def validate(valid_loader, model, criterion, epoch):
start = time.strftime("%H:%M:%S")
print("\n")
print(f"Starting epoch: {epoch} | phase: Valid | ⏰: {start}")
batch_time = AverageMeter()
losses = AverageMeter()
dices = AverageMeter()
# switch to evaluate mode
model.eval()
end = time.time()
for it, iter_data in enumerate(valid_loader, 0):
images, masks = iter_data
images, masks = images.to(device), masks.to(device)
outputs = model(images)
#if lossfnc != None:
# loss = criterion(outputs, masks, epoch=epoch)
#else:
loss = criterion(outputs, masks)
probs = F.sigmoid(outputs)
dice = metric(probs, masks)
losses.update(loss.item())
dices.update(dice.item())
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
return losses.avg, dices.avg
def train(train_loader, model, criterion, optimizer, epoch, lr=1e-5):
start = time.strftime("%H:%M:%S")
print("\n")
print(f"Starting epoch: {epoch} | phase: Train | ⏰: {start}")
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
dices = AverageMeter()
# switch to train mode
model.train()
num_its = len(train_loader)
end = time.time()
iter = 0
print_freq = 1
# start = time.strftime("%H:%M:%S")
# print(f"Starting epoch: {epoch} | phase: {phase} | ⏰: {start}")
for iter, iter_data in enumerate(train_loader, 0):
# measure data loading time
data_time.update(time.time() - end)
images, masks = iter_data
images, masks = images.to(device), masks.to(device)
outputs = model(images)
#if lossfnc != None:
# loss = criterion(outputs, masks, epoch=epoch)
#else:
loss = criterion(outputs, masks)
losses.update(loss.item())
optimizer.zero_grad()
loss.backward()
optimizer.step()
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
probs = F.sigmoid(outputs)
dice = metric(probs, masks)
dices.update(dice.item())
if (iter + 1) % print_freq == 0 or iter == 0 or (iter + 1) == num_its:
print('\r%5.1f %5d %0.6f | %0.4f %0.4f | ... ' % \
(epoch - 1 + (iter + 1) / num_its, iter + 1, lr, losses.avg, dices.avg), \
end='', flush=True)
return iter, losses.avg, dices.avg
def validate(valid_loader, model, criterion, epoch):
start = time.strftime("%H:%M:%S")
print("\n")
print(f"Starting epoch: {epoch} | phase: Valid | ⏰: {start}")
batch_time = AverageMeter()
losses = AverageMeter()
dices = AverageMeter()
# switch to evaluate mode
model.eval()
end = time.time()
num_its = len(valid_loader)
print_freq = 1
for iter, iter_data in enumerate(valid_loader, 0):
images, masks = iter_data
images, masks = images.to(device), masks.to(device)
outputs = model(images)
if lossfnc == "SymmetricLovaszLoss":
loss = criterion(outputs, masks, epoch=epoch)
else:
loss = criterion(outputs, masks)
probs = F.sigmoid(outputs)
dice = metric(probs, masks)
losses.update(loss.item())
dices.update(dice.item())
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if (iter + 1) % print_freq == 0 or iter == 0 or (iter + 1) == num_its:
print('\r%5.1f %5d | %0.4f %0.4f | ... ' % \
(epoch - 1 + (iter + 1) / num_its, iter + 1, losses.avg, dices.avg), \
end='', flush=True)
return losses.avg, dices.avg