ComfyUI/comfy/ldm/wan/model.py

# original version: https://github.com/Wan-Video/Wan2.1/blob/main/wan/modules/model.py
# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved.
import math

import torch
import torch.nn as nn

from comfy.ldm.modules.attention import optimized_attention


def sinusoidal_embedding_1d(dim, position):
    # preprocess
    assert dim % 2 == 0
    half = dim // 2
    position = position.type(torch.float64)

    # calculation
    sinusoid = torch.outer(
        position, torch.pow(10000, -torch.arange(half).to(position).div(half)))
    x = torch.cat([torch.cos(sinusoid), torch.sin(sinusoid)], dim=1)
    return x


def rope_params(max_seq_len, dim, theta=10000):
    assert dim % 2 == 0
    freqs = torch.outer(
        torch.arange(max_seq_len),
        1.0 / torch.pow(theta,
                        torch.arange(0, dim, 2).to(torch.float64).div(dim)))
    freqs = torch.polar(torch.ones_like(freqs), freqs)
    return freqs


def rope_apply(x, grid_sizes, freqs):
    n, c = x.size(2), x.size(3) // 2

    # split freqs
    freqs = freqs.split([c - 2 * (c // 3), c // 3, c // 3], dim=1)

    # loop over samples
    output = []
    for i, (f, h, w) in enumerate(grid_sizes.tolist()):
        seq_len = f * h * w

        # precompute multipliers
        x_i = torch.view_as_complex(x[i, :seq_len].to(torch.float64).reshape(
            seq_len, n, -1, 2))
        freqs_i = torch.cat([
            freqs[0][:f].view(f, 1, 1, -1).expand(f, h, w, -1),
            freqs[1][:h].view(1, h, 1, -1).expand(f, h, w, -1),
            freqs[2][:w].view(1, 1, w, -1).expand(f, h, w, -1)
        ], dim=-1).reshape(seq_len, 1, -1)

        # apply rotary embedding
        x_i = torch.view_as_real(x_i * freqs_i).flatten(2)
        x_i = torch.cat([x_i, x[i, seq_len:]])

        # append to collection
        output.append(x_i)
    return torch.stack(output).to(dtype=x.dtype)


class WanRMSNorm(nn.Module):

    def __init__(self, dim, eps=1e-5, device=None, dtype=None):
        super().__init__()
        self.dim = dim
        self.eps = eps
        self.weight = nn.Parameter(torch.ones(dim, device=device, dtype=dtype))

    def forward(self, x):
        r"""
        Args:
            x(Tensor): Shape [B, L, C]
        """
        return self._norm(x.float()).type_as(x) * self.weight

    def _norm(self, x):
        return x * torch.rsqrt(x.pow(2).mean(dim=-1, keepdim=True) + self.eps)


class WanSelfAttention(nn.Module):

    def __init__(self,
                 dim,
                 num_heads,
                 window_size=(-1, -1),
                 qk_norm=True,
                 eps=1e-6, operation_settings={}):
        assert dim % num_heads == 0
        super().__init__()
        self.dim = dim
        self.num_heads = num_heads
        self.head_dim = dim // num_heads
        self.window_size = window_size
        self.qk_norm = qk_norm
        self.eps = eps

        # layers
        self.q = operation_settings.get("operations").Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.k = operation_settings.get("operations").Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.v = operation_settings.get("operations").Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.o = operation_settings.get("operations").Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.norm_q = WanRMSNorm(dim, eps=eps, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")) if qk_norm else nn.Identity()
        self.norm_k = WanRMSNorm(dim, eps=eps, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")) if qk_norm else nn.Identity()

    def forward(self, x, seq_lens, grid_sizes, freqs):
        r"""
        Args:
            x(Tensor): Shape [B, L, num_heads, C / num_heads]
            seq_lens(Tensor): Shape [B]
            grid_sizes(Tensor): Shape [B, 3], the second dimension contains (F, H, W)
            freqs(Tensor): Rope freqs, shape [1024, C / num_heads / 2]
        """
        b, s, n, d = *x.shape[:2], self.num_heads, self.head_dim

