def main():
args = argParser()
cifarLoader = CifarLoader(args)
net = args.model()
print('The log is recorded in ')
print(net.logFile.name)
if os.path.isfile('model.pth'):
net.load_state_dict(torch.load('model.pth'))
criterion = net.criterion()
optimizer = net.optimizer()
for epoch in range(args.epochs): # loop over the dataset multiple times
net.adjust_learning_rate(optimizer, epoch, args)
train(net, cifarLoader, optimizer, criterion, epoch)
if epoch % 10 == 0: # Comment out this part if you want a faster training
test(net, cifarLoader, 'Train')
test(net, cifarLoader, 'Test')
torch.save(net.state_dict(), 'model.pth')
print('The log is recorded in ')
print(net.logFile.name)
def main():
args = argParser()
cifarLoader = CifarLoader(args)
if not os.path.exists(args.logdir):
os.makedirs(args.logdir)
device = torch.device("cuda" if args.cuda else "cpu")
net = args.model(args.logdir, device).to(device)
print('The log is recorded in ')
print(net.logFile.name)
criterion = net.criterion().to(device)
optimizer = net.optimizer()
for epoch in trange(args.epochs): # loop over the dataset multiple times
net.adjust_learning_rate(optimizer, epoch, args)
train(net, cifarLoader, optimizer, criterion, epoch, device)
if epoch % 10 == 0: # Comment out this part if you want a faster training
test(net, cifarLoader, device, 'Train')
test(net, cifarLoader, device, 'Test')
print('The log is recorded in ')
print(net.logFile.name)
def main():
args = argParser()
cifarLoader = CifarLoader(args)
net = args.model()
net.to(device)
print('The log is recorded in ')
print(net.logFile.name)
criterion = net.criterion()
optimizer = net.optimizer()
for epoch in range(args.epochs): # loop over the dataset multiple times
net.adjust_learning_rate(optimizer, epoch, args)
train(net, cifarLoader, optimizer, criterion, epoch)
if epoch % 1 == 0: # Comment out this part if you want a faster training
test(net, cifarLoader, 'Train')
test(net, cifarLoader, 'Test')
print('The log is recorded in ')
print(net.logFile.name)
def main():
# cuda = torch.device('cuda')
# torch.set_default_tensor_type('torch.cuda.FloatTensor')
args = argParser()
cifarLoader = CifarLoader(args)
net = args.model()
# net.cuda()
print('The log is recorded in ')
print(net.logFile.name)
criterion = net.criterion()
optimizer = net.optimizer()
for epoch in range(args.epochs): # loop over the dataset multiple times
net.adjust_learning_rate(optimizer, epoch, args)
train(net, cifarLoader, optimizer, criterion, epoch)
if epoch % 1 == 0: # Comment out this part if you want a faster training
test(net, cifarLoader, 'Train')
test(net, cifarLoader, 'Test')
print('The log is recorded in ')
print(net.logFile.name)
def main():
args = argParser()
use_cuda = args.cuda and torch.cuda.is_available()
device = torch.device('cuda' if use_cuda else 'cpu')
cifarLoader = CifarLoader(args)
net = args.model().to(device)
print('The log is recorded in ')
print(net.logFile.name)
criterion = net.criterion(args.loss)
optimizer = net.optimizer()
for epoch in range(args.epochs): # loop over the dataset multiple times
net.adjust_learning_rate(optimizer, epoch, args)
train(net, cifarLoader, device, optimizer, criterion, epoch)
if epoch % 1 == 0: # Comment out this part if you want a faster training
test(net, cifarLoader, device, 'Train')
test(net, cifarLoader, device, 'Test')
print('The log is recorded in ')
print(net.logFile.name)