training problem about multilingual_text_to_speech

[michael-conrad]

Add the following to the top of the train python script

import matplotlib
matplotlib.use("Qt5Agg")

[Tomiinek]

Hello, thank you for reporting the error.

I cannot reproduce the issue, could you try to replace this line with the following code:

embedded = layer(encoded if condition is None else condition.long())

Does it help?

[kongziyue1234]

Hello，thank you for your suggestion.
I tried your suggestion, but the mistake still came up.Later, I conducted the experiment in Ubuntu system, and such an error occurred after a warning:audio amplitude out of range, auto clipped. appeared

[Tomiinek]

Hello, I have never seen the error. I am also getting the warnings, but they do not cause any errors.

After some googling, it seems to me like an issue with your X server or drivers.
It is probably not related to the code.

[kongziyue1234]

Hello,I switched the experiment to different gpus and the problem was solved.However, in the half of the training, the following mistakes occurred.I have updated the tkinter package and reduced the batch_size,but they didn't work. Have you ever encountered them?