File size: 8,906 Bytes
58b87c6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 | """mDiffAE v2 decoder: skip-concat topology with dual PDG (token masking + path drop).
No outer RMSNorms (use_other_outer_rms_norms=False during training):
norm_in, latent_norm, and norm_out are all absent.
"""
from __future__ import annotations
import math
import torch
from torch import Tensor, nn
from .adaln import AdaLNZeroLowRankDelta, AdaLNZeroProjector
from .dico_block import DiCoBlock
from .straight_through_encoder import Patchify
from .time_embed import SinusoidalTimeEmbeddingMLP
class Decoder(nn.Module):
"""VP diffusion decoder conditioned on encoder latents and timestep.
Architecture (skip-concat, 2+4+2 default):
Patchify x_t -> Fuse with upsampled z
-> Start blocks (2) -> Middle blocks (4) -> Skip fuse -> End blocks (2)
-> Conv1x1 -> PixelShuffle
Dual PDG at inference:
- Path drop: replace middle block output with ``path_drop_mask_feature``.
- Token mask: replace a fraction of upsampled latent tokens with
``latent_mask_feature`` before fusion.
"""
def __init__(
self,
in_channels: int,
patch_size: int,
model_dim: int,
depth: int,
start_block_count: int,
end_block_count: int,
bottleneck_dim: int,
mlp_ratio: float,
depthwise_kernel_size: int,
adaln_low_rank_rank: int,
pdg_mask_ratio: float = 0.75,
) -> None:
super().__init__()
self.patch_size = int(patch_size)
self.model_dim = int(model_dim)
self.pdg_mask_ratio = float(pdg_mask_ratio)
# Input processing (no norm_in — use_other_outer_rms_norms=False)
self.patchify = Patchify(in_channels, patch_size, model_dim)
# Latent conditioning path (no latent_norm)
self.latent_up = nn.Conv2d(bottleneck_dim, model_dim, kernel_size=1, bias=True)
self.fuse_in = nn.Conv2d(2 * model_dim, model_dim, kernel_size=1, bias=True)
# Time embedding
self.time_embed = SinusoidalTimeEmbeddingMLP(model_dim)
# AdaLN: shared base projector + per-block low-rank deltas
self.adaln_base = AdaLNZeroProjector(d_model=model_dim, d_cond=model_dim)
self.adaln_deltas = nn.ModuleList(
[
AdaLNZeroLowRankDelta(
d_model=model_dim, d_cond=model_dim, rank=adaln_low_rank_rank
)
for _ in range(depth)
]
)
# Block layout: start + middle + end
middle_count = depth - start_block_count - end_block_count
self._middle_start_idx = start_block_count
self._end_start_idx = start_block_count + middle_count
def _make_blocks(count: int) -> nn.ModuleList:
return nn.ModuleList(
[
DiCoBlock(
model_dim,
mlp_ratio,
depthwise_kernel_size=depthwise_kernel_size,
use_external_adaln=True,
)
for _ in range(count)
]
)
self.start_blocks = _make_blocks(start_block_count)
self.middle_blocks = _make_blocks(middle_count)
self.fuse_skip = nn.Conv2d(2 * model_dim, model_dim, kernel_size=1, bias=True)
self.end_blocks = _make_blocks(end_block_count)
# Learned mask features for dual PDG
self.latent_mask_feature = nn.Parameter(torch.zeros((1, model_dim, 1, 1)))
self.path_drop_mask_feature = nn.Parameter(torch.zeros((1, model_dim, 1, 1)))
# Output head (no norm_out — use_other_outer_rms_norms=False)
self.out_proj = nn.Conv2d(
model_dim, in_channels * (patch_size**2), kernel_size=1, bias=True
)
self.unpatchify = nn.PixelShuffle(patch_size)
def _adaln_m_for_layer(self, cond: Tensor, layer_idx: int) -> Tensor:
"""Compute packed AdaLN modulation = shared_base + per-layer delta."""
act = self.adaln_base.act(cond)
base_m = self.adaln_base.forward_activated(act)
delta_m = self.adaln_deltas[layer_idx](act)
return base_m + delta_m
def _run_blocks(
self, blocks: nn.ModuleList, x: Tensor, cond: Tensor, start_index: int
) -> Tensor:
"""Run a group of decoder blocks with per-block AdaLN modulation."""
for local_idx, block in enumerate(blocks):
adaln_m = self._adaln_m_for_layer(cond, layer_idx=start_index + local_idx)
x = block(x, adaln_m=adaln_m)
return x
def _apply_latent_token_mask(self, z_up: Tensor) -> Tensor:
"""Replace a fraction of upsampled latent tokens with latent_mask_feature.
