Method __init__

src/diffusers/models/unets/unet_2d_blocks.py:1998–2041

(
        self,
        in_channels: int,
        out_channels: int,
        temb_channels: int,
        dropout: float = 0.0,
        num_layers: int = 4,
        resnet_eps: float = 1e-5,
        resnet_act_fn: str = "gelu",
        resnet_group_size: int = 32,
        add_downsample: bool = False,
    )

class KDownBlock2D(nn.Module):
    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        temb_channels: int,
        dropout: float = 0.0,
        num_layers: int = 4,
        resnet_eps: float = 1e-5,
        resnet_act_fn: str = "gelu",
        resnet_group_size: int = 32,
        add_downsample: bool = False,
    ):
        super().__init__()
        resnets = []

        for i in range(num_layers):
            in_channels = in_channels if i == 0 else out_channels
            groups = in_channels // resnet_group_size
            groups_out = out_channels // resnet_group_size

            resnets.append(
                ResnetBlockCondNorm2D(
                    in_channels=in_channels,
                    out_channels=out_channels,
                    dropout=dropout,
                    temb_channels=temb_channels,
                    groups=groups,
                    groups_out=groups_out,
                    eps=resnet_eps,
                    non_linearity=resnet_act_fn,
                    time_embedding_norm="ada_group",
                    conv_shortcut_bias=False,
                )
            )

        self.resnets = nn.ModuleList(resnets)

        if add_downsample:
            # YiYi's comments- might be able to use FirDownsample2D, look into details later
            self.downsamplers = nn.ModuleList([KDownsample2D()])
        else:
            self.downsamplers = None

        self.gradient_checkpointing = False

    def forward(
        self, hidden_states: torch.Tensor, temb: torch.Tensor | None = None, *args, **kwargs