layers.py

import math
import collections.abc

import torch
import torch.nn as nn
import torch.nn.functional as F
import functools

from einops import rearrange
from itertools import repeat
from functools import partial
from .utils import approx_gelu, get_layernorm, t2i_modulate
from typing import Optional


try:
    import xformers
    HAS_XFORMERS = True
except:
    HAS_XFORMERS = False


# =================
# STDiT2Block
# =================
class STDiT2Block(nn.Module):
    def __init__(
        self,
        hidden_size,
        num_heads,
        mlp_ratio=4.0,
        drop_path=0.0,
        enable_flash_attn=False,
        enable_layernorm_kernel=False,
        enable_sequence_parallelism=False,
        rope=None,
        qk_norm=False,
    ):
        super().__init__()
        self.hidden_size = hidden_size
        self.enable_flash_attn = enable_flash_attn
        self._enable_sequence_parallelism = enable_sequence_parallelism

        assert not self._enable_sequence_parallelism, "Sequence parallelism is not supported."
        if enable_sequence_parallelism:
            self.attn_cls = SeqParallelAttention
            self.mha_cls = SeqParallelMultiHeadCrossAttention
        else:
            self.attn_cls = Attention
            self.mha_cls = MultiHeadCrossAttention

        # spatial branch
        self.norm1 = get_layernorm(hidden_size, eps=1e-6, affine=False, use_kernel=enable_layernorm_kernel)
        self.attn = self.attn_cls(
            hidden_size,
            num_heads=num_heads,
            qkv_bias=True,
            enable_flash_attn=enable_flash_attn,
            qk_norm=qk_norm,
        )
        self.scale_shift_table = nn.Parameter(torch.randn(6, hidden_size) / hidden_size**0.5)

        # cross attn
        self.cross_attn = self.mha_cls(hidden_size, num_heads)

        # mlp branch
        self.norm2 = get_layernorm(hidden_size, eps=1e-6, affine=False, use_kernel=enable_layernorm_kernel)
        self.mlp = Mlp(
            in_features=hidden_size, hidden_features=int(hidden_size * mlp_ratio), act_layer=approx_gelu, drop=0
        )
        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()

        # temporal branch
        self.norm_temp = get_layernorm(hidden_size, eps=1e-6, affine=False, use_kernel=enable_layernorm_kernel)  # new
        self.attn_temp = self.attn_cls(
            hidden_size,
            num_heads=num_heads,
            qkv_bias=True,
            enable_flash_attn=self.enable_flash_attn,
            rope=rope,
            qk_norm=qk_norm,
        )
        self.scale_shift_table_temporal = nn.Parameter(torch.randn(3, hidden_size) / hidden_size**0.5)  # new

    def t_mask_select(self, x_mask, x, masked_x, T, S):
        # x: [B, (T, S), C]
        # mased_x: [B, (T, S), C]
        # x_mask: [B, T]
        x = rearrange(x, "B (T S) C -> B T S C", T=T, S=S)
        masked_x = rearrange(masked_x, "B (T S) C -> B T S C", T=T, S=S)
        x = torch.where(x_mask[:, :, None, None], x, masked_x)
        x = rearrange(x, "B T S C -> B (T S) C")
        return x

    def forward(self, x, y, t, t_tmp, mask=None, x_mask=None, t0=None, t0_tmp=None, T=None, S=None):
        B, N, C = x.shape

        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (
            self.scale_shift_table[None] + t.reshape(B, 6, -1)
        ).chunk(6, dim=1)
        shift_tmp, scale_tmp, gate_tmp = (self.scale_shift_table_temporal[None] + t_tmp.reshape(B, 3, -1)).chunk(
            3, dim=1
        )
        if x_mask is not None:
            shift_msa_zero, scale_msa_zero, gate_msa_zero, shift_mlp_zero, scale_mlp_zero, gate_mlp_zero = (
                self.scale_shift_table[None] + t0.reshape(B, 6, -1)
            ).chunk(6, dim=1)
            shift_tmp_zero, scale_tmp_zero, gate_tmp_zero = (
                self.scale_shift_table_temporal[None] + t0_tmp.reshape(B, 3, -1)
            ).chunk(3, dim=1)

        # modulate
        x_m = t2i_modulate(self.norm1(x), shift_msa, scale_msa)
        if x_mask is not None:
            x_m_zero = t2i_modulate(self.norm1(x), shift_msa_zero, scale_msa_zero)
            x_m = self.t_mask_select(x_mask, x_m, x_m_zero, T, S)

