-
Notifications
You must be signed in to change notification settings - Fork 3
/
train.py
53 lines (38 loc) · 1.47 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
import numpy as np
import torch
import random
from model import Net
from data_loader import data_loader
import torch.nn.functional as F
def test(model, data):
model.eval()
logits, accs = model(data), []
test_loss = F.nll_loss(model(data)[data.test_mask], data.y[data.test_mask]).detach().cpu().numpy()
for _, mask in data('train_mask', 'test_mask'):
pred = logits[mask].max(1)[1]
acc = pred.eq(data.y[mask]).sum().item() / mask.sum().item()
accs.append(acc)
return [test_loss] + accs
def train(model, optimizer, data):
model.train()
losses = []
for epoch in range(1, 200):
optimizer.zero_grad()
loss = F.nll_loss(model(data)[data.train_mask], data.y[data.train_mask])
loss.backward()
optimizer.step()
train_loss = loss.detach().cpu().numpy()
log = 'Epoch: {:03d}, train_loss: {:.3f}, test_loss:{:.3f}, train_acc: {:.2f}, test_acc: {:.2f}'
test_loss = test(model, data)[0]
losses.append([train_loss, test_loss])
test_loss, train_acc, test_acc = test(model, data)
print(log.format(epoch, train_loss, test_loss, train_acc, test_acc))
def main():
data = data_loader()
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model, data = Net().to(device), data.to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)
train(model, optimizer, data)
test(model, data)
if __name__ == "__main__":
main()