Class AttnDownEncoderBlock2D

src/diffusers/models/unets/unet_2d_blocks.py:1451–1551

class AttnDownEncoderBlock2D(nn.Module):
    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        dropout: float = 0.0,
        num_layers: int = 1,
        resnet_eps: float = 1e-6,
        resnet_time_scale_shift: str = "default",
        resnet_act_fn: str = "swish",
        resnet_groups: int = 32,
        resnet_pre_norm: bool = True,
        attention_head_dim: int = 1,
        output_scale_factor: float = 1.0,
        add_downsample: bool = True,
        downsample_padding: int = 1,
    ):
        super().__init__()
        resnets = []
        attentions = []

        if attention_head_dim is None:
            logger.warning(
                f"It is not recommend to pass `attention_head_dim=None`. Defaulting `attention_head_dim` to `in_channels`: {out_channels}."
            )
            attention_head_dim = out_channels

        for i in range(num_layers):
            in_channels = in_channels if i == 0 else out_channels
            if resnet_time_scale_shift == "spatial":
                resnets.append(
                    ResnetBlockCondNorm2D(
                        in_channels=in_channels,
                        out_channels=out_channels,
                        temb_channels=None,
                        eps=resnet_eps,
                        groups=resnet_groups,
                        dropout=dropout,
                        time_embedding_norm="spatial",
                        non_linearity=resnet_act_fn,
                        output_scale_factor=output_scale_factor,
                    )
                )
            else:
                resnets.append(
                    ResnetBlock2D(
                        in_channels=in_channels,
                        out_channels=out_channels,
                        temb_channels=None,
                        eps=resnet_eps,
                        groups=resnet_groups,
                        dropout=dropout,
                        time_embedding_norm=resnet_time_scale_shift,
                        non_linearity=resnet_act_fn,
                        output_scale_factor=output_scale_factor,
                        pre_norm=resnet_pre_norm,
                    )
                )