        # spatial branch
        x_s = rearrange(x_m, "B (T S) C -> (B T) S C", T=T, S=S)
        x_s = self.attn(x_s)
        x_s = rearrange(x_s, "(B T) S C -> B (T S) C", T=T, S=S)
        if x_mask is not None:
            x_s_zero = gate_msa_zero * x_s
            x_s = gate_msa * x_s
            x_s = self.t_mask_select(x_mask, x_s, x_s_zero, T, S)
        else:
            x_s = gate_msa * x_s
        x = x + self.drop_path(x_s)

        # modulate
        x_m = t2i_modulate(self.norm_temp(x), shift_tmp, scale_tmp)
        if x_mask is not None:
            x_m_zero = t2i_modulate(self.norm_temp(x), shift_tmp_zero, scale_tmp_zero)
            x_m = self.t_mask_select(x_mask, x_m, x_m_zero, T, S)

        # temporal branch
        x_t = rearrange(x_m, "B (T S) C -> (B S) T C", T=T, S=S)
        x_t = self.attn_temp(x_t)
        x_t = rearrange(x_t, "(B S) T C -> B (T S) C", T=T, S=S)
        if x_mask is not None:
            x_t_zero = gate_tmp_zero * x_t
            x_t = gate_tmp * x_t
            x_t = self.t_mask_select(x_mask, x_t, x_t_zero, T, S)
        else:
            x_t = gate_tmp * x_t
        x = x + self.drop_path(x_t)

        # cross attn
        x = x + self.cross_attn(x, y, mask)

        # modulate
        x_m = t2i_modulate(self.norm2(x), shift_mlp, scale_mlp)
        if x_mask is not None:
            x_m_zero = t2i_modulate(self.norm2(x), shift_mlp_zero, scale_mlp_zero)
            x_m = self.t_mask_select(x_mask, x_m, x_m_zero, T, S)

        # mlp
        x_mlp = self.mlp(x_m)
        if x_mask is not None:
            x_mlp_zero = gate_mlp_zero * x_mlp
            x_mlp = gate_mlp * x_mlp
            x_mlp = self.t_mask_select(x_mask, x_mlp, x_mlp_zero, T, S)
        else:
            x_mlp = gate_mlp * x_mlp
        x = x + self.drop_path(x_mlp)

        return x
    

# =================
# Attention
# =================
class LlamaRMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        """
        LlamaRMSNorm is equivalent to T5LayerNorm
        """
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.variance_epsilon = eps

    def forward(self, hidden_states):
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(torch.float32)
        variance = hidden_states.pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
        return self.weight * hidden_states.to(input_dtype)
    
class Attention(nn.Module):
    def __init__(
        self,
        dim: int,
        num_heads: int = 8,
        qkv_bias: bool = False,
        qk_norm: bool = False,
        attn_drop: float = 0.0,
        proj_drop: float = 0.0,
        norm_layer: nn.Module = LlamaRMSNorm,
        enable_flash_attn: bool = False,
        rope=None,
    ) -> None:
        super().__init__()
        assert dim % num_heads == 0, "dim should be divisible by num_heads"
        self.dim = dim
        self.num_heads = num_heads
        self.head_dim = dim // num_heads
        self.scale = self.head_dim**-0.5
        self.enable_flash_attn = enable_flash_attn

        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.q_norm = norm_layer(self.head_dim) if qk_norm else nn.Identity()
        self.k_norm = norm_layer(self.head_dim) if qk_norm else nn.Identity()
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        self.rope = False
        if rope is not None:
            self.rope = True
            self.rotary_emb = rope

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, N, C = x.shape
        # flash attn is not memory efficient for small sequences, this is empirical
        enable_flash_attn = self.enable_flash_attn and (N > B)
        qkv = self.qkv(x)
        qkv_shape = (B, N, 3, self.num_heads, self.head_dim)

        qkv = qkv.view(qkv_shape).permute(2, 0, 3, 1, 4)
        q, k, v = qkv.unbind(0)
        if self.rope:
            q = self.rotary_emb(q)
            k = self.rotary_emb(k)
        q, k = self.q_norm(q), self.k_norm(k)

        if enable_flash_attn:
            from flash_attn import flash_attn_func

            # (B, #heads, N, #dim) -> (B, N, #heads, #dim)
            q = q.permute(0, 2, 1, 3)
            k = k.permute(0, 2, 1, 3)
            v = v.permute(0, 2, 1, 3)
            x = flash_attn_func(
                q,
                k,
                v,
                dropout_p=self.attn_drop.p if self.training else 0.0,
                softmax_scale=self.scale,
            )
        else:
            dtype = q.dtype
            q = q * self.scale
            attn = q @ k.transpose(-2, -1)  # translate attn to float32
            attn = attn.to(torch.float32)
            attn = attn.softmax(dim=-1)
            attn = attn.to(dtype)  # cast back attn to original dtype
            attn = self.attn_drop(attn)
            x = attn @ v

        x_output_shape = (B, N, C)
        if not enable_flash_attn:
            x = x.transpose(1, 2)
        x = x.reshape(x_output_shape)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x


