fixing torch device error when synthesizer is called with use_cuda=True

[TRY-ER]

whenever I declare synthesizer with use_cuda = True. there will be an error that different tensors are in different devices.
After observation it was found that input tensor was on "cpu" while weight is on the gpu.
Hence I modified the code to transfer the input to gpu.
It works fine with my code.

Hope this helps !

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discussion channels.