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
|
import copy
import os.path as osp
import torch
import torch.nn.functional as F
from torch.nn import GRU, Linear, ReLU, Sequential
import torch_geometric.transforms as T
from torch_geometric.datasets import QM9
from torch_geometric.loader import DataLoader
from torch_geometric.nn import NNConv, Set2Set
from torch_geometric.utils import remove_self_loops
target = 0
dim = 64
class MyTransform:
def __call__(self, data):
data = copy.copy(data)
data.y = data.y[:, target] # Specify target.
return data
class Complete:
def __call__(self, data):
data = copy.copy(data)
device = data.edge_index.device
row = torch.arange(data.num_nodes, dtype=torch.long, device=device)
col = torch.arange(data.num_nodes, dtype=torch.long, device=device)
row = row.view(-1, 1).repeat(1, data.num_nodes).view(-1)
col = col.repeat(data.num_nodes)
edge_index = torch.stack([row, col], dim=0)
edge_attr = None
if data.edge_attr is not None:
idx = data.edge_index[0] * data.num_nodes + data.edge_index[1]
size = list(data.edge_attr.size())
size[0] = data.num_nodes * data.num_nodes
edge_attr = data.edge_attr.new_zeros(size)
edge_attr[idx] = data.edge_attr
edge_index, edge_attr = remove_self_loops(edge_index, edge_attr)
data.edge_attr = edge_attr
data.edge_index = edge_index
return data
path = osp.join(osp.dirname(osp.realpath(__file__)), '..', 'data', 'QM9')
transform = T.Compose([MyTransform(), Complete(), T.Distance(norm=False)])
dataset = QM9(path, transform=transform).shuffle()
# Normalize targets to mean = 0 and std = 1.
mean = dataset.data.y.mean(dim=0, keepdim=True)
std = dataset.data.y.std(dim=0, keepdim=True)
dataset.data.y = (dataset.data.y - mean) / std
mean, std = mean[:, target].item(), std[:, target].item()
# Split datasets.
test_dataset = dataset[:10000]
val_dataset = dataset[10000:20000]
train_dataset = dataset[20000:]
test_loader = DataLoader(test_dataset, batch_size=128, shuffle=False)
val_loader = DataLoader(val_dataset, batch_size=128, shuffle=False)
train_loader = DataLoader(train_dataset, batch_size=128, shuffle=True)
class Net(torch.nn.Module):
def __init__(self):
super().__init__()
self.lin0 = torch.nn.Linear(dataset.num_features, dim)
nn = Sequential(Linear(5, 128), ReLU(), Linear(128, dim * dim))
self.conv = NNConv(dim, dim, nn, aggr='mean')
self.gru = GRU(dim, dim)
self.set2set = Set2Set(dim, processing_steps=3)
self.lin1 = torch.nn.Linear(2 * dim, dim)
self.lin2 = torch.nn.Linear(dim, 1)
def forward(self, data):
out = F.relu(self.lin0(data.x))
h = out.unsqueeze(0)
for i in range(3):
m = F.relu(self.conv(out, data.edge_index, data.edge_attr))
out, h = self.gru(m.unsqueeze(0), h)
out = out.squeeze(0)
out = self.set2set(out, data.batch)
out = F.relu(self.lin1(out))
out = self.lin2(out)
return out.view(-1)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = Net().to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, mode='min',
factor=0.7, patience=5,
min_lr=0.00001)
def train(epoch):
model.train()
loss_all = 0
for data in train_loader:
data = data.to(device)
optimizer.zero_grad()
loss = F.mse_loss(model(data), data.y)
loss.backward()
loss_all += loss.item() * data.num_graphs
optimizer.step()
return loss_all / len(train_loader.dataset)
def test(loader):
model.eval()
error = 0
for data in loader:
data = data.to(device)
error += (model(data) * std - data.y * std).abs().sum().item() # MAE
return error / len(loader.dataset)
best_val_error = None
for epoch in range(1, 301):
lr = scheduler.optimizer.param_groups[0]['lr']
loss = train(epoch)
val_error = test(val_loader)
scheduler.step(val_error)
if best_val_error is None or val_error <= best_val_error:
test_error = test(test_loader)
best_val_error = val_error
print(f'Epoch: {epoch:03d}, LR: {lr:7f}, Loss: {loss:.7f}, '
f'Val MAE: {val_error:.7f}, Test MAE: {test_error:.7f}')
|