# ========================
# MultiHeadCrossAttention
# ========================
class MultiHeadCrossAttention(nn.Module):
    def __init__(self, d_model, num_heads, attn_drop=0.0, proj_drop=0.0):
        super(MultiHeadCrossAttention, self).__init__()
        assert d_model % num_heads == 0, "d_model must be divisible by num_heads"

        self.d_model = d_model
        self.num_heads = num_heads
        self.head_dim = d_model // num_heads

        self.q_linear = nn.Linear(d_model, d_model)
        self.kv_linear = nn.Linear(d_model, d_model * 2)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(d_model, d_model)
        self.proj_drop = nn.Dropout(proj_drop)

    def forward(self, x, cond, mask=None):
        # query/value: img tokens; key: condition; mask: if padding tokens
        B, N, C = x.shape

        q = self.q_linear(x).view(1, -1, self.num_heads, self.head_dim)
        kv = self.kv_linear(cond).view(1, -1, 2, self.num_heads, self.head_dim)
        k, v = kv.unbind(2)

        attn_bias = None
        if mask is not None:
            attn_bias = xformers.ops.fmha.BlockDiagonalMask.from_seqlens([N] * B, mask)
        x = xformers.ops.memory_efficient_attention(q, k, v, p=self.attn_drop.p, attn_bias=attn_bias)

        x = x.view(B, -1, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x
    

# =================
# Timm Components 
# =================
def drop_path(x, drop_prob: float = 0., training: bool = False, scale_by_keep: bool = True):
    """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).
    This is the same as the DropConnect impl I created for EfficientNet, etc networks, however,
    the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper...
    See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... I've opted for
    changing the layer and argument names to 'drop path' rather than mix DropConnect as a layer name and use
    'survival rate' as the argument.
    """
    if drop_prob == 0. or not training:
        return x
    keep_prob = 1 - drop_prob
    shape = (x.shape[0],) + (1,) * (x.ndim - 1)  # work with diff dim tensors, not just 2D ConvNets
    random_tensor = x.new_empty(shape).bernoulli_(keep_prob)
    if keep_prob > 0.0 and scale_by_keep:
        random_tensor.div_(keep_prob)
    return x * random_tensor

class DropPath(nn.Module):
    """Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks).
    """
    def __init__(self, drop_prob: float = 0., scale_by_keep: bool = True):
        super(DropPath, self).__init__()
        self.drop_prob = drop_prob
        self.scale_by_keep = scale_by_keep

    def forward(self, x):
        return drop_path(x, self.drop_prob, self.training, self.scale_by_keep)

    def extra_repr(self):
        return f'drop_prob={round(self.drop_prob,3):0.3f}'
    
def _ntuple(n):
    def parse(x):
        if isinstance(x, collections.abc.Iterable) and not isinstance(x, str):
            return tuple(x)
        return tuple(repeat(x, n))
    return parse

to_2tuple = _ntuple(2)

class Mlp(nn.Module):
    """ MLP as used in Vision Transformer, MLP-Mixer and related networks
    """
    def __init__(
            self,
            in_features,
            hidden_features=None,
            out_features=None,
            act_layer=nn.GELU,
            norm_layer=None,
            bias=True,
            drop=0.,
            use_conv=False,
    ):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        bias = to_2tuple(bias)
        drop_probs = to_2tuple(drop)
        linear_layer = partial(nn.Conv2d, kernel_size=1) if use_conv else nn.Linear

        self.fc1 = linear_layer(in_features, hidden_features, bias=bias[0])
        self.act = act_layer()
        self.drop1 = nn.Dropout(drop_probs[0])
        self.norm = norm_layer(hidden_features) if norm_layer is not None else nn.Identity()
        self.fc2 = linear_layer(hidden_features, out_features, bias=bias[1])
        self.drop2 = nn.Dropout(drop_probs[1])

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop1(x)
        x = self.norm(x)
        x = self.fc2(x)
        x = self.drop2(x)
        return x
    

