humo/models/wan_modules/model_humo.py

import torch
from torch import nn

from common.distributed import get_device
from models.audio.audio_proj import AudioProjModel

import torch.cuda.amp as amp
import math
from humo.models.wan_modules.attention import flash_attention
from common.distributed.advanced import is_unified_parallel_initialized

import types

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


@amp.autocast(enabled=False)
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.float32).div(dim)))
    freqs = torch.polar(torch.ones_like(freqs), freqs)
    return freqs


@amp.autocast(enabled=False)
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.float32).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).float()


class WanRMSNorm(nn.Module):

    def __init__(self, dim, eps=1e-5):
        super().__init__()
        self.dim = dim
        self.eps = eps
        self.weight = nn.Parameter(torch.ones(dim))

    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 WanLayerNorm(nn.LayerNorm):

    def __init__(self, dim, eps=1e-6, elementwise_affine=False):
        super().__init__(dim, elementwise_affine=elementwise_affine, eps=eps)

    def forward(self, x):
        r"""

        Args:

            x(Tensor): Shape [B, L, C]

        """
        return super().forward(x.float()).type_as(x)


class WanSelfAttention(nn.Module):

    def __init__(self,

                 dim,

                 num_heads,

                 window_size=(-1, -1),

                 qk_norm=True,

                 eps=1e-6):
        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 = nn.Linear(dim, dim)
        self.k = nn.Linear(dim, dim)
        self.v = nn.Linear(dim, dim)
        self.o = nn.Linear(dim, dim)
        self.norm_q = WanRMSNorm(dim, eps=eps) if qk_norm else nn.Identity()
        self.norm_k = WanRMSNorm(dim, eps=eps) 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], torch.Size([1, 9360, 5120])

            seq_lens(Tensor): Shape [B], tensor([9360])

            grid_sizes(Tensor): Shape [B, 3], the second dimension contains (F, H, W), tensor([[ 6, 30, 52]])

            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 = flash_attention(
            q=rope_apply(q, grid_sizes, freqs),
            k=rope_apply(k, grid_sizes, freqs),
            v=v,
            k_lens=seq_lens,
            window_size=self.window_size)

        # output
        x = x.flatten(2)
        x = self.o(x)
        return x


class WanSelfAttentionSepKVDim(nn.Module):

    def __init__(self,

                 kv_dim,

                 dim,

                 num_heads,

                 window_size=(-1, -1),

                 qk_norm=True,

                 eps=1e-6):
        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 = nn.Linear(dim, dim)
        self.k = nn.Linear(kv_dim, dim)
        self.v = nn.Linear(kv_dim, dim)
        self.o = nn.Linear(dim, dim)
        self.norm_q = WanRMSNorm(dim, eps=eps) if qk_norm else nn.Identity()
        self.norm_k = WanRMSNorm(dim, eps=eps) 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], torch.Size([1, 9360, 5120])

            seq_lens(Tensor): Shape [B], tensor([9360])

            grid_sizes(Tensor): Shape [B, 3], the second dimension contains (F, H, W), tensor([[ 6, 30, 52]])

            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 = flash_attention(
            q=rope_apply(q, grid_sizes, freqs),
            k=rope_apply(k, grid_sizes, freqs),
            v=v,
            k_lens=seq_lens,
            window_size=self.window_size)

        # output
        x = x.flatten(2)
        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]

        """
        b, n, d = x.size(0), self.num_heads, self.head_dim

        # compute query, key, value
        q = self.norm_q(self.q(x)).view(b, -1, n, d)
        k = self.norm_k(self.k(context)).view(b, -1, n, d)
        v = self.v(context).view(b, -1, n, d)

        # compute attention
        x = flash_attention(q, k, v, k_lens=context_lens)

        # output
        x = x.flatten(2)
        x = self.o(x)
        return x


class WanT2VCrossAttentionGather(WanSelfAttentionSepKVDim):

    def forward(self, x, context, context_lens, grid_sizes, freqs, audio_seq_len):
        b, n, d = x.size(0), self.num_heads, self.head_dim

        q = self.norm_q(self.q(x)).view(b, -1, n, d)
        k = self.norm_k(self.k(context)).view(b, -1, n, d)
        v = self.v(context).view(b, -1, n, d)

