Class UNetMidBlock1D

src/diffusers/models/unets/unet_1d_blocks.py:408–445

class UNetMidBlock1D(nn.Module):
    def __init__(self, mid_channels: int, in_channels: int, out_channels: int | None = None):
        super().__init__()

        out_channels = in_channels if out_channels is None else out_channels

        # there is always at least one resnet
        self.down = Downsample1d("cubic")
        resnets = [
            ResConvBlock(in_channels, mid_channels, mid_channels),
            ResConvBlock(mid_channels, mid_channels, mid_channels),
            ResConvBlock(mid_channels, mid_channels, mid_channels),
            ResConvBlock(mid_channels, mid_channels, mid_channels),
            ResConvBlock(mid_channels, mid_channels, mid_channels),
            ResConvBlock(mid_channels, mid_channels, out_channels),
        ]
        attentions = [
            SelfAttention1d(mid_channels, mid_channels // 32),
            SelfAttention1d(mid_channels, mid_channels // 32),
            SelfAttention1d(mid_channels, mid_channels // 32),
            SelfAttention1d(mid_channels, mid_channels // 32),
            SelfAttention1d(mid_channels, mid_channels // 32),
            SelfAttention1d(out_channels, out_channels // 32),
        ]
        self.up = Upsample1d(kernel="cubic")

        self.attentions = nn.ModuleList(attentions)
        self.resnets = nn.ModuleList(resnets)

    def forward(self, hidden_states: torch.Tensor, temb: torch.Tensor | None = None) -> torch.Tensor:
        hidden_states = self.down(hidden_states)
        for attn, resnet in zip(self.attentions, self.resnets):
            hidden_states = resnet(hidden_states)
            hidden_states = attn(hidden_states)

        hidden_states = self.up(hidden_states)

        return hidden_states


class AttnDownBlock1D(nn.Module):