/
airfoilgcnn.py
210 lines (170 loc) · 7.79 KB
/
airfoilgcnn.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
import torch
from torch import nn
import torch.nn.functional as F
from torch_geometric.data import Data,DataLoader
from torch_geometric.nn import GraphConv, TopKPooling, GCNConv, avg_pool, TAGConv, SAGEConv
from torch_geometric.nn import global_mean_pool as gap, global_max_pool as gmp
import tqdm
from tqdm import tqdm
import pandas as pd
import os
import numpy as np
import matplotlib.pyplot as plt
#from sklearn.model_selection import KFold
# Set torch device
if(torch.cuda.is_available()):
print("USING GPU")
device = torch.device("cuda:0")
else:
print("USING CPU")
device = torch.device("cpu")
#device = torch.device("cpu")
class NodeRemovalNet(torch.nn.Module):
def __init__(self, output_dim, conv_width=64, topk=0.5, initial_num_nodes=None):
super(NodeRemovalNet, self).__init__()
self.conv_width = conv_width
self.initial_num_nodes = initial_num_nodes
self.conv1 = SAGEConv(2, conv_width)
self.pool1 = TopKPooling(conv_width, ratio=topk)
self.conv2 = SAGEConv(conv_width, conv_width)
self.pool2 = TopKPooling(conv_width, ratio= topk)
self.conv3 = SAGEConv(conv_width, conv_width)
self.pool3 = TopKPooling(conv_width, ratio=topk)
self.conv4 = GCNConv(conv_width, conv_width)
self.pool4 = TopKPooling(conv_width, ratio=topk)
self.conv5 = GCNConv(conv_width, conv_width)
self.pool5 = TopKPooling(conv_width, ratio=topk)
self.conv6 = GCNConv(conv_width, conv_width)
self.pool6 = TopKPooling(conv_width, ratio=topk)
self.lin1 = torch.nn.Linear(2*conv_width, 128)
self.lin2 = torch.nn.Linear(128, 64)
self.lin3 = torch.nn.Linear(64, output_dim)
torch.manual_seed(0)
self.reset()
def reset(self):
# SAGEConv layers
nn.init.xavier_normal_(self.conv1.lin_l.weight, gain=0.9)
nn.init.normal_(self.conv1.lin_l.bias)
nn.init.xavier_normal_(self.conv1.lin_r.weight, gain=0.9)
nn.init.xavier_normal_(self.conv2.lin_l.weight, gain=0.9)
nn.init.normal_(self.conv2.lin_l.bias)
nn.init.xavier_normal_(self.conv2.lin_r.weight, gain=0.9)
nn.init.xavier_normal_(self.conv3.lin_l.weight, gain=0.9)
nn.init.normal_(self.conv3.lin_l.bias)
nn.init.xavier_normal_(self.conv3.lin_r.weight, gain=0.9)
# GCN layers
nn.init.xavier_normal_(self.conv4.lin.weight, gain=0.9)
nn.init.xavier_normal_(self.conv5.lin.weight, gain=0.9)
nn.init.xavier_normal_(self.conv6.lin.weight, gain=0.9)
nn.init.xavier_normal_(self.lin1.weight, gain=0.9)
nn.init.normal_(self.lin1.bias)
nn.init.xavier_normal_(self.lin2.weight, gain=0.9)
nn.init.normal_(self.lin2.bias)
nn.init.xavier_normal_(self.lin3.weight, gain=0.9)
nn.init.normal_(self.lin3.bias)
def set_num_nodes(self, initial_num_nodes):
self.initial_num_nodes = initial_num_nodes
self.conv1 = SAGEConv(self.initial_num_nodes, self.conv_width).to(device)
def set_removable(self, removable):
self.removable = removable
def forward(self, data, embedding=False):
"""
data: Batch of Pytorch Geometric data objects, containing node features, edge indices and batch size
returns: Predicted normalized drag value
"""
x, edge_index, batch = data.x.float(), data.edge_index, data.batch
#print(x.shape)
x = F.relu(self.conv1(x, edge_index))
#print(x.shape)
x, edge_index, _, batch, _, _ = self.pool1(x, edge_index, None, batch)
#print(x.shape)
x1 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv2(x, edge_index))
#print(x.shape)
