I have had mixed luck with this model, which is supposedly state-of-the-art: https://github.com/magenta/mt3

What kind of music are you trying to transcribe?

Feel free to email me.