# =================
# Embedding
# =================
class CaptionEmbedder(nn.Module):
    """
    Embeds class labels into vector representations. Also handles label dropout for classifier-free guidance.
    """

    def __init__(
        self,
        in_channels,
        hidden_size,
        uncond_prob,
        act_layer=nn.GELU(approximate="tanh"),
        token_num=120,
    ):
        super().__init__()
        self.y_proj = Mlp(
            in_features=in_channels,
            hidden_features=hidden_size,
            out_features=hidden_size,
            act_layer=act_layer,
            drop=0,
        )
        self.register_buffer(
            "y_embedding",
            torch.randn(token_num, in_channels) / in_channels**0.5,
        )
        self.uncond_prob = uncond_prob

    def token_drop(self, caption, force_drop_ids=None):
        """
        Drops labels to enable classifier-free guidance.
        """
        if force_drop_ids is None:
            drop_ids = torch.rand(caption.shape[0]).cuda() < self.uncond_prob
        else:
            drop_ids = force_drop_ids == 1
        caption = torch.where(drop_ids[:, None, None, None], self.y_embedding, caption)
        return caption

    def forward(self, caption, train, force_drop_ids=None):
        if train:
            assert caption.shape[2:] == self.y_embedding.shape
        use_dropout = self.uncond_prob > 0
        if (train and use_dropout) or (force_drop_ids is not None):
            caption = self.token_drop(caption, force_drop_ids)
        caption = self.y_proj(caption)
        return caption
    

class PatchEmbed3D(nn.Module):
    """Video to Patch Embedding.

    Args:
        patch_size (int): Patch token size. Default: (2,4,4).
        in_chans (int): Number of input video channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        norm_layer (nn.Module, optional): Normalization layer. Default: None
    """

    def __init__(
        self,
        patch_size=(2, 4, 4),
        in_chans=3,
        embed_dim=96,
        norm_layer=None,
        flatten=True,
    ):
        super().__init__()
        self.patch_size = patch_size
        self.flatten = flatten

        self.in_chans = in_chans
        self.embed_dim = embed_dim

        self.proj = nn.Conv3d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    def forward(self, x):
        """Forward function."""
        # padding
        _, _, D, H, W = x.size()
        if W % self.patch_size[2] != 0:
            x = F.pad(x, (0, self.patch_size[2] - W % self.patch_size[2]))
        if H % self.patch_size[1] != 0:
            x = F.pad(x, (0, 0, 0, self.patch_size[1] - H % self.patch_size[1]))
        if D % self.patch_size[0] != 0:
            x = F.pad(x, (0, 0, 0, 0, 0, self.patch_size[0] - D % self.patch_size[0]))

        x = self.proj(x)  # (B C T H W)
        if self.norm is not None:
            D, Wh, Ww = x.size(2), x.size(3), x.size(4)
            x = x.flatten(2).transpose(1, 2)
            x = self.norm(x)
            x = x.transpose(1, 2).view(-1, self.embed_dim, D, Wh, Ww)
        if self.flatten:
            x = x.flatten(2).transpose(1, 2)  # BCTHW -> BNC
        return x
    
class T2IFinalLayer(nn.Module):
    """
    The final layer of PixArt.
    """

    def __init__(self, hidden_size, num_patch, out_channels, d_t=None, d_s=None):
        super().__init__()
        self.norm_final = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
        self.linear = nn.Linear(hidden_size, num_patch * out_channels, bias=True)
        self.scale_shift_table = nn.Parameter(torch.randn(2, hidden_size) / hidden_size**0.5)
        self.out_channels = out_channels
        self.d_t = d_t
        self.d_s = d_s

    def t_mask_select(self, x_mask, x, masked_x, T, S):
        # x: [B, (T, S), C]
        # mased_x: [B, (T, S), C]
        # x_mask: [B, T]
        x = rearrange(x, "B (T S) C -> B T S C", T=T, S=S)
        masked_x = rearrange(masked_x, "B (T S) C -> B T S C", T=T, S=S)
        x = torch.where(x_mask[:, :, None, None], x, masked_x)
        x = rearrange(x, "B T S C -> B (T S) C")
        return x

    def forward(self, x, t, x_mask=None, t0=None, T=None, S=None):
        if T is None:
            T = self.d_t
        if S is None:
            S = self.d_s
        shift, scale = (self.scale_shift_table[None] + t[:, None]).chunk(2, dim=1)
        x = t2i_modulate(self.norm_final(x), shift, scale)
        if x_mask is not None:
            shift_zero, scale_zero = (self.scale_shift_table[None] + t0[:, None]).chunk(2, dim=1)
            x_zero = t2i_modulate(self.norm_final(x), shift_zero, scale_zero)
            x = self.t_mask_select(x_mask, x, x_zero, T, S)
        x = self.linear(x)
        return x
    
class TimestepEmbedder(nn.Module):
    """
    Embeds scalar timesteps into vector representations.
    """

    def __init__(self, hidden_size, frequency_embedding_size=256):
        super().__init__()
        self.mlp = nn.Sequential(
            nn.Linear(frequency_embedding_size, hidden_size, bias=True),
            nn.SiLU(),
            nn.Linear(hidden_size, hidden_size, bias=True),
        )
        self.frequency_embedding_size = frequency_embedding_size

