models/diffusion/video_model.py

# Adapted from https://github.com/Stability-AI/generative-models/blob/main/sgm/modules/diffusionmodules/video_model.py
from functools import partial
from typing import List, Optional, Union

from einops import rearrange

from models.svd.sgm.modules.diffusionmodules.openaimodel import *
from models.svd.sgm.modules.video_attention import SpatialVideoTransformer
from models.svd.sgm.util import default
from models.svd.sgm.modules.diffusionmodules.util import AlphaBlender
from functools import partial
from models.cam.conditioning import ConditionalModel


class VideoResBlock(ResBlock):
    def __init__(
        self,
        channels: int,
        emb_channels: int,
        dropout: float,
        video_kernel_size: Union[int, List[int]] = 3,
        merge_strategy: str = "fixed",
        merge_factor: float = 0.5,
        out_channels: Optional[int] = None,
        use_conv: bool = False,
        use_scale_shift_norm: bool = False,
        dims: int = 2,
        use_checkpoint: bool = False,
        up: bool = False,
        down: bool = False,
    ):
        super().__init__(
            channels,
            emb_channels,
            dropout,
            out_channels=out_channels,
            use_conv=use_conv,
            use_scale_shift_norm=use_scale_shift_norm,
            dims=dims,
            use_checkpoint=use_checkpoint,
            up=up,
            down=down,
        )

        self.time_stack = ResBlock(
            default(out_channels, channels),
            emb_channels,
            dropout=dropout,
            dims=3,
            out_channels=default(out_channels, channels),
            use_scale_shift_norm=False,
            use_conv=False,
            up=False,
            down=False,
            kernel_size=video_kernel_size,
            use_checkpoint=use_checkpoint,
            exchange_temb_dims=True,
        )
        self.time_mixer = AlphaBlender(
            alpha=merge_factor,
            merge_strategy=merge_strategy,
            rearrange_pattern="b t -> b 1 t 1 1",
        )

    def forward(
        self,
        x: th.Tensor,
        emb: th.Tensor,
        num_video_frames: int,
        image_only_indicator: Optional[th.Tensor] = None,
    ) -> th.Tensor:
        x = super().forward(x, emb)

        x_mix = rearrange(x, "(b t) c h w -> b c t h w", t=num_video_frames)
        x = rearrange(x, "(b t) c h w -> b c t h w", t=num_video_frames)

        x = self.time_stack(
            x, rearrange(emb, "(b t) ... -> b t ...", t=num_video_frames)
        )
        x = self.time_mixer(
            x_spatial=x_mix, x_temporal=x, image_only_indicator=image_only_indicator
        )
        x = rearrange(x, "b c t h w -> (b t) c h w")
        return x


class VideoUNet(nn.Module):
    '''
    Adapted from the vanilla SVD model. We add "cross_attention_merger_input_blocks" and "cross_attention_merger_mid_block" to incorporate the CAM control features. 

    '''

    def __init__(
        self,
        in_channels: int,
        model_channels: int,
        out_channels: int,
        num_res_blocks: int,
        num_conditional_frames: int,
        attention_resolutions: Union[List[int], int],
        dropout: float = 0.0,
        channel_mult: List[int] = (1, 2, 4, 8),
        conv_resample: bool = True,
        dims: int = 2,
        num_classes: Optional[Union[int, str]] = None,
        use_checkpoint: bool = False,
        num_heads: int = -1,
        num_head_channels: int = -1,
        num_heads_upsample: int = -1,
        use_scale_shift_norm: bool = False,
        resblock_updown: bool = False,
        transformer_depth: Union[List[int], int] = 1,
        transformer_depth_middle: Optional[int] = None,
        context_dim: Optional[int] = None,
        time_downup: bool = False,
        time_context_dim: Optional[int] = None,
        extra_ff_mix_layer: bool = False,
        use_spatial_context: bool = False,
        merge_strategy: str = "fixed",
        merge_factor: float = 0.5,
        spatial_transformer_attn_type: str = "softmax",
        video_kernel_size: Union[int, List[int]] = 3,
        use_linear_in_transformer: bool = False,
        adm_in_channels: Optional[int] = None,
        disable_temporal_crossattention: bool = False,
        max_ddpm_temb_period: int = 10000,
        merging_mode: str = "addition",
        controlnet_mode: bool = False,
        use_apm: bool = False,
    ):
        super().__init__()
        assert context_dim is not None
        self.controlnet_mode = controlnet_mode
        if controlnet_mode:
            assert merging_mode.startswith(
                "attention"), "other merging modes not implemented"
            AttentionCondModel = partial(
                ConditionalModel, conditional_model=merging_mode.split("attention_")[1])
            self.cross_attention_merger_input_blocks = nn.ModuleList([])
        if num_heads_upsample == -1:
            num_heads_upsample = num_heads

