File size: 15,718 Bytes
95d4bb7 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 |
import math
from dataclasses import dataclass
import torch
from einops import rearrange
from torch import Tensor, nn
from flux.math import attention, rope,apply_rope
import os
class EmbedND(nn.Module):
def __init__(self, dim: int, theta: int, axes_dim: list[int]):
super().__init__()
self.dim = dim
self.theta = theta
self.axes_dim = axes_dim
def forward(self, ids: Tensor) -> Tensor:
n_axes = ids.shape[-1]
emb = torch.cat(
[rope(ids[..., i], self.axes_dim[i], self.theta) for i in range(n_axes)], # [1, 1, 4592, 8, 2, 2]) [1, 1, 4592, 28, 2, 2]) [1, 1, 4592, 28, 2, 2])
dim=-3,
)
return emb.unsqueeze(1)
def timestep_embedding(t: Tensor, dim, max_period=10000, time_factor: float = 1000.0):
"""
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.
"""
t = time_factor * t
half = dim // 2
freqs = torch.exp(-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half).to(
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)
if torch.is_floating_point(t):
embedding = embedding.to(t)
return embedding
class MLPEmbedder(nn.Module):
def __init__(self, in_dim: int, hidden_dim: int):
super().__init__()
self.in_layer = nn.Linear(in_dim, hidden_dim, bias=True)
self.silu = nn.SiLU()
self.out_layer = nn.Linear(hidden_dim, hidden_dim, bias=True)
def forward(self, x: Tensor) -> Tensor:
return self.out_layer(self.silu(self.in_layer(x)))
class RMSNorm(torch.nn.Module):
def __init__(self, dim: int):
super().__init__()
self.scale = nn.Parameter(torch.ones(dim))
def forward(self, x: Tensor):
x_dtype = x.dtype
x = x.float()
rrms = torch.rsqrt(torch.mean(x**2, dim=-1, keepdim=True) + 1e-6)
return (x * rrms).to(dtype=x_dtype) * self.scale
class QKNorm(torch.nn.Module):
def __init__(self, dim: int):
super().__init__()
self.query_norm = RMSNorm(dim)
self.key_norm = RMSNorm(dim)
def forward(self, q: Tensor, k: Tensor, v: Tensor) -> tuple[Tensor, Tensor]:
q = self.query_norm(q)
k = self.key_norm(k)
return q.to(v), k.to(v)
class SelfAttention(nn.Module):
def __init__(self, dim: int, num_heads: int = 8, qkv_bias: bool = False):
super().__init__()
self.num_heads = num_heads
head_dim = dim // num_heads
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
self.norm = QKNorm(head_dim)
self.proj = nn.Linear(dim, dim)
def forward(self, x: Tensor, pe: Tensor) -> Tensor:
qkv = self.qkv(x)
q, k, v = rearrange(qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
q, k = self.norm(q, k, v)
x = attention(q, k, v, pe=pe)
x = self.proj(x)
return x
@dataclass
class ModulationOut:
shift: Tensor
scale: Tensor
gate: Tensor
class Modulation(nn.Module):
def __init__(self, dim: int, double: bool):
super().__init__()
self.is_double = double
self.multiplier = 6 if double else 3
self.lin = nn.Linear(dim, self.multiplier * dim, bias=True)
def forward(self, vec: Tensor) -> tuple[ModulationOut, ModulationOut | None]:
out = self.lin(nn.functional.silu(vec))[:, None, :].chunk(self.multiplier, dim=-1)
return (
ModulationOut(*out[:3]),
ModulationOut(*out[3:]) if self.is_double else None,
)
class DoubleStreamBlock(nn.Module):
def __init__(self, hidden_size: int, num_heads: int, mlp_ratio: float, qkv_bias: bool = False):
super().__init__()
mlp_hidden_dim = int(hidden_size * mlp_ratio)
self.num_heads = num_heads
self.hidden_size = hidden_size
self.img_mod = Modulation(hidden_size, double=True)
self.img_norm1 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.img_attn = SelfAttention(dim=hidden_size, num_heads=num_heads, qkv_bias=qkv_bias)
self.img_norm2 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.img_mlp = nn.Sequential(
nn.Linear(hidden_size, mlp_hidden_dim, bias=True),
nn.GELU(approximate="tanh"),
nn.Linear(mlp_hidden_dim, hidden_size, bias=True),
)
self.txt_mod = Modulation(hidden_size, double=True)
self.txt_norm1 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.txt_attn = SelfAttention(dim=hidden_size, num_heads=num_heads, qkv_bias=qkv_bias)
self.txt_norm2 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.txt_mlp = nn.Sequential(
nn.Linear(hidden_size, mlp_hidden_dim, bias=True),
nn.GELU(approximate="tanh"),
nn.Linear(mlp_hidden_dim, hidden_size, bias=True),
)
def forward(self, img: Tensor, txt: Tensor, vec: Tensor, pe: Tensor) -> tuple[Tensor, Tensor]:
