Method forward

src/diffusion/unet_triplane.py:257–267

Apply the block to a Tensor, conditioned on a timestep embedding. :param x: an [N x C x ...] Tensor of features. :param emb: an [N x emb_channels] Tensor of timestep embeddings. :return: an [N x C x ...] Tensor of outputs.

(self, x, emb)

            self.skip_connection = TriplaneConv(channels, self.out_channels, 1, padding=0, is_rollout=False)

    def forward(self, x, emb):
        """
        Apply the block to a Tensor, conditioned on a timestep embedding.

        :param x: an [N x C x ...] Tensor of features.
        :param emb: an [N x emb_channels] Tensor of timestep embeddings.
        :return: an [N x C x ...] Tensor of outputs.
        """
        return checkpoint(
            self._forward, (x, emb), self.parameters(), self.use_checkpoint
        )

    def _forward(self, x, emb):
        # x: (h_xy, h_xz, h_yz)