’t know what might be causing it, as you are excluding some issues I would have recommended to check, i.e. the model’s state_dict etc.
I’m also unsure if anything else was changed between these runs (such as the PyTorch release) so would generally recommend to use the same setup as before.