        if num_heads == -1:
            assert num_head_channels != -1

        if num_head_channels == -1:
            assert num_heads != -1

        self.in_channels = in_channels
        self.model_channels = model_channels
        self.out_channels = out_channels
        if isinstance(transformer_depth, int):
            transformer_depth = len(channel_mult) * [transformer_depth]
        transformer_depth_middle = default(
            transformer_depth_middle, transformer_depth[-1]
        )

        self.num_res_blocks = num_res_blocks
        self.attention_resolutions = attention_resolutions
        self.dropout = dropout
        self.channel_mult = channel_mult
        self.conv_resample = conv_resample
        self.num_classes = num_classes
        self.use_checkpoint = use_checkpoint
        self.num_heads = num_heads
        self.num_head_channels = num_head_channels
        self.num_heads_upsample = num_heads_upsample
        self.dims = dims
        self.use_scale_shift_norm = use_scale_shift_norm
        self.resblock_updown = resblock_updown
        self.transformer_depth = transformer_depth
        self.transformer_depth_middle = transformer_depth_middle
        self.context_dim = context_dim
        self.time_downup = time_downup
        self.time_context_dim = time_context_dim
        self.extra_ff_mix_layer = extra_ff_mix_layer
        self.use_spatial_context = use_spatial_context
        self.merge_strategy = merge_strategy
        self.merge_factor = merge_factor
        self.spatial_transformer_attn_type = spatial_transformer_attn_type
        self.video_kernel_size = video_kernel_size
        self.use_linear_in_transformer = use_linear_in_transformer
        self.adm_in_channels = adm_in_channels
        self.disable_temporal_crossattention = disable_temporal_crossattention
        self.max_ddpm_temb_period = max_ddpm_temb_period

        time_embed_dim = model_channels * 4
        self.time_embed = nn.Sequential(
            linear(model_channels, time_embed_dim),
            nn.SiLU(),
            linear(time_embed_dim, time_embed_dim),
        )

        if self.num_classes is not None:
            if isinstance(self.num_classes, int):
                self.label_emb = nn.Embedding(num_classes, time_embed_dim)
            elif self.num_classes == "continuous":
                print("setting up linear c_adm embedding layer")
                self.label_emb = nn.Linear(1, time_embed_dim)
            elif self.num_classes == "timestep":
                self.label_emb = nn.Sequential(
                    Timestep(model_channels),
                    nn.Sequential(
                        linear(model_channels, time_embed_dim),
                        nn.SiLU(),
                        linear(time_embed_dim, time_embed_dim),
                    ),
                )

            elif self.num_classes == "sequential":
                assert adm_in_channels is not None
                self.label_emb = nn.Sequential(
                    nn.Sequential(
                        linear(adm_in_channels, time_embed_dim),
                        nn.SiLU(),
                        linear(time_embed_dim, time_embed_dim),
                    )
                )
            else:
                raise ValueError()

        self.input_blocks = nn.ModuleList(
            [
                TimestepEmbedSequential(
                    conv_nd(dims, in_channels, model_channels, 3, padding=1)
                )
            ]
        )
        self._feature_size = model_channels
        input_block_chans = [model_channels]
        ch = model_channels
        ds = 1
        if controlnet_mode and merging_mode.startswith("attention"):
            self.cross_attention_merger_input_blocks.append(
                AttentionCondModel(input_channels=ch))

        def get_attention_layer(
            ch,
            num_heads,
            dim_head,
            depth=1,
            context_dim=None,
            use_checkpoint=False,
            disabled_sa=False,
            use_apm: bool = False,
        ):
            return SpatialVideoTransformer(
                ch,
                num_heads,
                dim_head,
                depth=depth,
                context_dim=context_dim,
                time_context_dim=time_context_dim,
                dropout=dropout,
                ff_in=extra_ff_mix_layer,
                use_spatial_context=use_spatial_context,
                merge_strategy=merge_strategy,
                merge_factor=merge_factor,
                checkpoint=use_checkpoint,
                use_linear=use_linear_in_transformer,
                attn_mode=spatial_transformer_attn_type,
                disable_self_attn=disabled_sa,
                disable_temporal_crossattention=disable_temporal_crossattention,
                max_time_embed_period=max_ddpm_temb_period,
                use_apm=use_apm,
            )

