-
Notifications
You must be signed in to change notification settings - Fork 1
/
wgnbase_embed-Copy1.py
220 lines (167 loc) · 7.61 KB
/
wgnbase_embed-Copy1.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
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
import torch
from torch import nn
from torchvision import models
from einops import rearrange, repeat
from einops.layers.torch import Rearrange
from timm.models import create_model
import math
# helpers
def pair(t):
return t if isinstance(t, tuple) else (t, t)
# classes
# class ResModel(nn.Module):
# def __init__(self):
# super(ResModel, self).__init__()
# self.premodel = models.resnet18(pretrained=True)
# self.model = nn.Sequential(*list(self.premodel.children())[:-1]) ###[:-2]
# out_chann = 512 ### 2048x4x4
# print ("resnet18 model loaded")
# self.compress = nn.Linear(out_chann, 768, bias=True) ## False
# def forward(self, x_input):
# x = self.model(x_input) ##### 25 512 7 7
# # print ('000:',x.shape)
# x_comp = self.compress(x) ### 25 2048
# return x_comp
# class vitembeding(nn.Module):
# def __init__(self):
# super(vitembeding, self).__init__()
# self.original_model = create_model(
# 'vit_base_patch16_224',
# pretrained=True,
# num_classes=1000,
# drop_rate=0.0,
# drop_path_rate=0.0,
# drop_block_rate=None,
# )
# def forward(self, x_input):
# for p in self.original_model.parameters():
# p.requires_grad = False
# x = self.original_model.forward_features(x_input)[:,0,:]
# return x
class PreNorm(nn.Module):
def __init__(self, dim, fn):
super().__init__()
self.norm = nn.LayerNorm(dim)
self.fn = fn
def forward(self, x, **kwargs):
return self.fn(self.norm(x), **kwargs)
class FeedForward(nn.Module):
def __init__(self, dim, hidden_dim, dropout = 0.):
super().__init__()
self.net = nn.Sequential(
nn.Linear(dim, hidden_dim),
nn.GELU(),
nn.Dropout(dropout),
nn.Linear(hidden_dim, dim),
nn.Dropout(dropout)
)
def forward(self, x):
return self.net(x)
class Attention(nn.Module):
def __init__(self, dim, heads = 8, dim_head = 64, dropout = 0.):
super().__init__()
inner_dim = dim_head * heads
project_out = not (heads == 1 and dim_head == dim)
self.heads = heads
self.scale = dim_head ** -0.5
self.attend = nn.Softmax(dim = -1)
self.to_qkv = nn.Linear(dim, inner_dim * 3, bias = False)
self.to_out = nn.Sequential(
nn.Linear(inner_dim, dim),
nn.Dropout(dropout)
) if project_out else nn.Identity()
def forward(self, x):
qkv = self.to_qkv(x).chunk(3, dim = -1)
q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h = self.heads), qkv)
dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale
attn = self.attend(dots)
out = torch.matmul(attn, v)
out = rearrange(out, 'b h n d -> b n (h d)')
return self.to_out(out)
class Transformer(nn.Module):
def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout = 0.):
super().__init__()
self.layers = nn.ModuleList([])
for _ in range(depth):
self.layers.append(nn.ModuleList([
PreNorm(dim, Attention(dim, heads = heads, dim_head = dim_head, dropout = dropout)),
PreNorm(dim, FeedForward(dim, mlp_dim, dropout = dropout))
]))
def forward(self, x):
for attn, ff in self.layers:
x = attn(x) + x
x = ff(x) + x
return x
class WGNbase(nn.Module):
def __init__(self, *, image_size, patch_size, num_classes, dim, depth, heads, mlp_dim, pool = 'cls', channels = 3, dim_head = 64, dropout = 0., emb_dropout = 0.):
super().__init__()
image_height, image_width = pair(image_size)
patch_height, patch_width = pair(patch_size)
self.num_classes = num_classes
assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.'
num_patches = (image_height // patch_height) * (image_width // patch_width)
patch_dim = channels * patch_height * patch_width
assert pool in {'cls', 'mean'}, 'pool type must be either cls (cls token) or mean (mean pooling)'
self.to_patch_embedding = nn.Sequential(
Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1 = patch_height, p2 = patch_width), ### 1,64,1024x3
nn.Linear(patch_dim, dim), ### 1,64,1024
)
self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim)) ### 1,65,1024
self.cls_token = nn.Parameter(torch.randn(1, 1, dim))
self.dropout = nn.Dropout(emb_dropout)
self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim, dropout)
# self.input_embedding = vitembeding() ### ResModel()
self.pool = pool
self.to_latent = nn.Identity()
self.mlp_head = nn.Sequential(
nn.LayerNorm(dim),
nn.Linear(dim, num_classes)
)
##########################
channel_a = 78
self.proj1_a = nn.Conv2d(channel_a, channel_a, kernel_size=1, stride=1, bias=True)
#self.proj2_a = nn.Conv2d(channel_a, 10, kernel_size=16, stride=16, bias=True)
self.proj2_a = nn.Linear(channel_a*256, 25)
channel_b = 75
self.proj1_b = nn.Conv2d(channel_b, 25, kernel_size=1, stride=1, bias=True)
self.proj2_b = nn.Conv2d(25, channel_b, kernel_size=1, stride=1, bias=True)
# self.proj2_b = nn.Linear(3*256, 768)
channel_c = 3
self.proj1_c = nn.Conv2d(channel_c, channel_c, kernel_size=1, stride=1, bias=True)
# self.proj2_c = nn.Conv2d(channel_b, 3, kernel_size=1, stride=1, bias=True)
self.proj2_c = nn.Linear(channel_c*256, 768)
self.acti_softmax = nn.Softmax(dim=-1)
self.acti_Sig = nn.Sigmoid()
self.pre_out = nn.Linear(768, self.num_classes)
self.head_size = 768
self.key_w = nn.Linear(768, self.head_size)
self.query_w = nn.Linear(768, self.head_size)
def l2_normalize(self, x, dim=None, epsilon=1e-12):
"""Normalizes a given vector or matrix."""
square_sum = torch.sum(x ** 2, dim=dim, keepdim=True)
# print('square_sum shape:',square_sum.shape) ### [10,1] for prompt [16,1] for x_embed_mean
x_inv_norm = torch.rsqrt(torch.maximum(square_sum, torch.tensor(epsilon, device=x.device)))
return x * x_inv_norm
def line_norm(self,x,dim=None):
ttt_max, ttindex_max = torch.max(x, dim=dim)
ttt_min, ttindex_min = torch.min(x, dim=dim)
# # # print('ttmax:',ttt_max.shape)
ttt_max = repeat(ttt_max, '... -> ... n', n = x.shape[dim])
ttt_min = repeat(ttt_min, '... -> ... n', n = x.shape[dim])
return (x-ttt_min)/(ttt_max- ttt_min)
def forward(self, x_embed, prompt_embed):
b,_,wh = x_embed.shape
key = self.key_w(x_embed)
# key = x_embed
query = self.query_w(prompt_embed)
# query = prompt_embed
Q = query
K = torch.transpose(key, 1, 2)
V = prompt_embed
scores = torch.matmul(Q, K)
scores = torch.mean(scores / math.sqrt(self.head_size), dim=-1)
probs = torch.sigmoid(scores)
probs_shape = list(probs.size())
probs = probs.unsqueeze(dim=-1).expand(probs_shape[0], probs_shape[1], 1)
prompt_embedding = torch.mul(V, probs)
return prompt_embedding