        # query, key, value function
        def qkv_fn(x):
            q = self.norm_q(self.q(x)).view(b, s, n, d)
            k = self.norm_k(self.k(x)).view(b, s, n, d)
            v = self.v(x).view(b, s, n * d)
            return q, k, v

        q, k, v = qkv_fn(x)

        x = optimized_attention(
            q=rope_apply(q, grid_sizes, freqs).view(b, s, n * d),
            k=rope_apply(k, grid_sizes, freqs).view(b, s, n * d),
            v=v,
            heads=self.num_heads,
        )

        x = self.o(x)
        return x


class WanT2VCrossAttention(WanSelfAttention):

    def forward(self, x, context, context_lens):
        r"""
        Args:
            x(Tensor): Shape [B, L1, C]
            context(Tensor): Shape [B, L2, C]
            context_lens(Tensor): Shape [B]
        """
        # compute query, key, value
        q = self.norm_q(self.q(x))
        k = self.norm_k(self.k(context))
        v = self.v(context)

        # compute attention
        x = optimized_attention(q, k, v, heads=self.num_heads)

        x = self.o(x)
        return x


class WanI2VCrossAttention(WanSelfAttention):

    def __init__(self,
                 dim,
                 num_heads,
                 window_size=(-1, -1),
                 qk_norm=True,
                 eps=1e-6, operation_settings={}):
        super().__init__(dim, num_heads, window_size, qk_norm, eps)

        self.k_img = operation_settings.get("operations").Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.v_img = operation_settings.get("operations").Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        # self.alpha = nn.Parameter(torch.zeros((1, )))
        self.norm_k_img = WanRMSNorm(dim, eps=eps, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")) if qk_norm else nn.Identity()

    def forward(self, x, context, context_lens):
        r"""
        Args:
            x(Tensor): Shape [B, L1, C]
            context(Tensor): Shape [B, L2, C]
            context_lens(Tensor): Shape [B]
        """
        context_img = context[:, :257]
        context = context[:, 257:]

        # compute query, key, value
        q = self.norm_q(self.q(x))
        k = self.norm_k(self.k(context))
        v = self.v(context)
        k_img = self.norm_k_img(self.k_img(context_img))
        v_img = self.v_img(context_img)
        img_x = optimized_attention(q, k_img, v_img, heads=self.num_heads)
        # compute attention
        x = optimized_attention(q, k, v, heads=self.num_heads)

        # output
        x = x + img_x
        x = self.o(x)
        return x


WAN_CROSSATTENTION_CLASSES = {
    't2v_cross_attn': WanT2VCrossAttention,
    'i2v_cross_attn': WanI2VCrossAttention,
}


class WanAttentionBlock(nn.Module):

    def __init__(self,
                 cross_attn_type,
                 dim,
                 ffn_dim,
                 num_heads,
                 window_size=(-1, -1),
                 qk_norm=True,
                 cross_attn_norm=False,
                 eps=1e-6, operation_settings={}):
        super().__init__()
        self.dim = dim
        self.ffn_dim = ffn_dim
        self.num_heads = num_heads
        self.window_size = window_size
        self.qk_norm = qk_norm
        self.cross_attn_norm = cross_attn_norm
        self.eps = eps

        # layers
        self.norm1 = operation_settings.get("operations").LayerNorm(dim, eps, elementwise_affine=False, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.self_attn = WanSelfAttention(dim, num_heads, window_size, qk_norm,
                                          eps, operation_settings=operation_settings)
        self.norm3 = operation_settings.get("operations").LayerNorm(
            dim, eps,
            elementwise_affine=True, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")) if cross_attn_norm else nn.Identity()
        self.cross_attn = WAN_CROSSATTENTION_CLASSES[cross_attn_type](dim,
                                                                      num_heads,
                                                                      (-1, -1),
                                                                      qk_norm,
                                                                      eps, operation_settings=operation_settings)
        self.norm2 = operation_settings.get("operations").LayerNorm(dim, eps, elementwise_affine=False, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.ffn = nn.Sequential(
            operation_settings.get("operations").Linear(dim, ffn_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")), nn.GELU(approximate='tanh'),
            operation_settings.get("operations").Linear(ffn_dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))