        # --- NEW: derive sizes from shapes (SymInts), no int(tensor) casts ---
        Lq = q.shape[1]                 # total video tokens per sample
        # audio has 16 tokens per frame -> frames = audio_tokens // 16
        frames = (context.shape[1] // 16)
        hlen_wlen = Lq // frames        # tokens per frame = H*W

        # Now reshape using SymInt-derived sizes
        q = q.reshape(-1, hlen_wlen, n, d)
        k = k.reshape(-1, 16, n, d)
        v = v.reshape(-1, 16, n, d)

        x = flash_attention(q, k, v, k_lens=None)
        x = x.view(b, -1, n, d).flatten(2)
        x = self.o(x)
        return x

    # def forward(self, x, context, context_lens, grid_sizes, freqs, audio_seq_len):
    #     r"""
    #     Args:
    #         x(Tensor): Shape [B, L1, C] - video tokens
    #         context(Tensor): Shape [B, L2, C] - audio tokens with shape [B, frames*16, 1536]
    #         context_lens(Tensor): Shape [B] - actually seq_lens from call (video sequence length)
    #         grid_sizes(Tensor): Shape [B, 3] - video grid dimensions (F, H, W)
    #         freqs(Tensor): RoPE frequencies
    #         audio_seq_len(Tensor): Actual audio sequence length (frames * 16)
    #     """
    #     b, n, d = x.size(0), self.num_heads, self.head_dim

    #     q = self.norm_q(self.q(x)).view(b, -1, n, d)
    #     k = self.norm_k(self.k(context)).view(b, -1, n, d)
    #     v = self.v(context).view(b, -1, n, d)

    #     # Handle video spatial structure
    #     hlen_wlen = int(grid_sizes[0][1] * grid_sizes[0][2])
    #     q = q.reshape(-1, hlen_wlen, n, d)
        
    #     # Handle audio temporal structure (16 tokens per frame)
    #     k = k.reshape(-1, 16, n, d)
    #     v = v.reshape(-1, 16, n, d)

    #     # Cross-attention
    #     x = flash_attention(q, k, v, k_lens=None)  # No masking for audio
        
    #     x = x.view(b, -1, n, d).flatten(2)
    #     x = self.o(x)
    #     return x


class AudioCrossAttentionWrapper(nn.Module):
    def __init__(self, dim, kv_dim, num_heads, qk_norm=True, eps=1e-6,):
        super().__init__()

        self.audio_cross_attn = WanT2VCrossAttentionGather(
                kv_dim, dim, num_heads, (-1, -1), qk_norm, eps)
        self.norm1_audio = WanLayerNorm(dim, eps,
            elementwise_affine=True)

    def forward(self, x, audio, seq_lens, grid_sizes, freqs, audio_seq_len):
        x = x + self.audio_cross_attn(
            self.norm1_audio(x), audio, seq_lens, grid_sizes, freqs, audio_seq_len)
        return x
        

class WanI2VCrossAttention(WanSelfAttention):

    def __init__(self,

                 dim,

                 num_heads,

                 window_size=(-1, -1),

                 qk_norm=True,

                 eps=1e-6):
        super().__init__(dim, num_heads, window_size, qk_norm, eps)

    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]

        """
        b, n, d = x.size(0), self.num_heads, self.head_dim

        # compute query, key, value
        q = self.norm_q(self.q(x)).view(b, -1, n, d)
        k = self.norm_k(self.k(context)).view(b, -1, n, d)
        v = self.v(context).view(b, -1, n, d)
        x = flash_attention(q, k, v, k_lens=context_lens)
        
        # output
        x = x.flatten(2)
        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,

                 use_audio=True):
        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 = WanLayerNorm(dim, eps)
        self.self_attn = WanSelfAttention(dim, num_heads, window_size, qk_norm,
                                          eps)
        self.norm3 = WanLayerNorm(
            dim, eps,
            elementwise_affine=True) if cross_attn_norm else nn.Identity()
        self.cross_attn = WAN_CROSSATTENTION_CLASSES[cross_attn_type](dim,
                                                                      num_heads,
                                                                      (-1, -1),
                                                                      qk_norm,
                                                                      eps)
        self.norm2 = WanLayerNorm(dim, eps)
        self.ffn = nn.Sequential(
            nn.Linear(dim, ffn_dim), nn.GELU(approximate='tanh'),
            nn.Linear(ffn_dim, dim))

