-
Notifications
You must be signed in to change notification settings - Fork 24
/
swapping_class.py
186 lines (154 loc) · 8.13 KB
/
swapping_class.py
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
from typing import List, Union, Tuple, Dict, Optional
import abc
import torch.nn.functional as nnf
import torch
import utils
from utils import get_refinement_mapper
class LocalBlend:
def get_mask(self, x_t, maps, alpha, use_pool):
k = 1
maps = (maps * alpha).sum(-1).mean(1)
if use_pool:
maps = nnf.max_pool2d(maps, (k * 2 + 1, k * 2 +1), (1, 1), padding=(k, k))
mask = nnf.interpolate(maps, size=(x_t.shape[2:]))
mask = mask / mask.max(2, keepdims=True)[0].max(3, keepdims=True)[0]
mask = mask.gt(self.th[1-int(use_pool)])
mask = mask[:1] + mask
return mask
def __call__(self, x_t, attention_store):
self.counter += 1
if self.counter > self.start_blend:
maps = attention_store["down_cross"][2:4] + attention_store["up_cross"][:3]
maps = [item.reshape(self.alpha_layers.shape[0], -1, 1, 16, 16, 77) for item in maps]
maps = torch.cat(maps, dim=1)
mask = self.get_mask(x_t, maps, self.alpha_layers, True)
if self.substruct_layers is not None:
maps_sub = ~self.get_mask(maps, self.substruct_layers, False)
mask = mask * maps_sub
mask = mask.float()
x_t = x_t[:1] + mask * (x_t - x_t[:1])
return x_t
def __init__(self, prompts: List[str], words: [List[List[str]]], tokenizer, device, NUM_DDIM_STEPS,
substruct_words=None, start_blend=0.2, th=(.3, .3)):
alpha_layers = torch.zeros(len(prompts), 1, 1, 1, 1, 77)
for i, (prompt, words_) in enumerate(zip(prompts, words)):
if type(words_) is str:
words_ = [words_]
for word in words_:
ind = utils.get_word_inds(prompt, word, tokenizer)
alpha_layers[i, :, :, :, :, ind] = 1
if substruct_words is not None:
substruct_layers = torch.zeros(len(prompts), 1, 1, 1, 1, 77)
for i, (prompt, words_) in enumerate(zip(prompts, substruct_words)):
if type(words_) is str:
words_ = [words_]
for word in words_:
ind = utils.get_word_inds(prompt, word, tokenizer)
substruct_layers[i, :, :, :, :, ind] = 1
self.substruct_layers = substruct_layers.to(device)
else:
self.substruct_layers = None
self.alpha_layers = alpha_layers.to(device)
self.start_blend = int(start_blend * NUM_DDIM_STEPS)
self.counter = 0
self.th=th
class AttentionControlEdit(abc.ABC):
@staticmethod
def get_empty_store():
return {"down_cross": [], "mid_cross": [], "up_cross": [],
"down_self": [], "mid_self": [], "up_self": []}
@property
def num_uncond_att_layers(self):
return self.num_att_layers if self.LOW_RESOURCE else 0
def __call__(self, attn, is_cross: bool, place_in_unet: str):
if self.cur_att_layer >= self.num_uncond_att_layers:
if self.LOW_RESOURCE:
attn = self.forward(attn, is_cross, place_in_unet)
else:
h = attn.shape[0]
attn[h // 2:] = self.forward(attn[h // 2:], is_cross, place_in_unet)
self.cur_att_layer += 1
if self.cur_att_layer == self.num_att_layers + self.num_uncond_att_layers:
self.cur_att_layer = 0
self.cur_step += 1
self.between_steps()
return attn
def step_callback(self, x_t):
if self.local_blend is not None:
x_t = self.local_blend(x_t, self.attention_store)
return x_t
def replace_self_attention(self, attn_base, att_replace, place_in_unet):
if att_replace.shape[2] <= 32 ** 2:
attn_base = attn_base.unsqueeze(0).expand(att_replace.shape[0], *attn_base.shape)
return attn_base
else:
return att_replace
def between_steps(self):
if len(self.attention_store) == 0:
self.attention_store = self.step_store
else:
for key in self.attention_store:
for i in range(len(self.attention_store[key])):
self.attention_store[key][i] += self.step_store[key][i]
self.step_store = self.get_empty_store()
def get_average_attention(self):
average_attention = {key: [item / self.cur_step for item in self.attention_store[key]] for key in self.attention_store}
return average_attention
def reset(self):
self.cur_step = 0
self.cur_att_layer = 0
self.step_store = self.get_empty_store()
self.attention_store = {}
@abc.abstractmethod
def replace_cross_attention(self, attn_base, att_replace):
raise NotImplementedError
def forward(self, attn, is_cross: bool, place_in_unet: str):
key = f"{place_in_unet}_{'cross' if is_cross else 'self'}"
if attn.shape[1] <= 32 ** 2:
self.step_store[key].append(attn)
if is_cross or (self.num_self_replace[0] <= self.cur_step < self.num_self_replace[1]):
h = attn.shape[0] // (self.batch_size)
attn = attn.reshape(self.batch_size, h, *attn.shape[1:])
attn_base, attn_repalce = attn[0], attn[1:]
if is_cross:
alpha_words = self.cross_replace_alpha[self.cur_step]
attn_repalce_new = self.replace_cross_attention(attn_base, attn_repalce) * alpha_words + (1 - alpha_words) * attn_repalce
attn[1:] = attn_repalce_new
else:
attn[1:] = self.replace_self_attention(attn_base, attn_repalce, place_in_unet)
attn = attn.reshape(self.batch_size * h, *attn.shape[2:])
return attn
def __init__(self, prompts, num_steps: int,
cross_replace_steps: Union[float, Tuple[float, float], Dict[str, Tuple[float, float]]],
self_replace_steps: Union[float, Tuple[float, float]],
local_blend: Optional[LocalBlend], tokenizer, device):
self.cur_step = 0
self.num_att_layers = -1
self.cur_att_layer = 0
self.step_store = self.get_empty_store()
self.attention_store = {}
self.batch_size = len(prompts)
self.cross_replace_alpha = utils.get_time_words_attention_alpha(prompts, num_steps, cross_replace_steps, tokenizer).to(device)
if type(self_replace_steps) is float:
self_replace_steps = 0, self_replace_steps
self.num_self_replace = int(num_steps * self_replace_steps[0]), int(num_steps * self_replace_steps[1])
self.local_blend = local_blend
class AttentionSwap(AttentionControlEdit):
def replace_cross_attention(self, attn_base, att_replace):
attn_base_replace = attn_base[:, :, self.mapper].permute(2, 0, 1, 3)
attn_replace = attn_base_replace * self.alphas + att_replace * (1 - self.alphas)
# attn_replace = attn_replace / attn_replace.sum(-1, keepdims=True)
return attn_replace
def __init__(self, prompts, num_steps: int, cross_map_replace_steps: float, self_map_replace_steps: float, self_output_replace_steps: float,
source_subject_word=None, target_subject_word=None, tokenizer=None, device=None, LOW_RESOURCE=False):
self_map_replace_steps = self_map_replace_steps + self_output_replace_steps
blend_word = (((source_subject_word,), (target_subject_word,)))
local_blend = LocalBlend(prompts, blend_word, tokenizer, device, num_steps)
super(AttentionSwap, self).__init__(prompts, num_steps, cross_map_replace_steps, self_map_replace_steps, local_blend, tokenizer, device)
self.cross_map_replace_steps = cross_map_replace_steps
self.self_map_replace_steps = self_map_replace_steps
self.self_output_replace_steps = self_output_replace_steps
self.mapper, alphas = get_refinement_mapper(prompts, tokenizer)
self.mapper, alphas = self.mapper.to(device), alphas.to(device)
self.alphas = alphas.reshape(alphas.shape[0], 1, 1, alphas.shape[1])
self.LOW_RESOURCE = LOW_RESOURCE