Skip to content

update the code to torch1.7 and python 3.8 #10

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
LLLL-Hui wants to merge 1 commit into
base: master
Choose a base branch
from
Open

update the code to torch1.7 and python 3.8 #10

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions data/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,7 +99,7 @@ def multiclass_noisify(y, P, random_state=0):
""" Flip classes according to transition probability matrix T.
It expects a number between 0 and the number of classes - 1.
"""
print np.max(y), P.shape[0]
print( np.max(y), P.shape[0])
assert P.shape[0] == P.shape[1]
assert np.max(y) < P.shape[0]

Expand All @@ -108,7 +108,7 @@ def multiclass_noisify(y, P, random_state=0):
assert (P >= 0.0).all()

m = y.shape[0]
print m
print( m)
new_y = y.copy()
flipper = np.random.RandomState(random_state)

Expand Down Expand Up @@ -142,7 +142,7 @@ def noisify_pairflip(y_train, noise, random_state=None, nb_classes=10):
assert actual_noise > 0.0
print('Actual noise %.2f' % actual_noise)
y_train = y_train_noisy
print P
print( P)

return y_train, actual_noise

Expand All @@ -167,7 +167,7 @@ def noisify_multiclass_symmetric(y_train, noise, random_state=None, nb_classes=1
assert actual_noise > 0.0
print('Actual noise %.2f' % actual_noise)
y_train = y_train_noisy
print P
print( P)

return y_train, actual_noise

Expand Down
4 changes: 2 additions & 2 deletions loss.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,11 +7,11 @@
# Loss functions
def loss_coteaching(y_1, y_2, t, forget_rate, ind, noise_or_not):
loss_1 = F.cross_entropy(y_1, t, reduce = False)
ind_1_sorted = np.argsort(loss_1.data).cuda()
ind_1_sorted = np.argsort(loss_1.data.cpu())
loss_1_sorted = loss_1[ind_1_sorted]

loss_2 = F.cross_entropy(y_2, t, reduce = False)
ind_2_sorted = np.argsort(loss_2.data).cuda()
ind_2_sorted = np.argsort(loss_2.data.cpu())
loss_2_sorted = loss_2[ind_2_sorted]

remember_rate = 1 - forget_rate
Expand Down
21 changes: 11 additions & 10 deletions main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,6 @@
# Seed
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)

# Hyper Parameters
batch_size = 128
learning_rate = args.lr
Expand Down Expand Up @@ -159,13 +158,13 @@ def accuracy(logit, target, topk=(1,)):

res = []
for k in topk:
correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
correct_k = correct[:k].contiguous().view(-1).float().sum(0, keepdim=True)
res.append(correct_k.mul_(100.0 / batch_size))
return res

# Train the Model
def train(train_loader,epoch, model1, optimizer1, model2, optimizer2):
print 'Training %s...' % model_str
print( 'Training %s...' % model_str)
pure_ratio_list=[]
pure_ratio_1_list=[]
pure_ratio_2_list=[]
Expand Down Expand Up @@ -205,15 +204,16 @@ def train(train_loader,epoch, model1, optimizer1, model2, optimizer2):
optimizer2.step()
if (i+1) % args.print_freq == 0:
print ('Epoch [%d/%d], Iter [%d/%d] Training Accuracy1: %.4F, Training Accuracy2: %.4f, Loss1: %.4f, Loss2: %.4f, Pure Ratio1: %.4f, Pure Ratio2 %.4f'
%(epoch+1, args.n_epoch, i+1, len(train_dataset)//batch_size, prec1, prec2, loss_1.data[0], loss_2.data[0], np.sum(pure_ratio_1_list)/len(pure_ratio_1_list), np.sum(pure_ratio_2_list)/len(pure_ratio_2_list)))
%(epoch+1, args.n_epoch, i+1, len(train_dataset)//batch_size, prec1, prec2, loss_1.data, loss_2.data, np.sum(pure_ratio_1_list)/len(pure_ratio_1_list), np.sum(pure_ratio_2_list)/len(pure_ratio_2_list)))

train_acc1=float(train_correct)/float(train_total)
train_acc2=float(train_correct2)/float(train_total2)
return train_acc1, train_acc2, pure_ratio_1_list, pure_ratio_2_list

# Evaluate the Model
def evaluate(test_loader, model1, model2):
print 'Evaluating %s...' % model_str
print('Evaluating %s...' % model_str)
del1 = model1.cuda()
model1.eval() # Change model to 'eval' mode.
correct1 = 0
total1 = 0
Expand All @@ -225,7 +225,8 @@ def evaluate(test_loader, model1, model2):
total1 += labels.size(0)
correct1 += (pred1.cpu() == labels).sum()

model2.eval() # Change model to 'eval' mode
model2 = model2.cuda()
model2.eval() # Change model to 'eval' mode
correct2 = 0
total2 = 0
for images, labels, _ in test_loader:
Expand All @@ -243,7 +244,7 @@ def evaluate(test_loader, model1, model2):

def main():
# Data Loader (Input Pipeline)
print 'loading dataset...'
print('loading dataset...')
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
num_workers=args.num_workers,
Expand All @@ -256,15 +257,15 @@ def main():
drop_last=True,
shuffle=False)
# Define models
print 'building model...'
print( 'building model...')
cnn1 = CNN(input_channel=input_channel, n_outputs=num_classes)
cnn1.cuda()
print cnn1.parameters
print( cnn1.parameters)
optimizer1 = torch.optim.Adam(cnn1.parameters(), lr=learning_rate)

cnn2 = CNN(input_channel=input_channel, n_outputs=num_classes)
cnn2.cuda()
print cnn2.parameters
print( cnn2.parameters)
optimizer2 = torch.optim.Adam(cnn2.parameters(), lr=learning_rate)

mean_pure_ratio1=0
Expand Down