        # modulation
        self.modulation = nn.Parameter(torch.randn(1, 6, dim) / dim**0.5)

        self.use_audio = use_audio
        if use_audio:
            self.audio_cross_attn_wrapper = AudioCrossAttentionWrapper(dim, 1536, num_heads, qk_norm, eps)

    def forward(

        self,

        x, # torch.Size([1, 9360, 5120])

        e, # torch.Size([1, 6, 5120])

        seq_lens, # tensor([9360])

        grid_sizes, # tensor([[ 6, 30, 52]])

        freqs, # torch.Size([1024, 64])

        context, # torch.Size([1, 512, 5120])

        context_lens, # None

        audio=None, # None

        audio_seq_len=None,

        ref_num_list=None,

    ):
        r"""

        Args:

            x(Tensor): Shape [B, L, C]

            e(Tensor): Shape [B, L, C]

            audio(Tensor): Shape [B, L, 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]

            ref_num_list: 配合seq_lens可以查到reference image在倒数第几个

        """
        assert e.dtype == torch.float32
        with torch.amp.autocast('cuda', 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).float() * (1 + e[1]) + e[0], seq_lens, grid_sizes,
            freqs)
        with torch.amp.autocast('cuda', dtype=torch.float32):
            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)
            
            if self.use_audio:
                x = self.audio_cross_attn_wrapper(x, audio, seq_lens, grid_sizes, freqs, audio_seq_len)

            y = self.ffn(self.norm2(x).float() * (1 + e[4]) + e[3])
            with torch.amp.autocast('cuda', dtype=torch.float32):
                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):
        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 = WanLayerNorm(dim, eps)
        self.head = nn.Linear(dim, out_dim)

        # modulation
        self.modulation = nn.Parameter(torch.randn(1, 2, dim) / dim**0.5)

    def forward(self, x, e):
        r"""

        Args:

            x(Tensor): Shape [B, L1, C]

            e(Tensor): Shape [B, C]

        """
        assert e.dtype == torch.float32
        with torch.amp.autocast('cuda', 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):
        super().__init__()

        self.proj = torch.nn.Sequential(
            torch.nn.LayerNorm(in_dim), torch.nn.Linear(in_dim, in_dim),
            torch.nn.GELU(), torch.nn.Linear(in_dim, out_dim),
            torch.nn.LayerNorm(out_dim))

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


class WanModel(nn.Module):
    r"""

    Wan diffusion backbone supporting both text-to-video and image-to-video.

    """

    ignore_for_config = [
        'patch_size', 'cross_attn_norm', 'qk_norm', 'text_dim', 'window_size'
    ]
    _no_split_modules = ['WanAttentionBlock']

    gradient_checkpointing = False

    def __init__(self,

                 model_type='t2v',

                 patch_size=(1, 2, 2),

                 text_len=512,

                 in_dim=16,

                 dim=2048,

                 ffn_dim=13824,

                 freq_dim=256,

                 text_dim=4096,

                 out_dim=16,

                 num_heads=40,

                 num_layers=40,

                 window_size=(-1, -1),

                 qk_norm=True,

                 cross_attn_norm=True,

                 eps=1e-6,

                 audio_token_num=16,

                 insert_audio=True):
        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__()

        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 = nn.Conv3d(
            in_dim, dim, kernel_size=patch_size, stride=patch_size)
        self.text_embedding = nn.Sequential(
            nn.Linear(text_dim, dim), nn.GELU(approximate='tanh'),
            nn.Linear(dim, dim))

        self.time_embedding = nn.Sequential(
            nn.Linear(freq_dim, dim), nn.SiLU(), nn.Linear(dim, dim))
        self.time_projection = nn.Sequential(nn.SiLU(), nn.Linear(dim, dim * 6))

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

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

        if self.insert_audio:
            self.audio_proj = AudioProjModel(seq_len=8, blocks=5, channels=1280, 
                intermediate_dim=512, output_dim=1536, context_tokens=audio_token_num)

        # RoPE freqs: register as a buffer so it moves with .to() / DDP and is tracked by compile
        assert (dim % num_heads) == 0 and (dim // num_heads) % 2 == 0
        d = dim // num_heads

