def test(epoch):
global best_acc, current_best_epoch
#pdb.set_trace()
net.eval()
with torch.no_grad():
all_embeddings = np.zeros((len(test_loader.dataset), 84))
all_targets = np.zeros(len(test_loader.dataset))
for batch_idx, (inputs, targets) in enumerate(test_loader):
inputs, targets = inputs.to(device), targets.to(device)
_, embeddings = net(inputs)
all_embeddings[batch_idx *
args.batch_size:batch_idx * args.batch_size +
embeddings.shape[0], :] = embeddings.cpu().numpy()
all_targets[batch_idx *
args.batch_size:batch_idx * args.batch_size +
targets.shape[0]] = targets.cpu().numpy()
progress_bar(batch_idx, len(test_loader))
# Save checkpoint.
acc = KNNEvaluation(all_embeddings, all_targets)
#pdb.set_trace()
if acc > best_acc:
print('Saving..')
state = {
'net': net.state_dict(),
'acc': acc,
'epoch': epoch,
}
if not os.path.isdir('checkpoint'):
os.mkdir('checkpoint')
torch.save(state, './checkpoint/lifted_ckpt.t7')
best_acc = acc
current_best_epoch = epoch
print('1NN accuracy on test set:%.3f | current best epoch: %d' %
(acc, current_best_epoch))
def train(epoch):
print('\nEpoch: %d' % epoch)
net.train()
train_loss = 0
for batch_idx, (inputs, targets) in enumerate(train_loader):
inputs, targets = inputs.to(device), targets.to(device)
optimizer.zero_grad()
_,embeddings = net(inputs)
loss = criterion(embeddings, targets)
loss.backward()
optimizer.step()
train_loss += loss.item()
progress_bar(batch_idx, len(train_loader), 'Loss: %.3f ' % (1.0*train_loss/(batch_idx+1)))
def train(epoch):
print('\nEpoch: %d' % epoch)
net.train()
train_loss = 0
correct = 0
total = 0
for batch_idx, (inputs, targets) in enumerate(trainloader):
inputs, targets = inputs.to(device), targets.to(device)
optimizer.zero_grad()
outputs = net(inputs)
loss = criterion(outputs, targets)
loss.backward()
optimizer.step()
train_loss += loss.item()
_, predicted = outputs.max(1)
total += targets.size(0)
correct += predicted.eq(targets).sum().item()
progress_bar(
batch_idx, len(trainloader), 'Loss: %.3f | Acc: %.3f%% (%d/%d)' %
(train_loss /
(batch_idx + 1), 100. * correct / total, correct, total))
def test(epoch):
global best_acc
net.eval()
test_loss = 0
correct = 0
total = 0
with torch.no_grad():
for batch_idx, (inputs, targets) in enumerate(testloader):
inputs, targets = inputs.to(device), targets.to(device)
outputs = net(inputs)
loss = criterion(outputs, targets)
test_loss += loss.item()
_, predicted = outputs.max(1)
total += targets.size(0)
correct += predicted.eq(targets).sum().item()
progress_bar(
batch_idx, len(testloader),
'Loss: %.3f | Acc: %.3f%% (%d/%d)' %
(test_loss /
(batch_idx + 1), 100. * correct / total, correct, total))
# Save checkpoint.
acc = 100. * correct / total
if acc > best_acc:
print('Saving..')
state = {
'net': net.state_dict(),
'acc': acc,
'epoch': epoch,
}
if not os.path.isdir('checkpoint'):
os.mkdir('checkpoint')
torch.save(state, './checkpoint/classification_ckpt.t7')
best_acc = acc
device = 'cuda' if torch.cuda.is_available() else 'cpu'
net = ResNet18()
net = net.to(device)
if device == 'cuda':
net = torch.nn.DataParallel(net)
cudnn.benchmark = True
# fileList=file_name('checkpoint','.t7')
fileList = ['checkpoint/Fisher2_ckpt.t7']
for ckpt_file in fileList:
print('==> Resuming from checkpoint ' + ckpt_file)
assert os.path.isdir(
'checkpoint'), 'Error: no checkpoint directory found!'
checkpoint = torch.load(ckpt_file)
net.load_state_dict(checkpoint['net'])
best_acc = checkpoint['acc']
net.eval()
with torch.no_grad():
all_embeddings = np.zeros((len(test_loader.dataset), 128))
all_targets = np.zeros(len(test_loader.dataset))
for batch_idx, (inputs, targets) in enumerate(test_loader):
inputs, targets = inputs.to(device), targets.to(device)
_, embeddings = net(inputs)
all_embeddings[
batch_idx * bs:batch_idx * bs +
embeddings.shape[0], :] = embeddings.cpu().numpy()
all_targets[batch_idx * bs:batch_idx * bs +
targets.shape[0]] = targets.cpu().numpy()
progress_bar(batch_idx, len(test_loader))
evaluate(ckpt_file, all_embeddings, all_targets)