Uses 2x2 groupwise masking: divides the spatial grid into 2x2 groups
and masks floor(ratio * 4) tokens per group (lowest random scores).
Args:
z_up: [B, C, H, W] upsampled latent conditioning.
Returns:
Masked tensor with same shape.
"""
b, c, h, w = z_up.shape
# Pad to even dims if needed
h_pad = (2 - h % 2) % 2
w_pad = (2 - w % 2) % 2
if h_pad > 0 or w_pad > 0:
z_up = torch.nn.functional.pad(z_up, (0, w_pad, 0, h_pad))
_, _, h, w = z_up.shape
# Reshape into 2x2 groups: [B, C, H/2, 2, W/2, 2] -> [B, C, H/2, W/2, 4]
x = z_up.reshape(b, c, h // 2, 2, w // 2, 2)
x = x.permute(0, 1, 2, 4, 3, 5).reshape(b, c, h // 2, w // 2, 4)
# Random scores for each token in each group
scores = torch.rand(b, 1, h // 2, w // 2, 4, device=z_up.device)
# Mask the floor(ratio * 4) lowest-scoring tokens per group
num_mask = math.floor(self.pdg_mask_ratio * 4)
if num_mask > 0:
_, indices = scores.sort(dim=-1)
mask = torch.zeros_like(scores, dtype=torch.bool)
mask.scatter_(-1, indices[..., :num_mask], True)
else:
mask = torch.zeros_like(scores, dtype=torch.bool)
# Apply mask: replace masked tokens with latent_mask_feature
mask_feat = self.latent_mask_feature.to(device=z_up.device, dtype=z_up.dtype)
mask_feat = mask_feat.squeeze(-1).squeeze(-1) # [1, C]
mask_feat = mask_feat.view(1, c, 1, 1, 1).expand_as(x)
mask_expanded = mask.expand_as(x)
x = torch.where(mask_expanded, mask_feat, x)
# Reshape back to [B, C, H, W]
x = x.reshape(b, c, h // 2, w // 2, 2, 2)
x = x.permute(0, 1, 2, 4, 3, 5).reshape(b, c, h, w)
# Remove padding
if h_pad > 0 or w_pad > 0:
x = x[:, :, : h - h_pad, : w - w_pad]
return x
def forward(
self,
x_t: Tensor,
t: Tensor,
latents: Tensor,
*,
drop_middle_blocks: bool = False,
mask_latent_tokens: bool = False,
) -> Tensor:
"""Single decoder forward pass.
Args:
x_t: Noised image [B, C, H, W].
t: Timestep [B] in [0, 1].
latents: Encoder latents [B, bottleneck_dim, h, w].
drop_middle_blocks: If True, replace middle block output with
path_drop_mask_feature (for path-drop PDG).
mask_latent_tokens: If True, mask a fraction of upsampled latent
tokens with latent_mask_feature (for token-mask PDG).
Returns:
x0 prediction [B, C, H, W].
"""
# Patchify x_t (no norm_in)
x_feat = self.patchify(x_t)
# Upsample latents (no latent_norm)
z_up = self.latent_up(latents)
# Token masking for PDG (replaces latent tokens with latent_mask_feature)
if mask_latent_tokens:
z_up = self._apply_latent_token_mask(z_up)
# Fuse x_feat and z_up
fused = torch.cat([x_feat, z_up], dim=1)
fused = self.fuse_in(fused)
# Time conditioning
cond = self.time_embed(t.to(torch.float32).to(device=x_t.device))
# Start blocks
start_out = self._run_blocks(self.start_blocks, fused, cond, start_index=0)
# Middle blocks (or path_drop_mask_feature for PDG)
if drop_middle_blocks:
middle_out = self.path_drop_mask_feature.to(
device=x_t.device, dtype=x_t.dtype
).expand_as(start_out)
else:
middle_out = self._run_blocks(
self.middle_blocks,
start_out,
cond,
start_index=self._middle_start_idx,
)
# Skip fusion
skip_fused = torch.cat([start_out, middle_out], dim=1)
skip_fused = self.fuse_skip(skip_fused)
# End blocks
end_out = self._run_blocks(
self.end_blocks, skip_fused, cond, start_index=self._end_start_idx
)
# Output head (no norm_out)
patches = self.out_proj(end_out)
return self.unpatchify(patches)
|