forked from brohrer/sharpened-cosine-similarity
-
Notifications
You must be signed in to change notification settings - Fork 0
/
models.py
88 lines (71 loc) · 2.83 KB
/
models.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
from torch import nn
from sharpened_cosine_similarity import SharpenedCosineSimilarity
from absolute_pooling import MaxAbsPool2d
class ResBlk(nn.Module):
def __init__(self, in_ch, out_ch, pool, use_residual):
super().__init__()
self.mod1 = SharpenedCosineSimilarity(in_ch, out_ch, kernel_size=3, padding=1)
self.mod2 = SharpenedCosineSimilarity(out_ch, out_ch, kernel_size=3, padding=1)
if use_residual:
self.conv1x1 = SharpenedCosineSimilarity(in_ch, out_ch, kernel_size=1, padding=0)
if pool:
self.pool = MaxAbsPool2d(kernel_size=2, stride=2, ceil_mode=True)
def forward(self, x):
y = x
x = self.mod1(x)
x = self.mod2(x)
if hasattr(self, "pool"):
x = self.pool(x)
if hasattr(self, "conv1x1"):
if hasattr(self, "pool"):
y = self.pool(y)
y = self.conv1x1(y)
x += y
return x
class ResidualNetwork(nn.Module):
def __init__(
self,
start_ch=32,
num_blocks_per_level=1,
use_residual=True
):
super().__init__()
self.backbone = nn.Sequential(
SharpenedCosineSimilarity(3, 32, 7, padding=3),
)
for level in range(3):
cur_ch = start_ch * 2**(level)
for i in range(num_blocks_per_level-1):
self.backbone.add_module(f"level{level}_block{i}",
ResBlk(cur_ch, cur_ch, pool=False, use_residual=use_residual)
)
self.backbone.add_module(f"level{level}",
ResBlk(cur_ch, cur_ch*2, pool=True,use_residual=use_residual))
self.out = nn.Linear(in_features=start_ch*8*4*4, out_features=10)
def forward(self, t):
t = self.backbone(t).flatten(start_dim=1)
t = self.out(t)
return t
class OriginalModel(nn.Module):
def __init__(self):
super().__init__()
self.scs1 = SharpenedCosineSimilarity(
in_channels=3, out_channels=24, kernel_size=3, padding=0)
self.pool1 = MaxAbsPool2d(kernel_size=2, stride=2, ceil_mode=True)
self.scs2 = SharpenedCosineSimilarity(
in_channels=24, out_channels=48, kernel_size=3, padding=1)
self.pool2 = MaxAbsPool2d(kernel_size=2, stride=2, ceil_mode=True)
self.scs3 = SharpenedCosineSimilarity(
in_channels=48, out_channels=96, kernel_size=3, padding=1)
self.pool3 = MaxAbsPool2d(kernel_size=2, stride=2, ceil_mode=True)
self.out = nn.Linear(in_features=96*4*4, out_features=10)
def forward(self, t):
t = self.scs1(t)
t = self.pool1(t)
t = self.scs2(t)
t = self.pool2(t)
t = self.scs3(t)
t = self.pool3(t)
t = t.reshape(-1, 96*4*4)
t = self.out(t)
return t