Spaces:
Sleeping
Sleeping
File size: 13,437 Bytes
1bf3c29 c85ad6e 1bf3c29 c85ad6e 1bf3c29 c85ad6e 1bf3c29 c85ad6e 1bf3c29 c85ad6e 1bf3c29 c85ad6e 1bf3c29 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 | import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.utils.checkpoint import checkpoint
from typing import Tuple
from .pixeldit_c2i import PatchTokenEmbedder, PixelTokenEmbedder, PiTBlock
from .modules import (
FinalLayer,
FeedForward,
PixelDiTJointAttnProcessor,
RMSNorm,
TimestepConditioner,
apply_adaln,
apply_rotary_emb,
precompute_freqs_cis_2d,
)
class MMDiTJointAttention(nn.Module):
def __init__(
self,
dim: int,
num_heads: int = 8,
qkv_bias: bool = False,
attn_drop: float = 0.,
proj_drop: float = 0.,
) -> 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.qkv_x = nn.Linear(dim, dim * 3, bias=qkv_bias)
self.qkv_y = nn.Linear(dim, dim * 3, bias=qkv_bias)
self.q_norm_x = RMSNorm(self.head_dim)
self.k_norm_x = RMSNorm(self.head_dim)
self.q_norm_y = RMSNorm(self.head_dim)
self.k_norm_y = RMSNorm(self.head_dim)
self.proj_x = nn.Linear(dim, dim)
self.proj_y = nn.Linear(dim, dim)
self.attn_drop = nn.Dropout(attn_drop)
self.proj_drop_x = nn.Dropout(proj_drop)
self.proj_drop_y = nn.Dropout(proj_drop)
self.processor = PixelDiTJointAttnProcessor()
def set_processor(self, processor) -> None:
self.processor = processor
def get_processor(self):
return self.processor
def forward(
self,
x: torch.Tensor,
y: torch.Tensor,
pos_img: torch.Tensor,
pos_txt: torch.Tensor = None,
attn_mask: torch.Tensor = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
B, Nx, C = x.shape
By, Ny, Cy = y.shape
assert B == By and C == Cy, "x and y must share batch and channel dims"
qkv_x = self.qkv_x(x).reshape(B, Nx, 3, self.num_heads, C // self.num_heads).permute(2, 0, 1, 3, 4)
qx, kx, vx = qkv_x[0], qkv_x[1], qkv_x[2]
qx = self.q_norm_x(qx)
kx = self.k_norm_x(kx)
qkv_y = self.qkv_y(y).reshape(B, Ny, 3, self.num_heads, C // self.num_heads).permute(2, 0, 1, 3, 4)
qy, ky, vy = qkv_y[0], qkv_y[1], qkv_y[2]
qy = self.q_norm_y(qy)
ky = self.k_norm_y(ky)
qx, kx = apply_rotary_emb(qx, kx, freqs_cis=pos_img)
if pos_txt is not None:
qy, ky = apply_rotary_emb(qy, ky, freqs_cis=pos_txt)
qx = qx.transpose(1, 2)
kx = kx.transpose(1, 2)
vx = vx.transpose(1, 2)
qy = qy.transpose(1, 2)
ky = ky.transpose(1, 2)
vy = vy.transpose(1, 2)
q_joint = torch.cat([qy, qx], dim=2)
k_joint = torch.cat([ky, kx], dim=2)
v_joint = torch.cat([vy, vx], dim=2)
out_joint = self.processor(self, q_joint, k_joint, v_joint, attn_mask=attn_mask)
out_y = out_joint[:, :, :Ny, :]
out_x = out_joint[:, :, Ny:, :]
out_y = out_y.transpose(1, 2).reshape(B, Ny, C)
out_x = out_x.transpose(1, 2).reshape(B, Nx, C)
out_x = self.proj_drop_x(self.proj_x(out_x))
out_y = self.proj_drop_y(self.proj_y(out_y))
return out_x, out_y
class MMDiTBlockT2I(nn.Module):
def __init__(self, hidden_size, groups, mlp_ratio=4.0, adaLN_modulation_img=None, adaLN_modulation_txt=None):
super().__init__()
self.hidden_size = hidden_size
self.groups = groups
self.head_dim = hidden_size // groups
self.norm_x1 = RMSNorm(hidden_size, eps=1e-6)
self.norm_y1 = RMSNorm(hidden_size, eps=1e-6)
self.attn = MMDiTJointAttention(hidden_size, num_heads=groups, qkv_bias=False)
self.norm_x2 = RMSNorm(hidden_size, eps=1e-6)
self.norm_y2 = RMSNorm(hidden_size, eps=1e-6)
mlp_hidden_dim = int(hidden_size * mlp_ratio)
self.mlp_x = FeedForward(hidden_size, mlp_hidden_dim)
self.mlp_y = FeedForward(hidden_size, mlp_hidden_dim)
self.adaLN_modulation_img = adaLN_modulation_img if adaLN_modulation_img is not None else nn.Sequential(nn.Linear(hidden_size, 6 * hidden_size, bias=True))