        def get_resblock(
            merge_factor,
            merge_strategy,
            video_kernel_size,
            ch,
            time_embed_dim,
            dropout,
            out_ch,
            dims,
            use_checkpoint,
            use_scale_shift_norm,
            down=False,
            up=False,
        ):
            return VideoResBlock(
                merge_factor=merge_factor,
                merge_strategy=merge_strategy,
                video_kernel_size=video_kernel_size,
                channels=ch,
                emb_channels=time_embed_dim,
                dropout=dropout,
                out_channels=out_ch,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
                down=down,
                up=up,
            )

        for level, mult in enumerate(channel_mult):
            for _ in range(num_res_blocks):
                layers = [
                    get_resblock(
                        merge_factor=merge_factor,
                        merge_strategy=merge_strategy,
                        video_kernel_size=video_kernel_size,
                        ch=ch,
                        time_embed_dim=time_embed_dim,
                        dropout=dropout,
                        out_ch=mult * model_channels,
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                    )
                ]
                ch = mult * model_channels
                if ds in attention_resolutions:
                    if num_head_channels == -1:
                        dim_head = ch // num_heads
                    else:
                        num_heads = ch // num_head_channels
                        dim_head = num_head_channels

                    layers.append(
                        get_attention_layer(
                            ch,
                            num_heads,
                            dim_head,
                            depth=transformer_depth[level],
                            context_dim=context_dim,
                            use_checkpoint=use_checkpoint,
                            disabled_sa=False,
                            use_apm=use_apm,
                        )
                    )
                self.input_blocks.append(TimestepEmbedSequential(*layers))
                if controlnet_mode and merging_mode.startswith("attention"):
                    self.cross_attention_merger_input_blocks.append(
                        AttentionCondModel(input_channels=ch))
                self._feature_size += ch
                input_block_chans.append(ch)
            if level != len(channel_mult) - 1:
                ds *= 2
                out_ch = ch
                self.input_blocks.append(
                    TimestepEmbedSequential(
                        get_resblock(
                            merge_factor=merge_factor,
                            merge_strategy=merge_strategy,
                            video_kernel_size=video_kernel_size,
                            ch=ch,
                            time_embed_dim=time_embed_dim,
                            dropout=dropout,
                            out_ch=out_ch,
                            dims=dims,
                            use_checkpoint=use_checkpoint,
                            use_scale_shift_norm=use_scale_shift_norm,
                            down=True,
                        )
                        if resblock_updown
                        else Downsample(
                            ch,
                            conv_resample,
                            dims=dims,
                            out_channels=out_ch,
                            third_down=time_downup,
                        )
                    )
                )
                ch = out_ch
                input_block_chans.append(ch)

                if controlnet_mode and merging_mode.startswith("attention"):
                    self.cross_attention_merger_input_blocks.append(
                        AttentionCondModel(input_channels=ch))
                self._feature_size += ch

        if num_head_channels == -1:
            dim_head = ch // num_heads
        else:
            num_heads = ch // num_head_channels
            dim_head = num_head_channels

        self.middle_block = TimestepEmbedSequential(
            get_resblock(
                merge_factor=merge_factor,
                merge_strategy=merge_strategy,
                video_kernel_size=video_kernel_size,
                ch=ch,
                time_embed_dim=time_embed_dim,
                out_ch=None,
                dropout=dropout,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
            ),
            get_attention_layer(
                ch,
                num_heads,
                dim_head,
                depth=transformer_depth_middle,
                context_dim=context_dim,
                use_checkpoint=use_checkpoint,
                use_apm=use_apm,
            ),
            get_resblock(
                merge_factor=merge_factor,
                merge_strategy=merge_strategy,
                video_kernel_size=video_kernel_size,
                ch=ch,
                out_ch=None,
                time_embed_dim=time_embed_dim,
                dropout=dropout,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
            ),
        )
        self._feature_size += ch
        if controlnet_mode and merging_mode.startswith("attention"):
            self.cross_attention_merger_mid_block = AttentionCondModel(
                input_channels=ch)

        self.output_blocks = nn.ModuleList([])
        for level, mult in list(enumerate(channel_mult))[::-1]:
            for i in range(num_res_blocks + 1):
                ich = input_block_chans.pop()
                layers = [
                    get_resblock(
                        merge_factor=merge_factor,
                        merge_strategy=merge_strategy,
                        video_kernel_size=video_kernel_size,
                        ch=ch + ich,
                        time_embed_dim=time_embed_dim,
                        dropout=dropout,
                        out_ch=model_channels * mult,
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                    )
                ]
                ch = model_channels * mult
                if ds in attention_resolutions:
                    if num_head_channels == -1:
                        dim_head = ch // num_heads
                    else:
                        num_heads = ch // num_head_channels
                        dim_head = num_head_channels

