forked from PetarV-/DGI
-
Notifications
You must be signed in to change notification settings - Fork 0
/
torch_geometric_dgi_batched-Copy1.py
189 lines (153 loc) · 6.18 KB
/
torch_geometric_dgi_batched-Copy1.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
import os.path as osp
import torch
import torch.nn as nn
import numpy as np
from sklearn import metrics
from tqdm import tqdm
from torch_geometric.datasets import Reddit
from torch_geometric.loader import NeighborSampler
from torch_geometric.nn import SAGEConv
from torch_geometric.nn import DeepGraphInfomax
from torch_geometric.utils import to_undirected, add_remaining_self_loops
from ogb.nodeproppred import PygNodePropPredDataset
device = torch.device('cuda:2' if torch.cuda.is_available() else 'cpu')
# path = osp.join(osp.dirname(osp.realpath(__file__)), '..', 'data', 'Reddit')
# dataset = Reddit(path)
# data = dataset[0]
def Kmeans(x, K=-1, Niter=10, verbose=False):
#start = time.time()
N, D = x.shape # Number of samples, dimension of the ambient space
x_temp = x.detach()
temp = set()
while len(temp)<K:
temp.add(np.random.randint(0, N))
c = x_temp[list(temp), :].clone()
x_i = x_temp.view(N, 1, D) # (N, 1, D) samples
cutoff = 1
if K>cutoff:
c_j = []
niter=K//cutoff
rem = K%cutoff
if rem>0:
rem=1
for i in range(niter+rem):
c_j.append(c[i*cutoff:min(K,(i+1)*cutoff),:].view(1, min(K,(i+1)*cutoff)-(i*cutoff), D))
else:
c_j = c.view(1, K, D) # (1, K, D) centroids
# K-means loop:
# - x is the (N, D) point cloud,
# - cl is the (N,) vector of class labels
# - c is the (K, D) cloud of cluster centroids
for i in range(Niter):
#print("iteration: " + str(i))
# E step: assign points to the closest cluster -------------------------
if K>cutoff:
for j in range(len(c_j)):
if j==0:
D_ij = ((x_i - c_j[j]) ** 2).sum(-1)
else:
D_ij = torch.cat((D_ij,((x_i - c_j[j]) ** 2).sum(-1)), dim=-1)
# D_ij += ((x_i - c_j[j]) ** 2).sum(-1)
else:
D_ij = ((x_i - c_j) ** 2).sum(-1) # (N, K) symbolic squared distances
assert D_ij.size(1)==K
cl = D_ij.argmin(dim=1).long().view(-1) # Points -> Nearest cluster
c.zero_()
c.scatter_add_(0, cl[:, None].repeat(1, D), x_temp)
# Divide by the number of points per cluster:
Ncl = torch.bincount(cl, minlength=K).type_as(c).view(K, 1)
# print(Ncl[:10])
Ncl += 0.00000000001
c /= Ncl # in-place division to compute the average
return cl, c
dataset = PygNodePropPredDataset(name = "ogbn-arxiv", root="/home/devvrit_03/GraphNN/clean_codes/dataset")
split_idx = dataset.get_idx_split()
train_idx, valid_idx, test_idx = split_idx["train"], split_idx["valid"], split_idx["test"]
data = dataset[0].to(device)
data.edge_index = to_undirected(add_remaining_self_loops(data.edge_index)[0])
train_loader = NeighborSampler(data.edge_index, node_idx=None,
sizes=[25, 20, 10], batch_size=2048,
shuffle=True, num_workers=12)
test_loader = NeighborSampler(data.edge_index, node_idx=None,
sizes=[25, 20, 10], batch_size=2048,
shuffle=False, num_workers=12)
class Encoder(nn.Module):
def __init__(self, in_channels, hidden_channels):
super().__init__()
self.convs = torch.nn.ModuleList([
SAGEConv(in_channels, hidden_channels),
SAGEConv(hidden_channels, hidden_channels),
SAGEConv(hidden_channels, hidden_channels)
])
self.activations = torch.nn.ModuleList()
self.activations.extend([
nn.PReLU(hidden_channels),
nn.PReLU(hidden_channels),
nn.PReLU(hidden_channels)
])
def forward(self, x, adjs):
for i, (edge_index, _, size) in enumerate(adjs):
x_target = x[:size[1]] # Target nodes are always placed first.
x = self.convs[i]((x, x_target), edge_index)
x = self.activations[i](x)
return x
def corruption(x, edge_index):
return x[torch.randperm(x.size(0))], edge_index
model = DeepGraphInfomax(
hidden_channels=512, encoder=Encoder(dataset.num_features, 512),
summary=lambda z, *args, **kwargs: torch.sigmoid(z.mean(dim=0)),
corruption=corruption).to(device)
model = model.to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=0.0001)
x, y = data.x.to(device), data.y.view(-1).to(device)
def train(epoch):
model.train()
total_loss = total_examples = 0
it=0
for batch_size, n_id, adjs in tqdm(train_loader,
desc=f'Epoch {epoch:02d}'):
# `adjs` holds a list of `(edge_index, e_id, size)` tuples.
adjs = [adj.to(device) for adj in adjs]
optimizer.zero_grad()
pos_z, neg_z, summary = model(x[n_id], adjs)
loss = model.loss(pos_z, neg_z, summary)
loss.backward()
optimizer.step()
total_loss += float(loss) * pos_z.size(0)
total_examples += pos_z.size(0)
it+=1
# if it==10:
# break
zs = []
with torch.no_grad():
model.eval()
for batch_size, n_id, adjs in tqdm(test_loader, desc=f'Test Epoch {epoch:02d}'):
adjs = [adj.to(device) for adj in adjs]
zs.append(model(x[n_id], adjs)[0])
zs = torch.cat(zs, dim=0)
zs = torch.nn.functional.normalize(zs)
y_pred,_ = Kmeans(zs.detach(), y.max()+1)
nmi = metrics.normalized_mutual_info_score(y.cpu().numpy(), y_pred.cpu().numpy())
print("nmi: " + str(nmi))
return total_loss / total_examples
@torch.no_grad()
def test(e):
model.eval()
zs = []
for i, (batch_size, n_id, adjs) in enumerate(test_loader):
adjs = [adj.to(device) for adj in adjs]
zs.append(model(x[n_id], adjs)[0])
z = torch.cat(zs, dim=0)
torch.save(z, "embedding_products.pt_epoch_" + str(e))
return
train_val_mask = data.train_idx | data.valid_idx
acc = model.test(z[train_val_mask], y[train_val_mask], z[data.test_idx],
y[data.test_idx], max_iter=10000)
return acc
for epoch in range(1, 21):
model.train()
loss = train(epoch)
print(f'Epoch {epoch:02d}, Loss: {loss:.4f}')
# test(epoch)
test_acc = test()
print(f'Test Accuracy: {test_acc:.4f}')