img_mod1, img_mod2 = self.img_mod(vec)
txt_mod1, txt_mod2 = self.txt_mod(vec)
# prepare image for attention
img_modulated = self.img_norm1(img)
img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift
img_qkv = self.img_attn.qkv(img_modulated)
img_q, img_k, img_v = rearrange(img_qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads) # torch.Size([1, 24, 4080, 128])
img_q, img_k = self.img_attn.norm(img_q, img_k, img_v)
# prepare txt for attention
txt_modulated = self.txt_norm1(txt)
txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift
txt_qkv = self.txt_attn.qkv(txt_modulated)
txt_q, txt_k, txt_v = rearrange(txt_qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
txt_q, txt_k = self.txt_attn.norm(txt_q, txt_k, txt_v)
# run actual attention
q = torch.cat((txt_q, img_q), dim=2) # [8, 24, 512, 128] + [8, 24, 900, 128] -> [8, 24, 1412, 128]
k = torch.cat((txt_k, img_k), dim=2)
v = torch.cat((txt_v, img_v), dim=2)
# import pdb;pdb.set_trace()
attn = attention(q, k, v, pe=pe)
txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1] :]
# calculate the img bloks
img = img + img_mod1.gate * self.img_attn.proj(img_attn)
img = img + img_mod2.gate * self.img_mlp((1 + img_mod2.scale) * self.img_norm2(img) + img_mod2.shift)
# calculate the txt bloks
txt = txt + txt_mod1.gate * self.txt_attn.proj(txt_attn)
txt = txt + txt_mod2.gate * self.txt_mlp((1 + txt_mod2.scale) * self.txt_norm2(txt) + txt_mod2.shift)
return img, txt
class SingleStreamBlock(nn.Module):
"""
A DiT block with parallel linear layers as described in
https://arxiv.org/abs/2302.05442 and adapted modulation interface.
"""
def __init__(
self,
hidden_size: int,
num_heads: int,
mlp_ratio: float = 4.0,
qk_scale: float | None = None,
):
super().__init__()
self.hidden_dim = hidden_size
self.num_heads = num_heads
head_dim = hidden_size // num_heads
self.scale = qk_scale or head_dim**-0.5
self.mlp_hidden_dim = int(hidden_size * mlp_ratio)
# qkv and mlp_in
self.linear1 = nn.Linear(hidden_size, hidden_size * 3 + self.mlp_hidden_dim)
# proj and mlp_out
self.linear2 = nn.Linear(hidden_size + self.mlp_hidden_dim, hidden_size)
self.norm = QKNorm(head_dim)
self.hidden_size = hidden_size
self.pre_norm = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.mlp_act = nn.GELU(approximate="tanh")
self.modulation = Modulation(hidden_size, double=False)
def forward(self, x: Tensor, vec: Tensor, pe: Tensor) -> Tensor:
mod, _ = self.modulation(vec)
x_mod = (1 + mod.scale) * self.pre_norm(x) + mod.shift
qkv, mlp = torch.split(self.linear1(x_mod), [3 * self.hidden_size, self.mlp_hidden_dim], dim=-1)
q, k, v = rearrange(qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
q, k = self.norm(q, k, v)
# compute attention
attn = attention(q, k, v, pe=pe)
# compute activation in mlp stream, cat again and run second linear layer
output = self.linear2(torch.cat((attn, self.mlp_act(mlp)), 2))
return x + mod.gate * output
class LastLayer(nn.Module):
def __init__(self, hidden_size: int, patch_size: int, out_channels: int):
super().__init__()
self.norm_final = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.linear = nn.Linear(hidden_size, patch_size * patch_size * out_channels, bias=True)
self.adaLN_modulation = nn.Sequential(nn.SiLU(), nn.Linear(hidden_size, 2 * hidden_size, bias=True))
def forward(self, x: Tensor, vec: Tensor) -> Tensor:
shift, scale = self.adaLN_modulation(vec).chunk(2, dim=1)
x = (1 + scale[:, None, :]) * self.norm_final(x) + shift[:, None, :]
x = self.linear(x)
return x
class LastLayer(nn.Module):
def __init__(self, hidden_size: int, patch_size: int, out_channels: int):
super().__init__()
self.norm_final = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
self.linear = nn.Linear(hidden_size, patch_size * patch_size * out_channels, bias=True)
self.adaLN_modulation = nn.Sequential(nn.SiLU(), nn.Linear(hidden_size, 2 * hidden_size, bias=True))
def forward(self, x: Tensor, vec: Tensor) -> Tensor:
shift, scale = self.adaLN_modulation(vec).chunk(2, dim=1)
x = (1 + scale[:, None, :]) * self.norm_final(x) + shift[:, None, :]
x = self.linear(x)
return x
class DoubleStreamBlock_kv(DoubleStreamBlock):
def __init__(self, hidden_size: int, num_heads: int, mlp_ratio: float, qkv_bias: bool = False):
