if cmd_args.phase == 'train':
optimizer = optim.Adam(classifier.parameters(),
lr=cmd_args.learning_rate)
train_idxes = list(range(len(train_glist)))
best_loss = None
for epoch in range(cmd_args.num_epochs):
random.shuffle(train_idxes)
avg_loss = loop_dataset(train_glist,
classifier,
train_idxes,
optimizer=optimizer)
print(
'\033[92maverage training of epoch %d: loss %.5f acc %.5f\033[0m'
% (epoch, avg_loss[0], avg_loss[1]))
test_loss = loop_dataset(test_glist, classifier,
list(range(len(test_glist))))
print(
'\033[93maverage test of epoch %d: loss %.5f acc %.5f\033[0m' %
(epoch, test_loss[0], test_loss[1]))
if best_loss is None or test_loss[0] < best_loss:
best_loss = test_loss[0]
print(
'----saving to best model since this is the best valid loss so far.----'
)
torch.save(classifier.state_dict(),
cmd_args.save_dir + '/epoch-best.model')
save_args(cmd_args.save_dir + '/epoch-best-args.pkl', cmd_args)
acc_train = acc_train.sum() / float(len(idx_train))
loss_train.backward()
optimizer.step()
gcn.eval()
_, loss_val, acc_val = gcn(features, orig_adj, idx_val, labels)
acc_val = acc_val.sum() / float(len(idx_val))
print('Epoch: {:04d}'.format(epoch + 1),
'loss_train: {:.4f}'.format(loss_train.data[0]),
'acc_train: {:.4f}'.format(acc_train),
'loss_val: {:.4f}'.format(loss_val.data[0]),
'acc_val: {:.4f}'.format(acc_val),
'time: {:.4f}s'.format(time.time() - t))
if best_val is None or acc_val > best_val:
best_val = acc_val
print(
'----saving to best model since this is the best valid loss so far.----'
)
torch.save(
gcn.state_dict(),
cmd_args.save_dir + '/model-%s-epoch-best-%.2f.model' %
(cmd_args.gm, cmd_args.del_rate))
save_args(
cmd_args.save_dir + '/model-%s-epoch-best-%.2f-args.pkl' %
(cmd_args.gm, cmd_args.del_rate), cmd_args)
run_test(gcn, features, orig_adj, idx_test, labels)
# pred = gcn(features, adh)