-
Notifications
You must be signed in to change notification settings - Fork 3
/
inceptionMixer_iFormer.py
145 lines (114 loc) · 5 KB
/
inceptionMixer_iFormer.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
""" (ref1) https://github.com/sail-sg/iFormer
(ref2) https://arxiv.org/abs/2205.12956
"""
import torch
import torch.nn as nn
class HighMixer(nn.Module):
""" Mixer for high-frequency features
via 2D convolution & max-pooling
"""
def __init__(self, dim, kernel_size=3, stride=1, padding=1, **kwargs):
super().__init__()
self.cnn_in = cnn_in = dim // 2
pool_in = dim // 2
cnn_dim = cnn_in * 2 # expand dim for output
pool_dim = pool_in * 2 # expand dim for output
# == Branch 1 == #
self.conv1 = nn.Conv2d(cnn_in, cnn_dim, kernel_size=1, stride=1, padding=0, bias=False)
self.proj1 = nn.Conv2d(cnn_dim, cnn_dim, kernel_size=kernel_size, stride=stride, padding=padding, bias=False, groups=cnn_dim)
self.act1 = nn.GELU()
# == Branch 2 == #
self.Maxpool = nn.MaxPool2d(kernel_size, stride=stride, padding=padding)
self.proj2 = nn.Conv2d(pool_in, pool_dim, kernel_size=1, stride=1, padding=0)
self.act2 = nn.GELU()
def forward(self, x):
""" x: (B,C,H,W)
"""
# -- Branch 1
cx = x[:, :self.cnn_in, :, :].contiguous() # split dim
cx = self.conv1(cx)
cx = self.proj1(cx)
cx = self.act1(cx)
# -- Branch 2
px = x[:, self.cnn_in:, :, :].contiguous()
px = self.Maxpool(px)
px = self.proj2(px)
px = self.act2(px)
# -- Assemble
hx = torch.cat((cx, px), dim=1)
return hx
class LowMixer(nn.Module):
""" Mixer for low-frequency features
via multi-head self-attention (MHSA)
"""
def __init__(self, dim, num_heads=8, qkv_bias=False, attn_drop=0., pool_size=2, **kwargs):
super().__init__()
self.num_heads = num_heads
self.head_dim = head_dim = dim // num_heads
self.scale = head_dim ** -0.5
self.dim = dim
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
self.attn_drop = nn.Dropout(attn_drop)
self.pool = nn.AvgPool2d(pool_size, stride=pool_size, padding=0, count_include_pad=False) if pool_size > 1 else nn.Identity()
self.uppool = nn.Upsample(scale_factor=pool_size) if pool_size > 1 else nn.Identity()
def att_fun(self, q, k, v, B, N, C):
""" MHSA
"""
attn = (q @ k.transpose(-2, -1)) * self.scale
attn = attn.softmax(dim=-1)
attn = self.attn_drop(attn)
x = (attn @ v).transpose(2, 3).reshape(B, C, N)
return x
def forward(self, x):
""" x: (B,C,H,W)
"""
B, _, _, _ = x.shape
xa = self.pool(x)
xa = xa.permute(0, 2, 3, 1).view(B, -1, self.dim) # (B,C,H,W) -> (B, N, C)
B, N, C = xa.shape
qkv = self.qkv(xa).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
q, k, v = qkv.unbind(0) # make torchscript happy (cannot use tensor as tuple)
xa = self.att_fun(q, k, v, B, N, C)
xa = xa.view(B, C, int(N**0.5), int(N**0.5))#.permute(0, 3, 1, 2)
xa = self.uppool(xa)
return xa
class iMixer(nn.Module):
""" Inception Mixer with HighMixer & LowMixer
"""
def __init__(self, dim, num_heads=8, qkv_bias=False,
attn_drop=0., proj_drop=0., attention_head=1, pool_size=2, **kwargs):
super().__init__()
self.num_heads = num_heads
self.head_dim = head_dim = dim // num_heads
self.low_dim = low_dim = attention_head * head_dim
self.high_dim = high_dim = dim - low_dim
self.high_mixer = HighMixer(high_dim)
self.low_mixer = LowMixer(low_dim, num_heads=attention_head, qkv_bias=qkv_bias, attn_drop=attn_drop, pool_size=pool_size)
self.conv_fuse = nn.Conv2d(low_dim+high_dim*2, low_dim+high_dim*2, kernel_size=3, stride=1, padding=1, bias=False, groups=low_dim+high_dim*2)
self.proj = nn.Conv2d(low_dim+high_dim*2, dim, kernel_size=1, stride=1, padding=0)
self.proj_drop = nn.Dropout(proj_drop)
def forward(self, x):
B, H, W, C = x.shape
x = x.permute(0, 3, 1, 2) # (B,H,W,C) -> (B,C,H,W)
# -- branch for high-frequency features
hx = x[:,:self.high_dim,:,:].contiguous()
hx = self.high_mixer(hx)
# -- branch for low-frequency features
lx = x[:,self.high_dim:,:,:].contiguous()
lx = self.low_mixer(lx)
x = torch.cat((hx, lx), dim=1)
x = x + self.conv_fuse(x)
x = self.proj(x)
x = self.proj_drop(x)
x = x.permute(0, 2, 3, 1).contiguous() # (B,C,H,W) -> (B,H,W,C)
return x
if __name__ == "__main__":
embed_dims = [96, 192, 320, 384]
num_heads = [3, 6, 10, 12] # 96//3 , 192//6, 320//10, 384//12
attention_heads = [1]*3 + [3]*3 + [7] * 4 + [9] * 5 + [11] * 3
i = 3
x = torch.randn(1, 7, 7, embed_dims[i]) # input
attn = iMixer(dim=embed_dims[i], num_heads=num_heads[i], qkv_bias=True,
attn_drop=0., attention_head=attention_heads[3+3+4+5+1], pool_size=1)
output = attn(x)
print(output.shape)