so-vits-svc/vencoder/encoder.py

class SpeechEncoder(object):
    def __init__(self, vec_path="pretrain/checkpoint_best_legacy_500.pt", device=None):
        self.model = None  # This is Model
        self.hidden_dim = 768
        pass


    def encoder(self, wav):
        """
        input: wav:[signal_length]
        output: embedding:[batchsize,hidden_dim,wav_frame]
        """
        pass