x, edge_index, _, batch, _, _ = self.pool2(x, edge_index, None, batch)
#print(x.shape)
x2 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
#x = F.relu(self.conv3(x, edge_index))
#print(x.shape)
#x, edge_index, _, batch, _, _ = self.pool3(x, edge_index, None, batch)
#print(x.shape)
#x3 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv4(x, edge_index))
#print(x.shape)
x, edge_index, _, batch, _, _ = self.pool4(x, edge_index, None, batch)
#print(x.shape)
x4 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv5(x, edge_index))
#print(x.shape)
x, edge_index, _, batch, _, _ = self.pool5(x, edge_index, None, batch)
#print(x.shape)
x5 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
#x = F.relu(self.conv6(x, edge_index))
#print(x.shape)
#x, edge_index, _, batch, _, _ = self.pool6(x, edge_index, None, batch)
#print(x.shape)
#x6 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
#print(x1.shape, x2.shape, x3.shape, x4.shape, x5.shape, x6.shape)
#raise
#x = x1+x2+x3+x4+x5+x6
x = x1+x2+x4+x5
if(embedding):
return x
x = F.relu(self.lin1(x))
x = F.dropout(x, p=0.0, training=self.training)
x = F.relu(self.lin2(x))
x = self.lin3(x)
x = F.softmax(x, dim=1) # Pick which vertex to remove
return x
class AirfoilGCNN(torch.nn.Module):
def __init__(self, conv_width=64):
super(AirfoilGCNN, self).__init__()
#conv_width = 64
topk = 0.5
self.conv1 = SAGEConv(2, conv_width)
self.pool1 = TopKPooling(conv_width, ratio=topk)
self.conv2 = SAGEConv(conv_width, conv_width)
self.pool2 = TopKPooling(conv_width, ratio= topk)
self.conv3 = SAGEConv(conv_width, conv_width)
self.pool3 = TopKPooling(conv_width, ratio=topk)
self.conv4 = GCNConv(conv_width, conv_width)
self.pool4 = TopKPooling(conv_width, ratio=topk)
self.conv5 = GCNConv(conv_width, conv_width)
self.pool5 = TopKPooling(conv_width, ratio=topk)
self.conv6 = GCNConv(conv_width, conv_width)
self.pool6 = TopKPooling(conv_width, ratio=topk)
self.lin1 = torch.nn.Linear(2*conv_width, 128)
self.lin2 = torch.nn.Linear(128, 64)
self.lin3 = torch.nn.Linear(64, 1)
def forward(self, data):
"""
data: Batch of Pytorch Geometric data objects, containing node features, edge indices and batch size
returns: Predicted normalized drag value
"""
x, edge_index, batch = data.x.float(), data.edge_index, data.batch
x = x[:,[2,3]]
x = F.relu(self.conv1(x, edge_index))
x, edge_index, _, batch, _, _ = self.pool1(x, edge_index, None, batch)
x1 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv2(x, edge_index))
x, edge_index, _, batch, _, _ = self.pool2(x, edge_index, None, batch)
x2 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
#print(x.shape)
x = F.relu(self.conv3(x, edge_index))
x, edge_index, _, batch, _, _ = self.pool3(x, edge_index, None, batch)
x3 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv4(x, edge_index))
x, edge_index, _, batch, _, _ = self.pool4(x, edge_index, None, batch)
x4 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv5(x, edge_index))
x, edge_index, _, batch, _, _ = self.pool5(x, edge_index, None, batch)
x5 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = F.relu(self.conv6(x, edge_index))
x, edge_index, _, batch, _, _ = self.pool6(x, edge_index, None, batch)
x6 = torch.cat([gmp(x, batch), gap(x, batch)], dim=1)
x = x1+x2+x3+x4+x5+x6
x = F.relu(self.lin1(x))
x = F.dropout(x, p=0.0, training=self.training)
x = F.relu(self.lin2(x))
x = self.lin3(x)
return x