self.adaLN_modulation_txt = adaLN_modulation_txt if adaLN_modulation_txt is not None else nn.Sequential(nn.Linear(hidden_size, 6 * hidden_size, bias=True))
def forward(self, x, y, c, pos_img, pos_txt=None, attn_mask=None):
shift_msa_x, scale_msa_x, gate_msa_x, shift_mlp_x, scale_mlp_x, gate_mlp_x = self.adaLN_modulation_img(c).chunk(6, dim=-1)
shift_msa_y, scale_msa_y, gate_msa_y, shift_mlp_y, scale_mlp_y, gate_mlp_y = self.adaLN_modulation_txt(c).chunk(6, dim=-1)
x_norm = apply_adaln(self.norm_x1(x), shift_msa_x, scale_msa_x)
y_norm = apply_adaln(self.norm_y1(y), shift_msa_y, scale_msa_y)
attn_x, attn_y = self.attn(x_norm, y_norm, pos_img, pos_txt, attn_mask)
x = x + gate_msa_x * attn_x
y = y + gate_msa_y * attn_y
x = x + gate_mlp_x * self.mlp_x(apply_adaln(self.norm_x2(x), shift_mlp_x, scale_mlp_x))
y = y + gate_mlp_y * self.mlp_y(apply_adaln(self.norm_y2(y), shift_mlp_y, scale_mlp_y))
return x, y
class PixDiT_T2I(nn.Module):
def __init__(
self,
in_channels=3,
num_groups=16,
hidden_size=1152,
pixel_hidden_size=64,
pixel_attn_hidden_size=None,
pixel_num_groups=None,
patch_depth=26,
pixel_depth=2,
num_text_blocks=4,
patch_size=16,
txt_embed_dim=4096,
txt_max_length=1024,
use_text_rope: bool = True,
text_rope_theta: float = 10000.0,
repa_encoder_index: int = -1,
use_pixel_abs_pos: bool = True,
):
super().__init__()
self.in_channels = int(in_channels)
self.out_channels = int(in_channels)
self.hidden_size = int(hidden_size)
self.num_groups = int(num_groups)
self.patch_depth = int(patch_depth)
self.pixel_depth = int(pixel_depth)
self.num_text_blocks = int(num_text_blocks)
self.patch_size = int(patch_size)
self.pixel_hidden_size = int(pixel_hidden_size)
self.txt_embed_dim = int(txt_embed_dim)
self.txt_max_length = int(txt_max_length)
self.use_text_rope = bool(use_text_rope)
self.text_rope_theta = float(text_rope_theta)
self.repa_encoder_index = int(repa_encoder_index)
self.use_pixel_abs_pos = bool(use_pixel_abs_pos)
if self.pixel_depth <= 0:
raise ValueError("PixDiT_T2I expects pixel_depth > 0 to retain the pixel pathway")
self.pixel_embedder = PixelTokenEmbedder(in_channels, self.pixel_hidden_size, use_pixel_abs_pos=self.use_pixel_abs_pos)
self.s_embedder = PatchTokenEmbedder(in_channels * patch_size ** 2, hidden_size, bias=True)
self.t_embedder = TimestepConditioner(hidden_size)
self.y_embedder = PatchTokenEmbedder(self.txt_embed_dim, hidden_size, bias=True, norm_layer=RMSNorm)
self.y_pos_embedding = nn.Parameter(torch.randn(1, self.txt_max_length, hidden_size))
self._shared_cond_adaln = None
self._shared_cond_adaln_img = None
self._shared_cond_adaln_txt = None
self.patch_blocks = nn.ModuleList([
MMDiTBlockT2I(
self.hidden_size,
self.num_groups,
adaLN_modulation_img=self._shared_cond_adaln_img,
adaLN_modulation_txt=self._shared_cond_adaln_txt,
)
for _ in range(self.patch_depth)
])
self.text_refine_blocks = None
self.pixel_attn_hidden_size = (
int(pixel_attn_hidden_size) if pixel_attn_hidden_size is not None else self.hidden_size
)
self.pixel_num_groups = int(pixel_num_groups) if pixel_num_groups is not None else self.num_groups
self.pixel_blocks = nn.ModuleList(
[
PiTBlock(
self.pixel_hidden_size,
self.hidden_size,
patch_size=self.patch_size,
num_heads=self.num_groups,
mlp_ratio=4.0,
attn_hidden_size=self.pixel_attn_hidden_size,
attn_num_heads=self.pixel_num_groups,
rope_fn=precompute_freqs_cis_2d,
)
for _ in range(self.pixel_depth)
]
)
self.final_layer = FinalLayer(self.pixel_hidden_size, self.out_channels)
self.precompute_pos = dict()
self.precompute_pos_txt = dict()
self.last_repa_tokens = None
self.gradient_checkpointing = False
