def train_model(loss, batch_size, num_epochs, learning_rate):
enc = Encoder(**cifar_config().encoder_arc).to(device)
dec = Decoder(**cifar_config().decoder_arc).to(device)
model = [enc, dec]
gd = optim.Adam(chain(*[
x.parameters() for x in model
if (isinstance(x, nn.Module) or isinstance(x, nn.Parameter))
]),
learning_rate,
weight_decay=1e-5)
train_loader, test_loader = get_cifar(batch_size=batch_size,
num_workers=32)
train_losses = []
test_results = []
for cnt in range(num_epochs):
for i, (batch, _) in enumerate(train_loader):
total = len(train_loader)
gd.zero_grad()
batch = batch.to(device)
loss_value, _ = loss(batch, enc, dec)
loss_value.backward()
train_losses.append(loss_value.item())
if (i + 1) % 10 == 0:
print('\rTrain loss:',
train_losses[-1],
'Batch',
i + 1,
'of',
total,
' ' * 10,
end='',
flush=True)
gd.step()
test_elbo = 0.
test_mse = 0.
with torch.autograd.no_grad():
for i, (batch, _) in enumerate(test_loader):
batch = batch.to(device)
batch_loss, recon = loss(batch, enc, dec)
test_mse += (torch.nn.MSELoss()(recon, batch) -
test_mse) / (i + 1)
test_elbo += (batch_loss - test_elbo) / (i + 1)
print('\nTest elbo after at epoch {}: {}'.format(cnt, test_elbo))
print('Test mse after at epoch {}: {}'.format(cnt, test_mse))
test_results.append((test_elbo, test_mse))
enc.cpu()
dec.cpu()
torch.save(enc.state_dict(), "./ckpt/enc.pt")
torch.save(dec.state_dict(), "./ckpt/dec.pt")
with open('./log/log.txt', 'w') as f:
for item in test_results:
f.write("%s\n" % float(item))
# timer # ta
time_str = self.calculate_remaining(start_time, time.time(),
iter / batch_per_epoch)
# save model # ta
self.save(iter)
self.logger.write(disp_str)
sys.stdout.write(disp_str)
sys.stdout.write(time_str) # ta
sys.stdout.flush()
iter += 1
self.iter_cnt += 1
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='model_tester.py')
parser.add_argument('-suffix',
default='run0',
type=str,
help="Suffix added to the save images.")
parser.add_argument('-r',
default='',
type=str,
help="Suffix added to the save images.")
args = parser.parse_args()
tester = Tester(config.cifar_config(), args)
tester.test()
iter / batch_per_epoch)
# save model # ta
self.save(iter)
self.logger.write(disp_str)
sys.stdout.write(disp_str)
sys.stdout.write(time_str) # ta
sys.stdout.flush()
iter += 1
self.iter_cnt += 1
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='cifar_trainer.py')
cc = config.cifar_config()
parser.add_argument('-suffix',
default='run0',
type=str,
help="Suffix added to the save images.")
parser.add_argument('-r',
default='',
type=str,
help="Suffix added to the save images.")
parser.add_argument('-max_epochs',
default=cc.max_epochs,
type=int,
help="max epoches")
parser.add_argument('-ld',
'--size_labeled_data',
default=cc.size_labeled_data,
disp_str = '#{}-{}\ttrain: {:.4f}, {:.2f}% | dev: {:.4f}, {:.2f}%'.format(
int(epoch), iter, train_loss, train_accuracy * 100,
dev_loss, dev_accuracy * 100)
for k, v in monitor.items():
disp_str += ' | {}: {:.4f}'.format(k,
v / config.eval_period)
disp_str += '\n'
monitor = OrderedDict()
self.logger.write(disp_str)
self.logger.flush()
sys.stdout.write(disp_str)
sys.stdout.flush()
iter += 1
self.iter_cnt += 1
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='cifar_trainer.py')
parser.add_argument('--suffix',
default='run0',
type=str,
help="Suffix added to the save images.")
args = parser.parse_args()
trainer = Trainer(config.cifar_config(), args)
trainer.train()