super().__init__(hidden_size, num_heads, mlp_ratio, qkv_bias)
def forward(self, img: Tensor, txt: Tensor, vec: Tensor, pe: Tensor, info) -> tuple[Tensor, Tensor]:
img_mod1, img_mod2 = self.img_mod(vec)
txt_mod1, txt_mod2 = self.txt_mod(vec)
# prepare image for attention
img_modulated = self.img_norm1(img)
img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift
img_qkv = self.img_attn.qkv(img_modulated)
img_q, img_k, img_v = rearrange(img_qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads) # torch.Size([1, 24, 4080, 128])
img_q, img_k = self.img_attn.norm(img_q, img_k, img_v)
# prepare txt for attention
txt_modulated = self.txt_norm1(txt)
txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift
txt_qkv = self.txt_attn.qkv(txt_modulated)
txt_q, txt_k, txt_v = rearrange(txt_qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
txt_q, txt_k = self.txt_attn.norm(txt_q, txt_k, txt_v)
feature_k_name = str(info['t']) + '_' + str(info['id']) + '_' + 'MB' + '_' + 'K'
feature_v_name = str(info['t']) + '_' + str(info['id']) + '_' + 'MB' + '_' + 'V'
if info['inverse']:
info['feature'][feature_k_name] = img_k.cpu()
info['feature'][feature_v_name] = img_v.cpu()
q = torch.cat((txt_q, img_q), dim=2) # [B, 24, 512, 128] + [B, 24, 900, 128] -> [B, 24, 1412, 128]
k = torch.cat((txt_k, img_k), dim=2)
v = torch.cat((txt_v, img_v), dim=2)
if 'attention_mask' in info:
attn = attention(q, k, v, pe=pe,attention_mask=info['attention_mask'])
else:
attn = attention(q, k, v, pe=pe)
# elif feature_k_name in info['feature']:
else:
source_img_k = info['feature'][feature_k_name].to(img.device)
source_img_v = info['feature'][feature_v_name].to(img.device)
mask_indices = info['mask_indices'] # 图片seq坐标下的
source_img_k[:, :, mask_indices, ...] = img_k
source_img_v[:, :, mask_indices, ...] = img_v
q = torch.cat((txt_q, img_q), dim=2)
k = torch.cat((txt_k, source_img_k), dim=2)
v = torch.cat((txt_v, source_img_v), dim=2)
attn = attention(q, k, v, pe=pe, pe_q = info['pe_mask'])
txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1] :]
# calculate the img bloks
img = img + img_mod1.gate * self.img_attn.proj(img_attn)
img = img + img_mod2.gate * self.img_mlp((1 + img_mod2.scale) * self.img_norm2(img) + img_mod2.shift)
# calculate the txt bloks
txt = txt + txt_mod1.gate * self.txt_attn.proj(txt_attn)
txt = txt + txt_mod2.gate * self.txt_mlp((1 + txt_mod2.scale) * self.txt_norm2(txt) + txt_mod2.shift)
return img, txt
class SingleStreamBlock_kv(SingleStreamBlock):
"""
A DiT block with parallel linear layers as described in
https://arxiv.org/abs/2302.05442 and adapted modulation interface.
"""
def __init__(
self,
hidden_size: int,
num_heads: int,
mlp_ratio: float = 4.0,
qk_scale: float | None = None,
):
super().__init__(hidden_size, num_heads, mlp_ratio, qk_scale)
def forward(self,x: Tensor, vec: Tensor, pe: Tensor, info) -> Tensor:
mod, _ = self.modulation(vec)
x_mod = (1 + mod.scale) * self.pre_norm(x) + mod.shift
qkv, mlp = torch.split(self.linear1(x_mod), [3 * self.hidden_size, self.mlp_hidden_dim], dim=-1)
q, k, v = rearrange(qkv, "B L (K H D) -> K B H L D", K=3, H=self.num_heads)
q, k = self.norm(q, k, v)
img_k = k[:, :, 512:, ...]
img_v = v[:, :, 512:, ...]
txt_k = k[:, :, :512, ...]
txt_v = v[:, :, :512, ...]
feature_k_name = str(info['t']) + '_' + str(info['id']) + '_' + 'SB' + '_' + 'K'
feature_v_name = str(info['t']) + '_' + str(info['id']) + '_' + 'SB' + '_' + 'V'
if info['inverse']:
info['feature'][feature_k_name] = img_k.cpu()
info['feature'][feature_v_name] = img_v.cpu()
if 'attention_mask' in info:
attn = attention(q, k, v, pe=pe,attention_mask=info['attention_mask'])
else:
attn = attention(q, k, v, pe=pe)
else:
source_img_k = info['feature'][feature_k_name].to(x.device)
source_img_v = info['feature'][feature_v_name].to(x.device)
mask_indices = info['mask_indices'] # 图片seq坐标下的
source_img_k[:, :, mask_indices, ...] = img_k
source_img_v[:, :, mask_indices, ...] = img_v
k = torch.cat((txt_k, source_img_k), dim=2)
v = torch.cat((txt_v, source_img_v), dim=2)
attn = attention(q, k, v, pe=pe, pe_q = info['pe_mask'])
# compute attention
# attn = attention(q, k, v, pe=pe)
# compute activation in mlp stream, cat again and run second linear layer
output = self.linear2(torch.cat((attn, self.mlp_act(mlp)), 2))
return x + mod.gate * output |