b71f31eae4f74b6a426861f8085d73f2eb1c063a,models/tacotron2.py,Tacotron2,forward,#Tacotron2#Any#Any#Any#Any#Any#,78

Before Change

            self.compute_speaker_embedding(speaker_ids)
        if self.num_speakers > 1:
            // B x T_in x embed_dim + speaker_embed_dim
            encoder_outputs = self._add_speaker_embedding(encoder_outputs,
                                                          self.speaker_embeddings)
        encoder_outputs = encoder_outputs * input_mask.unsqueeze(2).expand_as(encoder_outputs)
        // global style token
        if self.gst:
            // B x gst_dim

After Change

            else:
                encoder_outputs = torch.cat([encoder_outputs, embedded_speakers], dim=-1)
        else:
            if hasattr(self, "gst"):
                // B x gst_dim
                encoder_outputs, embedded_gst = self.compute_gst(encoder_outputs, mel_specs)
                encoder_outputs = torch.cat([encoder_outputs, embedded_gst], dim=-1)

        encoder_outputs = encoder_outputs * input_mask.unsqueeze(2).expand_as(encoder_outputs)

        // B x mel_dim x T_out -- B x T_out//r x T_in -- B x T_out//r
        decoder_outputs, alignments, stop_tokens = self.decoder(

Instances