-
Notifications
You must be signed in to change notification settings - Fork 2
/
train.py
220 lines (161 loc) · 7.71 KB
/
train.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
210
211
212
213
214
215
216
217
218
219
220
import torch
import torch.nn.functional as F
from torch.optim import Adam
from torch import tensor
from utils import get_batch, Ncontrast, accuracy
from tqdm import tqdm
from torch_geometric.utils.convert import to_networkx
import time
import networkx as nx
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
def train_citation(alpha, beta, tau, batch_size, adj_label, idx_train, features, labels, model, optimizer):
features_batch, adj_label_batch = get_batch(batch_size, adj_label, idx_train, features)
model.train()
optimizer.zero_grad()
x_bar, x_dis, output = model(features_batch)
loss_train_class = F.nll_loss(output[idx_train], labels[idx_train])
loss_Ncontrast = Ncontrast(x_dis, adj_label_batch, tau=tau)
loss_train = loss_train_class + alpha * loss_Ncontrast + beta * F.mse_loss(x_bar, features_batch)
acc_train = accuracy(output[idx_train], labels[idx_train])
loss_train.backward()
optimizer.step()
return
def evaluate_citation(model, features, labels, idx_train, idx_val, idx_test):
model.eval()
with torch.no_grad():
logits = model(features)
outs = {}
train_loss = F.nll_loss(logits[idx_train], labels[idx_train]).item()
train_pred = logits[idx_train].max(1)[1]
train_acc = train_pred.eq(labels[idx_train]).sum().item() / len(idx_train)
outs['train_loss'] = train_loss
outs['train_acc'] = train_acc
val_loss = F.nll_loss(logits[idx_val], labels[idx_val]).item()
val_pred = logits[idx_val].max(1)[1]
val_acc = val_pred.eq(labels[idx_val]).sum().item() / len(idx_val)
outs['val_loss'] = val_loss
outs['val_acc'] = val_acc
test_loss = F.nll_loss(logits[idx_test], labels[idx_test]).item()
test_pred = logits[idx_test].max(1)[1]
test_acc = test_pred.eq(labels[idx_test]).sum().item() / len(idx_test)
outs['test_loss'] = test_loss
outs['test_acc'] = test_acc
return outs
def train_coauthor_amazon(data, alpha, beta, tau, adj_label, model, optimizer):
model.train()
optimizer.zero_grad()
x_bar, x_dis, output = model(data.x)
loss_train_class = F.nll_loss(output[data.train_mask], data.y[data.train_mask])
loss_Ncontrast = Ncontrast(x_dis, adj_label, tau)
loss_train = loss_train_class + alpha * loss_Ncontrast + beta * F.mse_loss(x_bar, data.x)
loss_train.backward()
optimizer.step()
return
def evaluate_coauthor_amazon(model, data):
model.eval()
with torch.no_grad():
logits = model(data.x)
outs = {}
for key in ['train', 'val', 'test']:
mask = data['{}_mask'.format(key)]
loss = F.nll_loss(logits[mask], data.y[mask]).item()
pred = logits[mask].max(1)[1]
acc = pred.eq(data.y[mask]).sum().item() / mask.sum().item()
outs['{}_loss'.format(key)] = loss
outs['{}_acc'.format(key)] = acc
return outs
def run_citation(model, runs, epochs, lr, weight_decay, early_stopping, alpha, beta, tau, batch_size, adj_label,
features, labels, idx_train, idx_val,
idx_test):
val_losses, accs, durations = [], [], []
print(
f'raw_data_dim:{features.shape[1]}, classes:{torch.max(labels) + 1},',
f'Train_data_num:{idx_train.shape[0]}, Val_data_num:{idx_val.shape[0]}, Test_data_num:{idx_test.shape[0]}')
pbar = tqdm(range(runs), unit='run')
for _ in pbar:
model.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
if torch.cuda.is_available():
torch.cuda.synchronize()
t_start = time.perf_counter()
best_val_loss = float('inf')
test_acc = 0
val_loss_history = []
for epoch in range(1, epochs + 1):
out = train_citation(alpha, beta, tau, batch_size, adj_label, idx_train, features, labels, model, optimizer)
eval_info = evaluate_citation(model, features, labels, idx_train, idx_val, idx_test)
eval_info['epoch'] = epoch
if eval_info['val_loss'] < best_val_loss:
best_val_loss = eval_info['val_loss']
test_acc = eval_info['test_acc']
val_loss_history.append(eval_info['val_loss'])
if early_stopping > 0 and epoch > epochs // 2:
tmp = tensor(val_loss_history[-(early_stopping + 1):-1])
if eval_info['val_loss'] > tmp.mean().item():
break
if torch.cuda.is_available():
torch.cuda.synchronize()
t_end = time.perf_counter()
val_losses.append(best_val_loss)
accs.append(test_acc)
durations.append(t_end - t_start)
loss, acc, duration = tensor(val_losses), tensor(accs), tensor(durations)
print('Val Loss: {:.4f}, Test Accuracy: {:.3f} ± {:.3f}, Duration: {:.3f}'.
format(loss.mean().item(),
acc.mean().item(),
acc.std().item(),
duration.mean().item()))
def run_coauthor_amazon(dataset, model, runs, epochs, lr, weight_decay, early_stopping, alpha, beta, tau, adj_label,
permute_masks=None, lcc=False):
val_losses, accs, durations = [], [], []
lcc_mask = None
if lcc: # select largest connected component
data_ori = dataset[0]
data_nx = to_networkx(data_ori)
data_nx = data_nx.to_undirected()
print("Original #nodes:", data_nx.number_of_nodes())
data_nx = data_nx.subgraph(max(nx.connected_components(data_nx), key=len))
print("#Nodes after lcc:", data_nx.number_of_nodes())
lcc_mask = list(data_nx.nodes)
pbar = tqdm(range(runs), unit='run')
data = dataset[0]
for _ in pbar:
if permute_masks is not None:
data = permute_masks(data, dataset.num_classes, lcc_mask)
data = data.to(device)
if _ == 0:
print(
f'raw_data_dim:{data.x.shape[1]}, classes:{torch.max(data.y) + 1},',
f'Train_data_num:{sum(data.train_mask)}, Val_data_num:{sum(data.val_mask)}, Test_data_num:{sum(data.test_mask)}')
model.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
if torch.cuda.is_available():
torch.cuda.synchronize()
t_start = time.perf_counter()
best_val_loss = float('inf')
test_acc = 0
val_loss_history = []
for epoch in range(1, epochs + 1):
out = train_coauthor_amazon(data, alpha, beta, tau, adj_label, model, optimizer)
eval_info = evaluate_coauthor_amazon(model, data)
eval_info['epoch'] = epoch
if eval_info['val_loss'] < best_val_loss:
best_val_loss = eval_info['val_loss']
test_acc = eval_info['test_acc']
val_loss_history.append(eval_info['val_loss'])
if early_stopping > 0 and epoch > epochs // 2:
tmp = tensor(val_loss_history[-(early_stopping + 1):-1])
if eval_info['val_loss'] > tmp.mean().item():
break
if torch.cuda.is_available():
torch.cuda.synchronize()
t_end = time.perf_counter()
val_losses.append(best_val_loss)
accs.append(test_acc)
durations.append(t_end - t_start)
loss, acc, duration = tensor(val_losses), tensor(accs), tensor(durations)
print('Val Loss: {:.4f}, Test Accuracy: {:.3f} ± {:.3f}, Duration: {:.3f}'.
format(loss.mean().item(),
acc.mean().item(),
acc.std().item(),
duration.mean().item()))