                    layers.append(
                        get_attention_layer(
                            ch,
                            num_heads,
                            dim_head,
                            depth=transformer_depth[level],
                            context_dim=context_dim,
                            use_checkpoint=use_checkpoint,
                            disabled_sa=False,
                            use_apm=use_apm,
                        )
                    )
                if level and i == num_res_blocks:
                    out_ch = ch
                    ds //= 2
                    layers.append(
                        get_resblock(
                            merge_factor=merge_factor,
                            merge_strategy=merge_strategy,
                            video_kernel_size=video_kernel_size,
                            ch=ch,
                            time_embed_dim=time_embed_dim,
                            dropout=dropout,
                            out_ch=out_ch,
                            dims=dims,
                            use_checkpoint=use_checkpoint,
                            use_scale_shift_norm=use_scale_shift_norm,
                            up=True,
                        )
                        if resblock_updown
                        else Upsample(
                            ch,
                            conv_resample,
                            dims=dims,
                            out_channels=out_ch,
                            third_up=time_downup,
                        )
                    )

                self.output_blocks.append(TimestepEmbedSequential(*layers))
                self._feature_size += ch

        self.out = nn.Sequential(
            normalization(ch),
            nn.SiLU(),
            zero_module(conv_nd(dims, model_channels,
                        out_channels, 3, padding=1)),
        )

    def forward(
        self,
        # [28,8,72,128], i.e. (B F) (2 C) H W = concat([z_t,<cond_frames>])
        x: th.Tensor,
        timesteps: th.Tensor,  # [28], i.e. (B F)
        # [28, 1, 1024], i.e. (B F) 1 T, for cross attention from clip image encoder, <cond_frames_without_noise>
        context: Optional[th.Tensor] = None,
        # [28, 768], i.e. (B F) T ? concat([<fps_id>,<motion_bucket_id>,<cond_aug>]
        y: Optional[th.Tensor] = None,
        time_context: Optional[th.Tensor] = None,  # NONE
        num_video_frames: Optional[int] = None,  # 14
        num_conditional_frames: Optional[int] = None,  # 8
        # zeros, [2,14], i.e. [B, F]
        image_only_indicator: Optional[th.Tensor] = None,
        hs_control_input: Optional[th.Tensor] = None,  # cam features
        hs_control_mid: Optional[th.Tensor] = None,  # cam features
    ):
        assert (y is not None) == (
            self.num_classes is not None
        ), "must specify y if and only if the model is class-conditional -> no, relax this TODO"
        hs = []
        t_emb = timestep_embedding(
            timesteps, self.model_channels, repeat_only=False).to(x.dtype)
        emb = self.time_embed(t_emb)

        if self.num_classes is not None:
            assert y.shape[0] == x.shape[0]
            emb = emb + self.label_emb(y)

        h = x
        for module in self.input_blocks:
            h = module(
                h,
                emb,
                context=context,
                image_only_indicator=image_only_indicator,
                time_context=time_context,
                num_video_frames=num_video_frames,
            )
            hs.append(h)

        # fusion of cam features with base features
        if hs_control_input is not None:
            new_hs = []

            assert len(hs) == len(hs_control_input) and len(
                hs) == len(self.cross_attention_merger_input_blocks)
            for h_no_ctrl, h_ctrl, merger in zip(hs, hs_control_input, self.cross_attention_merger_input_blocks):
                merged_h = merger(h_no_ctrl, h_ctrl, num_frames=num_video_frames,
                                  num_conditional_frames=num_conditional_frames)
                new_hs.append(merged_h)
            hs = new_hs

        h = self.middle_block(
            h,
            emb,
            context=context,
            image_only_indicator=image_only_indicator,
            time_context=time_context,
            num_video_frames=num_video_frames,
        )

        # fusion of cam features with base features
        if hs_control_mid is not None:
            h = self.cross_attention_merger_mid_block(
                h, hs_control_mid, num_frames=num_video_frames, num_conditional_frames=num_conditional_frames)

        for module in self.output_blocks:
            h = th.cat([h, hs.pop()], dim=1)
            h = module(
                h,
                emb,
                context=context,
                image_only_indicator=image_only_indicator,
                time_context=time_context,
                num_video_frames=num_video_frames,
            )
        h = h.type(x.dtype)
        return self.out(h)