        # modulation
        self.modulation = nn.Parameter(torch.empty(1, 6, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))

    def forward(
        self,
        x,
        e,
        seq_lens,
        grid_sizes,
        freqs,
        context,
        context_lens,
    ):
        r"""
        Args:
            x(Tensor): Shape [B, L, C]
            e(Tensor): Shape [B, 6, C]
            seq_lens(Tensor): Shape [B], length of each sequence in batch
            grid_sizes(Tensor): Shape [B, 3], the second dimension contains (F, H, W)
            freqs(Tensor): Rope freqs, shape [1024, C / num_heads / 2]
        """
        # assert e.dtype == torch.float32

        e = (self.modulation + e).chunk(6, dim=1)
        # assert e[0].dtype == torch.float32

        # self-attention
        y = self.self_attn(
            self.norm1(x) * (1 + e[1]) + e[0], seq_lens, grid_sizes,
            freqs)

        x = x + y * e[2]

        # cross-attention & ffn function
        def cross_attn_ffn(x, context, context_lens, e):
            x = x + self.cross_attn(self.norm3(x), context, context_lens)
            y = self.ffn(self.norm2(x) * (1 + e[4]) + e[3])
            x = x + y * e[5]
            return x

        x = cross_attn_ffn(x, context, context_lens, e)
        return x


class Head(nn.Module):

    def __init__(self, dim, out_dim, patch_size, eps=1e-6, operation_settings={}):
        super().__init__()
        self.dim = dim
        self.out_dim = out_dim
        self.patch_size = patch_size
        self.eps = eps

        # layers
        out_dim = math.prod(patch_size) * out_dim
        self.norm = operation_settings.get("operations").LayerNorm(dim, eps, elementwise_affine=False, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.head = operation_settings.get("operations").Linear(dim, out_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))

        # modulation
        self.modulation = nn.Parameter(torch.empty(1, 2, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))

    def forward(self, x, e):
        r"""
        Args:
            x(Tensor): Shape [B, L1, C]
            e(Tensor): Shape [B, C]
        """
        # assert e.dtype == torch.float32
        e = (self.modulation + e.unsqueeze(1)).chunk(2, dim=1)
        x = (self.head(self.norm(x) * (1 + e[1]) + e[0]))
        return x


class MLPProj(torch.nn.Module):

    def __init__(self, in_dim, out_dim, operation_settings={}):
        super().__init__()

        self.proj = torch.nn.Sequential(
            operation_settings.get("operations").LayerNorm(in_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")), operation_settings.get("operations").Linear(in_dim, in_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")),
            torch.nn.GELU(), operation_settings.get("operations").Linear(in_dim, out_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")),
            operation_settings.get("operations").LayerNorm(out_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))

    def forward(self, image_embeds):
        clip_extra_context_tokens = self.proj(image_embeds)
        return clip_extra_context_tokens


class WanModel(torch.nn.Module):
    r"""
    Wan diffusion backbone supporting both text-to-video and image-to-video.
    """

    def __init__(self,
                 model_type='t2v',
                 patch_size=(1, 2, 2),
                 text_len=512,
                 in_dim=16,
                 dim=2048,
                 ffn_dim=8192,
                 freq_dim=256,
                 text_dim=4096,
                 out_dim=16,
                 num_heads=16,
                 num_layers=32,
                 window_size=(-1, -1),
                 qk_norm=True,
                 cross_attn_norm=True,
                 eps=1e-6,
                 image_model=None,
                 device=None,
                 dtype=None,
                 operations=None,
                 ):
        r"""
        Initialize the diffusion model backbone.

