Code and vocoder issues during synthesis

[CriDora]

Hello, thank you for your open source! Is the if branch in the "assert_required_models_available" function in matha/cli.py written in reverse? When I specify checkpoint_path, it keeps downloading the matcha_ljspeech model, so I think there is something wrong with the code. In addition, how can I not use vocoder when using the multi-speaker model I trained myself? If I don't specify --vocoder, it will obviously report an error during synthesis.

[CriDora]

There is no problem with vocoder, I misunderstood