    @staticmethod
    def timestep_embedding(t, dim, max_period=10000):
        """
        Create sinusoidal timestep embeddings.
        :param t: a 1-D Tensor of N indices, one per batch element.
                          These may be fractional.
        :param dim: the dimension of the output.
        :param max_period: controls the minimum frequency of the embeddings.
        :return: an (N, D) Tensor of positional embeddings.
        """
        # https://github.com/openai/glide-text2im/blob/main/glide_text2im/nn.py
        half = dim // 2
        freqs = torch.exp(-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half)
        freqs = freqs.to(device=t.device)
        args = t[:, None].float() * freqs[None]
        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
        if dim % 2:
            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
        return embedding

    def forward(self, t, dtype):
        t_freq = self.timestep_embedding(t, self.frequency_embedding_size)
        if t_freq.dtype != dtype:
            t_freq = t_freq.to(dtype)
        t_emb = self.mlp(t_freq)
        return t_emb
    
class SizeEmbedder(TimestepEmbedder):
    """
    Embeds scalar timesteps into vector representations.
    """

    def __init__(self, hidden_size, frequency_embedding_size=256):
        super().__init__(hidden_size=hidden_size, frequency_embedding_size=frequency_embedding_size)
        self.mlp = nn.Sequential(
            nn.Linear(frequency_embedding_size, hidden_size, bias=True),
            nn.SiLU(),
            nn.Linear(hidden_size, hidden_size, bias=True),
        )
        self.frequency_embedding_size = frequency_embedding_size
        self.outdim = hidden_size

    def forward(self, s, bs):
        if s.ndim == 1:
            s = s[:, None]
        assert s.ndim == 2
        if s.shape[0] != bs:
            s = s.repeat(bs // s.shape[0], 1)
            assert s.shape[0] == bs
        b, dims = s.shape[0], s.shape[1]
        s = rearrange(s, "b d -> (b d)")
        s_freq = self.timestep_embedding(s, self.frequency_embedding_size).to(self.dtype)
        s_emb = self.mlp(s_freq)
        s_emb = rearrange(s_emb, "(b d) d2 -> b (d d2)", b=b, d=dims, d2=self.outdim)
        return s_emb

    @property
    def dtype(self):
        return next(self.parameters()).dtype


class PositionEmbedding2D(nn.Module):
    def __init__(self, dim: int) -> None:
        super().__init__()
        self.dim = dim
        assert dim % 4 == 0, "dim must be divisible by 4"
        half_dim = dim // 2
        inv_freq = 1.0 / (10000 ** (torch.arange(0, half_dim, 2).float() / half_dim))
        self.register_buffer("inv_freq", inv_freq, persistent=False)

    def _get_sin_cos_emb(self, t: torch.Tensor):
        out = torch.einsum("i,d->id", t, self.inv_freq)
        emb_cos = torch.cos(out)
        emb_sin = torch.sin(out)
        return torch.cat((emb_sin, emb_cos), dim=-1)

    @functools.lru_cache(maxsize=512)
    def _get_cached_emb(
        self,
        device: torch.device,
        dtype: torch.dtype,
        h: int,
        w: int,
        scale: float = 1.0,
        base_size: Optional[int] = None,
    ):
        grid_h = torch.arange(h, device=device) / scale
        grid_w = torch.arange(w, device=device) / scale
        if base_size is not None:
            grid_h *= base_size / h
            grid_w *= base_size / w
        grid_h, grid_w = torch.meshgrid(
            grid_w,
            grid_h,
            indexing="ij",
        )  # here w goes first
        grid_h = grid_h.t().reshape(-1)
        grid_w = grid_w.t().reshape(-1)
        emb_h = self._get_sin_cos_emb(grid_h)
        emb_w = self._get_sin_cos_emb(grid_w)
        return torch.concat([emb_h, emb_w], dim=-1).unsqueeze(0).to(dtype)

    def forward(
        self,
        x: torch.Tensor,
        h: int,
        w: int,
        scale: Optional[float] = 1.0,
        base_size: Optional[int] = None,
    ) -> torch.Tensor:
        return self._get_cached_emb(x.device, x.dtype, h, w, scale, base_size)