        Args:
            model_type (`str`, *optional*, defaults to 't2v'):
                Model variant - 't2v' (text-to-video) or 'i2v' (image-to-video)
            patch_size (`tuple`, *optional*, defaults to (1, 2, 2)):
                3D patch dimensions for video embedding (t_patch, h_patch, w_patch)
            text_len (`int`, *optional*, defaults to 512):
                Fixed length for text embeddings
            in_dim (`int`, *optional*, defaults to 16):
                Input video channels (C_in)
            dim (`int`, *optional*, defaults to 2048):
                Hidden dimension of the transformer
            ffn_dim (`int`, *optional*, defaults to 8192):
                Intermediate dimension in feed-forward network
            freq_dim (`int`, *optional*, defaults to 256):
                Dimension for sinusoidal time embeddings
            text_dim (`int`, *optional*, defaults to 4096):
                Input dimension for text embeddings
            out_dim (`int`, *optional*, defaults to 16):
                Output video channels (C_out)
            num_heads (`int`, *optional*, defaults to 16):
                Number of attention heads
            num_layers (`int`, *optional*, defaults to 32):
                Number of transformer blocks
            window_size (`tuple`, *optional*, defaults to (-1, -1)):
                Window size for local attention (-1 indicates global attention)
            qk_norm (`bool`, *optional*, defaults to True):
                Enable query/key normalization
            cross_attn_norm (`bool`, *optional*, defaults to False):
                Enable cross-attention normalization
            eps (`float`, *optional*, defaults to 1e-6):
                Epsilon value for normalization layers
        """

        super().__init__()
        self.dtype = dtype
        operation_settings = {"operations": operations, "device": device, "dtype": dtype}

        assert model_type in ['t2v', 'i2v']
        self.model_type = model_type

        self.patch_size = patch_size
        self.text_len = text_len
        self.in_dim = in_dim
        self.dim = dim
        self.ffn_dim = ffn_dim
        self.freq_dim = freq_dim
        self.text_dim = text_dim
        self.out_dim = out_dim
        self.num_heads = num_heads
        self.num_layers = num_layers
        self.window_size = window_size
        self.qk_norm = qk_norm
        self.cross_attn_norm = cross_attn_norm
        self.eps = eps

        # embeddings
        self.patch_embedding = operations.Conv3d(
            in_dim, dim, kernel_size=patch_size, stride=patch_size, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
        self.text_embedding = nn.Sequential(
            operations.Linear(text_dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")), nn.GELU(approximate='tanh'),
            operations.Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))

        self.time_embedding = nn.Sequential(
            operations.Linear(freq_dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")), nn.SiLU(), operations.Linear(dim, dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))
        self.time_projection = nn.Sequential(nn.SiLU(), operations.Linear(dim, dim * 6, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))

        # blocks
        cross_attn_type = 't2v_cross_attn' if model_type == 't2v' else 'i2v_cross_attn'
        self.blocks = nn.ModuleList([
            WanAttentionBlock(cross_attn_type, dim, ffn_dim, num_heads,
                              window_size, qk_norm, cross_attn_norm, eps, operation_settings=operation_settings)
            for _ in range(num_layers)
        ])

        # head
        self.head = Head(dim, out_dim, patch_size, eps, operation_settings=operation_settings)

        # buffers (don't use register_buffer otherwise dtype will be changed in to())
        assert (dim % num_heads) == 0 and (dim // num_heads) % 2 == 0
        d = dim // num_heads
        self.register_buffer("freqs", torch.cat([
            rope_params(1024, d - 4 * (d // 6)),
            rope_params(1024, 2 * (d // 6)),
            rope_params(1024, 2 * (d // 6))
        ], dim=1), persistent=False)

        if model_type == 'i2v':
            self.img_emb = MLPProj(1280, dim, operation_settings=operation_settings)

    def forward_orig(
        self,
        x,
        t,
        context,
        seq_len=None,
        clip_fea=None,
        y=None,
    ):
        r"""
        Forward pass through the diffusion model