        _freqs = torch.cat([
            rope_params(1024, d - 4 * (d // 6)),
            rope_params(1024, 2 * (d // 6)),
            rope_params(1024, 2 * (d // 6))
        ], dim=1)
        self.register_buffer("freqs", _freqs, persistent=False)

        # initialize weights
        self.init_weights()

        # initialize unified parallel
        if is_unified_parallel_initialized():
            print(f"Initializing WanModel with unified parallel initialized")
            from humo.models.distributed.dit_ulysses_sequence_parallel import ulysses_attn_forward, ulysses_dit_forward, ulysses_audio_cross_attn_forward
            for block in self.blocks:
                block.self_attn.forward = types.MethodType(ulysses_attn_forward, block.self_attn)
                if block.use_audio:
                    block.audio_cross_attn_wrapper.audio_cross_attn.forward = types.MethodType(ulysses_audio_cross_attn_forward, block.audio_cross_attn_wrapper.audio_cross_attn)
            self.forward = types.MethodType(ulysses_dit_forward, self)
        
    def forward(

        self,

        x,

        t,

        context,

        seq_len,

        audio=None,

        y=None,

        tea_cache=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 y is not None

        # params
        freqs = self.freqs

        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.unsqueeze(0)) 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)
        assert seq_lens.max() <= seq_len

        # pad to uniform length and batch
        x = torch.cat([
            torch.cat([u, u.new_zeros(1, seq_len - u.size(1), u.size(2))], dim=1)
            for u in x
        ])  # shape: [B, seq_len, C]

        # time embeddings
        with torch.amp.autocast('cuda', dtype=torch.float32):
            e = self.time_embedding(
                sinusoidal_embedding_1d(self.freq_dim, t).float()
            ).float()
            e0 = self.time_projection(e).unflatten(1, (6, self.dim)).float()
            assert e.dtype == torch.float32 and e0.dtype == torch.float32

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

        # audio (unchanged; not cached)
        if self.insert_audio:
            audio = [self.audio_proj(au.unsqueeze(0)).permute(0, 3, 1, 2) for au in audio]
            audio_seq_len = max(au.shape[2] for au in audio) * audio[0].shape[3]

            audio = [au.flatten(2).transpose(1, 2) for au in audio]  # [1, t*32, 1536]
            audio = torch.cat([
                torch.cat([au, au.new_zeros(1, int(audio_seq_len) - au.size(1), au.size(2))], dim=1)
                for au in audio
            ])
        else:
            audio = None
            audio_seq_len = None

        # ---- tea_cache integration (mirrors your working model) ----
        if tea_cache is not None:
            # Use the pre-block tokens 'x' and time-mod 'e0' to decide whether to reuse cache
            tea_cache_update = tea_cache.check(self, x, e0)
        else:
            tea_cache_update = False

        ori_x_len = x.shape[1]  # remember original token length before potential cache extension

        if tea_cache_update:
            # Let the cache inject/append any needed past states/tokens for reuse
            x = tea_cache.update(x)
        else:
            # arguments for blocks
            kwargs = dict(
                e=e0,
                seq_lens=seq_lens,
                grid_sizes=grid_sizes,
                freqs=freqs,
                context=context,
                context_lens=context_lens,
                audio=audio,
                audio_seq_len=audio_seq_len
            )

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

            if tea_cache is not None:
                x_cache = x[:, :ori_x_len]
                tea_cache.store(x_cache)

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

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


    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

    def init_weights(self):
        r"""

        Initialize model parameters using Xavier initialization.

        """

        # basic init
        for m in self.modules():
            if isinstance(m, nn.Linear):
                nn.init.xavier_uniform_(m.weight)
                if m.bias is not None:
                    nn.init.zeros_(m.bias)

        # init embeddings
        nn.init.xavier_uniform_(self.patch_embedding.weight.flatten(1))
        for m in self.text_embedding.modules():
            if isinstance(m, nn.Linear):
                nn.init.normal_(m.weight, std=.02)
        for m in self.time_embedding.modules():
            if isinstance(m, nn.Linear):
                nn.init.normal_(m.weight, std=.02)

        # init output layer
        nn.init.zeros_(self.head.head.weight)