Is this compatible with the new 4.1 svc models trained with vec768?

[quyaoning]

I tried inference with a model file trained with sovits 4.1 in encoder vec768112 and got the following error:

           ERROR    [09:27:21] Error in realtime:                                                                                                  gui.py:838
           ERROR    [09:27:21] Given groups=1, weight of size [192, 768, 5], expected input[1, 256, 103] to have 768 channels, but got 256 channels instead gui.py:839