        Args:
            x (List[Tensor]):
                List of input video tensors, each with shape [C_in, F, H, W]
            t (Tensor):
                Diffusion timesteps tensor of shape [B]
            context (List[Tensor]):
                List of text embeddings each with shape [L, C]
            seq_len (`int`):
                Maximum sequence length for positional encoding
            clip_fea (Tensor, *optional*):
                CLIP image features for image-to-video mode
            y (List[Tensor], *optional*):
                Conditional video inputs for image-to-video mode, same shape as x

        Returns:
            List[Tensor]:
                List of denoised video tensors with original input shapes [C_out, F, H / 8, W / 8]
        """
        if self.model_type == 'i2v':
            assert clip_fea is not None and y is not None
        # params
        # device = self.patch_embedding.weight.device
        # if self.freqs.device != device:
        #     self.freqs = self.freqs.to(device)

        if y is not None:
            x = [torch.cat([u, v], dim=0) for u, v in zip(x, y)]

        # embeddings
        x = [self.patch_embedding(u) for u in x]
        grid_sizes = torch.stack(
            [torch.tensor(u.shape[2:], dtype=torch.long) for u in x])
        x = [u.flatten(2).transpose(1, 2) for u in x]
        seq_lens = torch.tensor([u.size(1) for u in x], dtype=torch.long)
        if seq_len is not None:
            assert seq_lens.max() <= seq_len
            x = torch.cat([
                torch.cat([u, u.new_zeros(1, seq_len - u.size(1), u.size(2))],
                          dim=1) for u in x
            ])
        elif len(x) == 1:
            x = x[0]

        # time embeddings
        e = self.time_embedding(
            sinusoidal_embedding_1d(self.freq_dim, t).to(dtype=x[0].dtype))
        e0 = self.time_projection(e).unflatten(1, (6, self.dim))

        # context
        context_lens = None
        context = self.text_embedding(
            torch.cat([
                torch.cat(
                    [u, u.new_zeros(u.size(0), self.text_len - u.size(1), u.size(2))], dim=1)
                for u in context
            ], dim=0))

        if clip_fea is not None:
            context_clip = self.img_emb(clip_fea)  # bs x 257 x dim
            context = torch.concat([context_clip, context], dim=1)

        # arguments
        kwargs = dict(
            e=e0,
            seq_lens=seq_lens,
            grid_sizes=grid_sizes,
            freqs=self.freqs,
            context=context,
            context_lens=context_lens)

        for block in self.blocks:
            x = block(x, **kwargs)

        # head
        x = self.head(x, e)

        # unpatchify
        x = self.unpatchify(x, grid_sizes)
        return x
        # return [u.float() for u in x]

    def forward(self, x, t, context, y=None, image=None, **kwargs):
        return self.forward_orig([x], t, [context], clip_fea=y, y=image)[0]

    def unpatchify(self, x, grid_sizes):
        r"""
        Reconstruct video tensors from patch embeddings.

        Args:
            x (List[Tensor]):
                List of patchified features, each with shape [L, C_out * prod(patch_size)]
            grid_sizes (Tensor):
                Original spatial-temporal grid dimensions before patching,
                    shape [B, 3] (3 dimensions correspond to F_patches, H_patches, W_patches)

        Returns:
            List[Tensor]:
                Reconstructed video tensors with shape [C_out, F, H / 8, W / 8]
        """

        c = self.out_dim
        out = []
        for u, v in zip(x, grid_sizes.tolist()):
            u = u[:math.prod(v)].view(*v, *self.patch_size, c)
            u = torch.einsum('fhwpqrc->cfphqwr', u)
            u = u.reshape(c, *[i * j for i, j in zip(v, self.patch_size)])
            out.append(u)
        return out