-
Notifications
You must be signed in to change notification settings - Fork 1
/
cnn_target_clear.py
92 lines (59 loc) · 2.61 KB
/
cnn_target_clear.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
import torch
import torch.nn as nn
import torch.nn.functional as F
class Net(nn.Module):
def __init__(self, hidden_size):
super(Net, self).__init__()
# define layers of CNN
# input >> hidden layer
self.conv1 = nn.Conv2d(hidden_size, hidden_size,kernel_size=3, stride=1, padding=1)
self.batchn1 = nn.BatchNorm2d(hidden_size)
self.conv2 = nn.Conv2d(hidden_size, hidden_size, kernel_size=3, stride=1, padding=1)
self.batchn2 = nn.BatchNorm2d(hidden_size)
def forward(self, chessboard):
# define forward behavior
# activations and batch normalization
chessboard = self.conv1(chessboard)
chessboard = self.batchn1(chessboard)
chessboard = F.relu(chessboard)
chessboard = self.conv2(chessboard)
chessboard = self.batchn2(chessboard)
chessboard = F.relu(chessboard)
return chessboard
class PlainChessNET(nn.Module):
def __init__(self, hidden_layers=2, hidden_size=128):
super(PlainChessNET, self).__init__()
# define layers of CNN
# chessboard part (12,8,8) input >> hidden layer
self.hidden_layers = hidden_layers
self.input_layer = nn.Conv2d(12, hidden_size, kernel_size=3, stride=1, padding=1)
self.modulelist = nn.ModuleList([Net(hidden_size) for i in range(hidden_layers)])
self.flatten = nn.Flatten()
self.fc1 = nn.Linear(hidden_size * 64, 64)
# source (the selected squares)
self.input_source = nn.Linear(64,64)
self.batchn1d_1 = nn.BatchNorm1d(64)
# output target (the targeted square)
self.output_target = nn.Linear(64,64)
def forward(self, chessboard, source):
# define forward behavior
# chessboard context
chessboard = self.input_layer(chessboard)
chessboard = F.relu(chessboard)
for h in range(self.hidden_layers):
chessboard = self.modulelist[h](chessboard)
chessboard = self.flatten(chessboard)
chessboard = self.fc1(chessboard)
chessboard = F.relu(chessboard)
# source aka selected square to be moved to target
source = self.input_source(source)
# merging chessboard (context + selected source square)
merge = chessboard + source
merge = self.batchn1d_1(merge)
# nllloss crossentropyloss
# x_target = F.log_softmax(self.output_target(merge),dim=1)
# mseloss
#x_target = F.relu(self.output_target(merge))
# sigmoid
x_target = torch.sigmoid(self.output_target(merge))
return x_target