Dear all.
When i trained my images with Resnet model, my loss in both val and train was updated while my train, val accuracy did not, they stay in the same number. What happens, please help me?
Thank you a lot.
Finished training epoch 0: Train loss: 0.6814049482345581, Train Acc: 59.35077667236328. Val loss: 0.7146084308624268, Val Acc: 60.75581359863281
Finished training epoch 1: Train loss: 0.6787809729576111, Train Acc: 59.35077667236328. Val loss: 0.7859861254692078, Val Acc: 60.75581359863281
...
Here is my code:
#Create model
model = resnet50(
num_seg_classes=2,
no_cuda=False)
model = nn.DataParallel(model, device_ids=[0,1,2,3]).cuda()
net_dict = model.state_dict()
pretrain_path = os.path.join("H:\MinhBPL\ADNIDatabase\MedicalNet_pytorch_files2",
"resnet_50.pth")
print ('loading pretrained model {}'.format(pretrain_path))
pretrain = torch.load(pretrain_path)
pretrain_dict = {k: v for k, v in pretrain['state_dict'].items() if k in net_dict.keys()}
net_dict.update(pretrain_dict)
model.load_state_dict(net_dict)
class AverageMeter(object):
"""Computes and stores the average and current value
Imported from https://github.com/pytorch/examples/blob/master/imagenet/main.py#L247-L262
"""
def __init__(self):
self.reset()
def reset(self):
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def update(self, val, n=1):
self.val = val
self.sum += val * n
self.count += n
self.avg = self.sum / self.count
def accuracy(output, target, topk=(1,)):
"""Computes the precision@k for the specified values of k"""
maxk = max(topk)
batch_size = target.size(0)
_, pred = output.topk(maxk, 1, True, True)
pred = pred.t()
correct = pred.eq(target.view(1, -1).cuda().expand_as(pred))
res = []
for k in topk:
correct_k = correct[:k].view(-1).float().sum(0)
res.append(correct_k.mul_(100.0 / batch_size))
return res
def train(trainloader, model, criterion, optimizer, epoch, use_cuda):
model.train()
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top = AverageMeter()
end = time.time()
print(f'Start training epoch {epoch}')
for (inputs, targets) in tqdm(trainloader):
data_time.update(time.time() - end)
if use_cuda:
inputs, targets = inputs.cuda(), targets.cuda()
inputs, targets = torch.autograd.Variable(inputs), torch.autograd.Variable(targets)
outputs = model(inputs)
loss = criterion(outputs, targets)
prec= accuracy(outputs.data, targets.data, topk=(1,))
losses.update(loss, inputs.size(0))
top.update(prec[0], inputs.size(0))
optimizer.zero_grad()
loss.backward()
optimizer.step()
batch_time.update(time.time() - end)
end = time.time()
print(f'Finished training epoch {epoch}: Train loss: {losses.avg}, Train Acc: {top.avg}, Batch time: {batch_time} ')
return (losses.avg, top.avg)
def test(testloader, model, criterion, epoch, use_cuda):
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top = AverageMeter()
model.eval()
end = time.time()
print(f'Start val epoch {epoch}')
with torch.no_grad():
for (inputs, targets) in tqdm(testloader):
data_time.update(time.time() - end)
if use_cuda:
inputs, targets = inputs.cuda(), targets.cuda()
inputs, targets = torch.autograd.Variable(inputs, volatile=True), torch.autograd.Variable(targets)
outputs = model(inputs)
loss = criterion(outputs, targets)
prec= accuracy(outputs.data, targets.data, topk=(1,))
losses.update(loss, inputs.size(0))
top.update(prec[0], inputs.size(0))
batch_time.update(time.time() - end)
end = time.time()
print(f'Finished training epoch {epoch}: Val loss: {losses.avg}, Val Acc: {top.avg}, Batch time: {batch_time} ')
return (losses.avg, top.avg)
use_cuda = True
graph_loss = {}
graph_acc = {}
graph_loss['train'] = []
graph_loss['val'] = []
graph_acc['train'] = []
graph_acc['val'] = []
best_acc = 0
for epoch in range(100):
optimizer = torch.optim.SGD(model.parameters(), lr=0.00001)
criterion = nn.CrossEntropyLoss()
train_loss, train_acc = train(train_dataloader, model, criterion, optimizer, epoch, use_cuda)
test_loss, test_acc = test(val_dataloader, model, criterion, epoch, use_cuda)
graph_loss['train'].append(train_loss)
graph_loss['val'].append(test_loss)
graph_acc['val'].append(train_acc)
graph_acc['val'].append(test_acc)
is_best = test_acc > best_acc
if is_best:
best_acc = max(test_acc, best_acc)
saveModel('savedModel')