self.initialize_weights()
def fetch_pos(self, height, width, device):
if (height, width) in self.precompute_pos:
return self.precompute_pos[(height, width)].to(device)
else:
pos = precompute_freqs_cis_2d(self.hidden_size // self.num_groups, height, width).to(device)
self.precompute_pos[(height, width)] = pos
return pos
def fetch_pos_text(self, length, device):
if length in self.precompute_pos_txt:
return self.precompute_pos_txt[length].to(device)
head_dim = self.hidden_size // self.num_groups
freqs = 1.0 / (self.text_rope_theta ** (torch.arange(0, head_dim, 2, device=device).float() / head_dim))
positions = torch.arange(0, length, device=device).float().unsqueeze(1)
angles = positions * freqs.unsqueeze(0)
freqs_cis = torch.polar(torch.ones_like(angles), angles)
self.precompute_pos_txt[length] = freqs_cis
return freqs_cis
def initialize_weights(self):
w = self.s_embedder.proj.weight.data
nn.init.xavier_uniform_(w.view([w.shape[0], -1]))
nn.init.constant_(self.s_embedder.proj.bias, 0)
nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
nn.init.normal_(self.t_embedder.mlp[2].weight, std=0.02)
nn.init.zeros_(self.final_layer.linear.weight)
nn.init.zeros_(self.final_layer.linear.bias)
def forward(self, x, t, y, s=None, mask=None):
B, _, H, W = x.shape
Hs = H // self.patch_size
Ws = W // self.patch_size
L = Hs * Ws
pos = self.fetch_pos(Hs, Ws, x.device)
x_patches = torch.nn.functional.unfold(x, kernel_size=self.patch_size, stride=self.patch_size).transpose(1, 2)
t_emb = self.t_embedder(t.view(-1)).view(B, -1, self.hidden_size)
if y.dim() != 3:
raise ValueError("Text embedding y must be [B, L, D]")
Ltxt = min(y.shape[1], self.txt_max_length)
y = y[:, :Ltxt, :]
y_emb = self.y_embedder(y).view(B, Ltxt, self.hidden_size)
y_emb = y_emb + self.y_pos_embedding[:, :Ltxt, :].to(y_emb.dtype)
condition = torch.nn.functional.silu(t_emb)
if s is None:
s0 = self.s_embedder(x_patches)
pos_txt = self.fetch_pos_text(Ltxt, x.device) if self.use_text_rope else None
attn_mask_joint = None
if mask is not None and isinstance(mask, torch.Tensor):
m = mask
while m.dim() > 2 and m.size(1) == 1:
m = m.squeeze(1)
if m.dim() == 3 and m.size(1) == 1:
m = m.squeeze(1)
if m.dim() == 2:
pad = (m == 0)
pad_img = torch.zeros((B, L), dtype=torch.bool, device=x.device)
attn_mask_joint = torch.cat([pad[:, :Ltxt], pad_img], dim=1).view(B, 1, 1, Ltxt + L)
self.last_repa_tokens = None
s = s0
for i in range(self.patch_depth):
if self.gradient_checkpointing and self.training:
s, y_emb = checkpoint(
self.patch_blocks[i],
s, y_emb, condition, pos, pos_txt, attn_mask_joint,
use_reentrant=False,
)
else:
s, y_emb = self.patch_blocks[i](s, y_emb, condition, pos, pos_txt, attn_mask_joint)
if 0 < self.repa_encoder_index == (i + 1):
self.last_repa_tokens = s
s = torch.nn.functional.silu(t_emb + s)
if not (0 < self.repa_encoder_index <= self.patch_depth):
self.last_repa_tokens = s
batch_size, length, _ = s.shape
if length != L:
if length > L:
s = s[:, :L, :]
else:
pad_len = L - length
s = torch.cat([s, s.new_zeros(B, pad_len, s.shape[2])], dim=1)
length = L
s_cond = s.view(B * L, self.hidden_size)
x_pixels = self.pixel_embedder(x, img_height=H, img_width=W, patch_size=self.patch_size)
for blk in self.pixel_blocks:
x_pixels = blk(x_pixels, s_cond, H, W, self.patch_size, mask)
x_pixels = self.final_layer(x_pixels)
C_out = self.out_channels
P2 = self.patch_size * self.patch_size
x_pixels = x_pixels.view(B, L, P2, C_out).permute(0, 3, 2, 1).contiguous()
x_pixels = x_pixels.view(B, C_out * P2, L)
x_img = torch.nn.functional.fold(x_pixels, (H, W), kernel_size=self.patch_size, stride=self.patch_size)
return x_img
|