def forward()

in models/src/wavenet_vocoder/wavenet.py [0:0]

• 22 lines of code
• 8 McCabe index (conditional complexity)

    def forward(self, x, c=None, g=None, softmax=False):
        """Forward step

        Args:
            x (Tensor): One-hot encoded audio signal, shape (B x C x T)
            c (Tensor): Local conditioning features,
              shape (B x cin_channels x T)
            g (Tensor): Global conditioning features,
              shape (B x gin_channels x 1) or speaker Ids of shape (B x 1).
              Note that ``self.use_speaker_embedding`` must be False when you
              want to disable embedding layer and use external features
              directly (e.g., one-hot vector).
              Also type of input tensor must be FloatTensor, not LongTensor
              in case of ``self.use_speaker_embedding`` equals False.
            softmax (bool): Whether applies softmax or not.

        Returns:
            Tensor: output, shape B x out_channels x T
        """
        B, _, T = x.size()

        if g is not None:
            if self.embed_speakers is not None:
                # (B x 1) -> (B x 1 x gin_channels)
                g = self.embed_speakers(g.view(B, -1))
                # (B x gin_channels x 1)
                g = g.transpose(1, 2)
                assert g.dim() == 3
        # Expand global conditioning features to all time steps
        g_bct = _expand_global_features(B, T, g, bct=True)

        if c is not None and self.upsample_net is not None:
            c = self.upsample_net(c)
            assert c.size(-1) == x.size(-1)

        # Feed data to network
        x = self.first_conv(x)
        skips = 0
        for f in self.conv_layers:
            x, h = f(x, c, g_bct)
            skips += h
        skips *= math.sqrt(1.0 / len(self.conv_layers))

        x = skips
        for f in self.last_conv_layers:
            x = f(x)

        x = F.softmax(x, dim=1) if softmax else x

        return x