14a4d1a061872622d992ec39369bd24a78e292cc,models/tacotrongst.py,TacotronGST,__init__,#TacotronGST#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#,9

Before Change

        if num_speakers > 1:
            self.speaker_embedding = nn.Embedding(num_speakers, 256)
            self.speaker_embedding.weight.data.normal_(0, 0.3)
        self.encoder = Encoder(256)
        self.gst = GST(num_mel=80, num_heads=4, num_style_tokens=10, embedding_dim=256)
        self.decoder = Decoder(256, mel_dim, r, memory_size, attn_win,
                               attn_norm, prenet_type, prenet_dropout,
                               forward_attn, trans_agent, forward_attn_mask,

After Change

                        separate_stopnet)
        gst_embedding_dim = 256
        decoder_dim = 512 + gst_embedding_dim if num_speakers > 1 else 256 + gst_embedding_dim
        proj_speaker_dim = 80 if num_speakers > 1 else 0
        self.decoder = Decoder(decoder_dim, mel_dim, r, memory_size, attn_win,
                               attn_norm, prenet_type, prenet_dropout,
                               forward_attn, trans_agent, forward_attn_mask,
                               location_attn, separate_stopnet, proj_speaker_dim)

Instances